diva portalsu.diva-portal.org/smash/get/diva2:1115208/fulltext01.pdfeast and southeast asia by...

STATES OF DEVELOPMENT

Nathaniel Lane

States of Development

Essays on the Political Economy of Development in Asia

Nathaniel Lane

©Nathaniel Lane, Stockholm University 2017 ISBN print 978-91-7649-878-1ISBN PDF 978-91-7649-879-8ISSN 0346-6892 Cover Picture: President Ferdinand Marcos and his wife Imelda meet with President Lyndon B.Johnson in Manila. October, 1966. White House Photograph Office. Printed in Sweden by Universitetsservice US-AB, Stockholm 2017Distributor: Institute for International Economic Studies

To my grandparents,John and June.

Acknowledgments

I am deeply indebted to my supervisors Torsten Persson and David Strömberg.

As my main advisor, Torsten has been a wellspring of support and encourage-

ment. With him, I have found an unwavering mentor who inspired me to ask big

questions. Since my first year, I have been astonished by the commitment and

graciousness Torsten has shown his students. It is truly inspirational. Similarly,

I am enormously appreciative of David Strömberg, who has been a tremendous

advisor. His door has been constantly open, and he has sat patiently through

many incoherent ideas and ramblings. Honestly, I could not have done any of

this without their tireless advocacy. It was their research that inspired me to

switch into graduate work in economics, and I am lucky to have been their

student.

On the other side of the Atlantic, James Robinson and Melissa Dell have

been constant sources of inspiration and motivation. I met James at a crucial

time for me while studying at Harvard. I will never be able to repay him for

his mentorship and endless cafe bills. Similarly, meeting Melissa was one of

the most refreshing moments of my PhD. Before her (and Pablo Querubin),

I had not met someone who shared similar research interests— including an

appreciation for the Asian economic experience. Both James and Melissa gave

me tremendous confidence as a researcher.

While not my supervisors, Suresh Naidu, Nathan Nunn, and Pablo Queru-

bin’s input has meant the world to me. Without Suresh, I would have never

become an economist. There is not a single person I have pestered more than

him. Both Nathan and Pablo have been tireless sources of intellectual support.

Pablo shared his time and enthusiasm for political economy, as well as Philip-

pine politics, like no other. I was honored to have worked with him. In Nathan I

have found a fantastic intellectual ally. Without him, I would have never had

the courage to attack my job market paper.

I would have been adrift had it not been for the other PhD students who

have passed through IIES: Mounir Karadja, Pamela Campa, Ruixue Jia, Shuhei

Kitamura, Bei Qin, Thorsten Rogall, and David Seim. My current classmates,

especially Erik Prawitz, Arieda Muco, and Valeri Sokolovski, have been crucial

for my sanity—I could not have made it through this year without you. Since

coming back to IIES, the newer members of the IIES community have been

fantastic friends, officemates, and even temporary roommates: Sirus Dehdari,

Benedetta Lerva, Jaakko Meriläinen, and Matti Mitrunen.

Importantly, this dissertation has been made possible with the indispensable

help of Christina Lönnblad and Ulrika Gålnander. With them, I am deeply

indebted to the members IIES community who have supported me through my

final year(s) at IIES, in particular Jon de Quidt.

I want to acknowledge my community outside of Sweden who, regardless

of distance, have supported this dissertation. I am thankful for the economic

history and political economy students (and beyond) I met in Cambridge: Caitlin

Daniel, Vicky Fouka, Leander Heldring, Sara Lowes, Chris Muller, and Lisa Xu.

They have been my roommates; job market confidants; typo checkers; unofficial

referees; and allies. Ellora Derenoncourt deserves a special place here, as she

has been a reservoir of support and comradery.

Last, I want to thank my close friends and family—from Sweden, to the

Philippines, to the US. In Sweden, I owe my gratitude to Aksel vod Sydow

as well as Rebecca Johansson, Elin Svensson, and Hanna Tiensuu. In the

Philippines, I need to thank Wanda, Winnie, and the Albano clan; Henry and

Luke Shevlin; and Pauline Camille Prieto. My family, of course, occupies a

special place: Ana, John, Mark, Rachel, Rick, the Hardmans, and in particular,

June—you all mean the world to me.

Daytona Beach

August, 2017

Nathaniel Lane

Contents

1 Introduction 1References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7

2 Manufacturing Revolutions 92.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . 9

2.2 Institutional Context . . . . . . . . . . . . . . . . . . . . . . 14

2.3 Theoretical Framework . . . . . . . . . . . . . . . . . . . . . 26

2.4 Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30

2.5 Direct Effects of Industrial Policy . . . . . . . . . . . . . . . 34

2.6 Network Externalities . . . . . . . . . . . . . . . . . . . . . 50

2.7 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . 65

References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69

Tables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89

3 Waiting for the Great Leap Forward 1053.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . 105

3.2 Historical Context and Stylized Facts . . . . . . . . . . . . . . 109

3.3 Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 115

3.4 Empirics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117

3.5 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . 138

References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 140

Tables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 145

4 The Historical State, Local Collective Action, and Economic Devel-opment in Vietnam 1534.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . 153

CONTENTS

4.2 Historical Background . . . . . . . . . . . . . . . . . . . . . 160

4.3 Estimation Framework . . . . . . . . . . . . . . . . . . . . . 167

4.4 Long Run Effects on Economic Prosperity . . . . . . . . . . . 169

4.5 Mechanisms . . . . . . . . . . . . . . . . . . . . . . . . . . . 177

4.6 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . 187

References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 188

Figures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 198

Tables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 200

A Appendix . . . . . . . . . . . . . . . . . . . . . . . . . . . . 212

Sammanfattning 259Referenser . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 265

1. Introduction

This thesis consists of three self-contained essays on the political economics

of development, examining the comparative experiences of three Asian coun-

tries: South Korea, the Philippines, and Vietnam. However, this dissertation

tells a story of two bordering worlds: the comparative historical experiences of

East and Southeast Asia.

The ambition of Asian states is a common thread that runs through each

chapter. This thesis concerns the impacts of what political scientist (and Asian-

ist), James C. Scott termed, “high modernist" aspirations (Scott, 1998). Chapter

2 studies a large-scale industrial policy pursued by South Korea’s under the

autocrat, Park Chung Hee. Chapter 3 studies the impacts of an equally ambi-

tious, but quite distinct, modernization push pursued by Ferdinand Marcos in

the Philippines. Chapter 4 reconciles the distinct developmental trajectories of

East and Southeast Asia by exploring the historic institutions of Vietnam–a

country that straddled the two civilizations. Here, even the earliest modern state

building projects of Asian empires translated into contemporary development

outcomes.

Political scientist, Paul Hutchcroft, has called South Korean and Philippine

dictatorships “reverse images" of one another, not least of which because of the

stark differences in their outcomes (Hutchcroft, 2011). The chapters 2 and 3,

Park’s Yushin Fourth Republic and Marcos’ New Society, were both Western

Cold War allies. In the 1970s, each strongman transitioned their country from

democracy to a dictatorship in the midst of political crisis. In either setting,

consolidation was enabled by economic elites who saw autocracy as the price

of modernization.

The distinct sectoral bias of the policies in chapters 2 and 3 reveal the

elite politics beneath either regime (Kang, 2002). In the case of South Korea,

Park’s regime appealed to the interests of industrial capitalists, whose interests

aligned with the state under the existential threat of communist invasion. The

1

2 Introduction

Figure 1.1: The setting: Asia-Pacific Cold War Allies and U.S. President Lyndon

B. Johnson, 1966

Notes: (L-R:) Prime Minister Nguyen Cao Ky (South Vietnam), Prime Minister Harold Holt

(Australia), President Park Chung Hee (Korea), President Ferdinand Marcos(Philippines), Prime Minister Keith Holyoake (New Zealand), Lt. Gen. Nguyen Van Thieu

(South Vietnam), Prime Minister Thanom Kittikachorn (Thailand), President Lyndon B. Johnson

(United States). Source: (U.S.) National Archives, White House Photo Office Collection,

11/22/1963 - 1/20/1969. Photographed by Frank Wolfe.

3

peculiarities of this setting allowed for South Korea’s ambitious industrial

policies. Meanwhile, Ferdinand Marcos’ green revolution represented a project

aimed at modernizing and benefiting the traditional sources of power in the

Philippines, agrarian landlords.

More importantly, the cases of the Republic of Korea and the Republic of the

Philippines, as well as the tumultuous history of Vietnam, are parables for East

versus Southeast Asia. South Korea’s Heavy Chemical Industry big push was a

poster child for policies pursued across East Asia following WWII. South Korea

patterned itself off of Japan’s Meiji restoration, appealing to their 19th-century

transformation. On the other hand, the green revolution exemplified the dual

dreams of Southeast Asian developmentalists: modernization and assuaging

rural unrest (Cullather, 2004,1). The stakes of these genetic innovations were

high, from Malaysia’s reeling rice bowl (Barker, Herdt, and Rose, 1985) to

an embattled Southern Vietnam (Poppel, 2015). On the other hand, Chapter 4

argues that patterns of local state institutions within Vietnam speak to general

forms of state formation seen across the two regions: the Sinic state in East

Asia and the Indic states that typify Southeast Asia political institutions.

The South Korean growth episode was one of the most dramatic episodes of

post-war development. As Park Chung Hee assumed power in 1961, the country

had the same per capita GDP as Ghana. By the1980s, South Korea underwent

an industrial transformation that took Western nations over a century to achieve

(Nelson and Pack, 1998). Chapter 2, Manufacturing Revolutions: IndustrialPolicy and Networks in South Korea, studies the impact of a major industrial

intervention during this period: South Korea’s Heavy Chemical and Industry

(HCI) drive (1973-1979). The HCI big push was the cornerstone of Park Chung

Hee’s new dictatorship, an attempt to shift the country from an exporter of

plywood, wigs, and footwear, to an economy one day capable of producing

domestic weaponry. This paper uses the historical circumstances around South

Korea’s push, along with newly digitized data, to study the impact of industrial

policy on industrial development.

By studying South Korea’s big push, I make three contributions. First, I

estimate the impact of industrial policy on short-run industrial development

outcomes. I do so by comparing the evolution of targeted and non-targeted

manufacturing industries before and after the policy’s sudden announcement. I

4 Introduction

show the positive effects industrial policies had on output growth, employment,

and labor productivity in treated sectors over non-treated sectors. Second, I

evaluate the spillovers of the intervention, tracing how the policy propagated

through industry linkages. I disentangle the effects through forwards and back-

ward linkages, motivating my results using a simple model of the South Korean

economic network. In doing so, I find that the industrial policy promoted the

growth, entry, and capital accumulation in sectors downstream from treated in-

dustry. On the other hand, this analysis shows that upstream industries with the

strongest direct connections to treated industries contracted, as treated industries

imported competing products.

Finally, I test whether the effects of the drive persisted after the planning

period, both in sectors targeted by policy and those exposed to the policy through

linkages. I find evidence of persistent pecuniary externalities like those posited

by big push development theorists, such as Albert Hirschman (Hirschman,

1958). In other words, I find that South Korea’s controversial industrial policy

was successful in producing industrial development, the benefits of which

persisted through time and in industries not directly targeted by the policies. On

average, the HCI policies led to about 80 percent more growth and promoted

an 11 percent decline in output prices for treated versus non-treated industries.

Together, these results show that the industrial policy promoted South Korea’s

move up the supply chain.

Chapter 3, Waiting for the Great Leap Forward - Green Revolutionand Structural Change in the Philippines, studies a much different type

of intervention, the Philippine’s green revolution. While Korea’s HCI was

done in spite of Western institutions, the green revolution was their brainchild.

A product of Ford and Rockefeller Foundation grants, Robert McNamara’s

World Bank, and the Philippine government, the International Rice Research

Institute (IRRI) was founded in the Laguna province in 1960. The IRRI was

the agronomic research heart of new hybrid rice varieties that would define

the green revolution; political scientists, Lynn T. White referred to it as the

“highest-profile technology research program in the world" (White, 2009, 6). In

1966, the Philippines experienced the widespread introduction of so-called of

the IR-8 “miracle rice" varieties–the first decisive product of the IRRI–marking

the beginning of the revolution.

5

Accordingly, this chapter studies how the green revolution technologies

impacted structural change in their country of origin. Since the birth of the

sub-field, development economists long theorized that rising agricultural pro-

ductivity was the engine behind structural transformation: the reallocation of

economic activity from the agricultural sectors to modern manufacturing and

service sectors (Nurkse, 1953; Rostow, 1960). However, even after the rapid,

momentous rollout of early green revolution technologies across the islands,

modernization did not follow. In sharp contrast to its Asian contemporaries,

the share of manufacturing labor remained constant, and the agricultural sector

remained the dominant source of employment through the 1980s.

Using newly digitized data on the green revolution, I show that growth in

agricultural productivity produced structural change—but in ways not antici-

pated by planners and theoretical models. With a newly constructed panel of

Philippine municipalities, I trace how the expansion of new high yielding vari-

eties, known as HYVs or modern varieties, increased agricultural productivity

and reallocated economic activity across sectors–my measures of structural

change. I focus specifically on how the share of employment in agriculture,

manufacturing, and services changed over the next four decades, immediately

following the arrivals of HYVs in 1966.

I show that green revolutions technological shocks had quite different effects

on short- and long-run structural change, producing particularly unexpected

effects on peasant agricultural labor. I first confirm that after 1966, unlike many

Asian (and current African economies) HYVs were widely adopted across

Philippine townships and were subsequently related to a rapid increase in agri-

cultural productivity. I then show that in the short-run, 1970-1980, the green

revolution translated into labor-absorbing technological change: reallocating

labor into HYV-intensive rice economies. These results are consistent with the

increase in aggregate agricultural employment the decade after the introduc-

tion of modern rice varieties. However, in the long run, 1980-2000, I show

this pattern is reversed; the green revolution translated into labor-displacing

technological change. In particular, agricultural wage labor was dislocated from

agriculture and pushed into low-skilled service employment. I argue that rising

wages and declining prices of capital prompted rice farms to mechanized, and

thus promoted the long-run decline in agricultural employment.

6 Introduction

Chapter 4, The Historical State, Local Collective Action, and EconomicDevelopment in Vietnam, written with Melissa Dell and Pablo Queurubin,

explores the distinct developmental trajectories of East and Southeast Asia. The

efficacy of advanced development strategies depends on the capacity of states to

organize these policies, often at localized levels. However, the ability of states

to deploy ambitious development policies varies widely across Asia, from the

Weberian bureaucracies of East Asian to personalistic networks of Southeast

Asia. The advanced bureaucracies and planning agencies of East Asia have been

credited with the success of the East Asian growth miracle (Evans, 1992). For

example, Japan’s MITI(Johnson, 1982) and South Korea’s Economic Planning

Board (Chapter 2) were credited with implementing advanced industrial policies.

Similar initiatives in Southeast Asian policy succumbed to crony capitalism and

predatory politics–perhaps most represented by Macros’ New Society. Chapter

4 studies the historical role East versus Southeast Asian state institutions played

in the divergent experiences of the two neighboring regions.

Specifically, Chapter 4 examines how the historical state impacts long-run

development, using Vietnam as a laboratory for the East and Southeast Asian

experiences. As argued by a long lineage of historians and anthropologists, early

modern Vietnam represented the dividing line between the two civilizations

(Lieberman, 2003). Northern Vietnam, historically known as Dai Viet, was

ruled by a strong centralized state in which the village was the fundamental ad-

ministrative unit. These institutions were directly adopted from Imperial China,

one of the earliest modern states in the world. On the other hand, Southern

Vietnam was a peripheral tributary of the Khmer (Cambodian) Empire, which

followed a patron-client model with weaker, more personalized power relations

and no village intermediation.

Using a regression discontinuity design across the historic Dai Viet-Khmer

boundary, the study shows that areas historically under a strong state have

higher living standards today and better economic outcomes over the past 150

years. Rich historical data document that in villages with a strong historical

state, citizens have been better able to organize for public goods and redistribu-

tion through civil society and local government. This suggests that the strong

historical state crowded in village-level collective action and that these norms

persisted long after the original state disappeared. We consider the develop-

REFERENCES 7

mental experience of these two worlds to be a larger analogy for the divergent

experiences of East and Southeast Asia.

References

Barker, R., R. W. Herdt, and B. Rose (1985). The Rice Economy of Asia.

Washington, D.C. and Manila: Resources for the Future, Inc./Internaton

Rice Research Institute.

Cullather, N. (2004, apr). Miracles of Modernization: The Green Revolution

and the Apotheosis of Technology. Diplomatic History 28(2), 227–254.

Cullather, N. (2013). The Hungry World: America’s Cold War Battle Against

Poverty in Asia. Cambridge, Massachusetts: Harvard University Press.

Evans, P. B. (1992). The State as Problem and Solution: Predation, Embedded

Autonomy, and Structural Change. In S. Haggard and R. R. Kaufman

(Eds.), The Politics of Economic Adjustment: International Constraints,

Distributive Conflicts and the State, Chapter 3, pp. 139–181. Princeton,

New Jersey: Princeton University Press.

Hirschman, A. O. (1958). The Strategy of Economic Development (Third ed.).

New Haven, Connecticut: Yale University Press.

Hutchcroft, P. (2011). Reflections on a Reverse Image. In The Park Chung Hee

Era, Chapter 19, pp. 542–572. Cambridge, Massachusetts and London,

England: Harvard University Press.

Johnson, C. (1982). Miti and the Japanese Miracle: The Growth of Industrial

Policy : 1925-1975. Stanford, California: Stanford University Press.

Kang, D. C. (2002). Crony Capitalism: Corruption and Development in South

Korea and the Philippines (First ed.). Cambridge Studies in Comparative

Politics. Cambridge, U.K.: Cambridge University Press.

Lieberman, V. (2003). Strange Parallels: Southeast Asia in Global Context, c.

800-1830, Volume I: Integration on the Mainland. Cambridge, U.K and

New York, New York: Cambridge University Press.

Nelson, R. R. and H. Pack (1998). The Asian Miracle and Modern Growth

Theory. The Economic Journal 109(1881), 416–436.

Nurkse, R. (1953). Problems of Capital Formation in Underdeveloped Coun-

tries. Oxford, UK: Oxford University Press.

8 REFERENCES

Poppel, Z. D. (2015). Quick rice: international development and the Green

Revolution in Sierra Leone, 1960-1976. In C. Helstosky (Ed.), The

Routledge History of Food, Chapter 17, pp. 332–354. New York, New

York: Routledge and Taylor & Francis Group.

Rostow, W. W. (1960). The Stages of Economic Growth: A Non-Communist

Manifesto (3rd ed.). New York, New York: Cambridge University Press.

Scott, J. C. (1998). Seeing Like a State: How Certain Schemes to Improve the

Human Condition Have Failed. New Haven, Connecticut: Yale University

Press.

White, L. T. (2009). Political Booms: Local Money and Power in Taiwan, East

China, Thailand, and the Philippines. Series on Contemporary China –

Vol.16. Singapore: World Scientific Publishing.

Wolfe, F. (1966). Manila Conference: SEATO nations leaders group portrait.

2. Manufacturing Revolutions -

Industrial Policy and Networks in

South Korea*

2.1 Introduction

Miracles by nature are mysterious. The forces behind the East Asian growth

miracle are no exception. Industrial policy, however, has defined Asia’s striking

postwar transformation (Rodrik, 1995). The ambitious development strategies

pursued by Korea, Singapore, and Taiwan now shape interventions across the

world, from Southeast Asia to Sub-Saharan Africa (Rodrik, 2005; Robinson,

2010; Lin, 2012). Arguably, industrial policies have since become a ubiquitous

feature of modern economic development; with rare exception, every develop-

ing country has pursued industrial policy. While early development economists

argued these policies were key to structural transformation (Rosenstein-Rodan,

1943; Hirschman, 1958), many others warned of their deleterious consequences

(Baldwin, 1969; Krueger, 1990). Nonetheless, few empirical studies have ex-

plored the effects of industrial policy on development—and none have addressed

their role in Asia’s postwar transformation.

*I thank my advisers, Torsten Persson, Melissa Dell, James Robinson, and David Stromberg. I

would also like to thank Suresh Naidu, Nathan Nunn, and Pablo Querubin for their encouragement

and support. This project has benefited from conversations with Philippe Aghion, Samuel

Bazzi, Roland Benabou, Timo Boppart, Francisco Buera, David Cole, Jon de Quidt, Ellora

Derenoncourt, Arin Dube, Rikard Forslid, Mikhail Golosov, Mounir Karadja, Max Kasy, Danial

Lashkari, Andreas Madestam, Sam Marden, Kurt Mitman, Matti Mitrunen, Chris Muller, Arash

Nekoei, Peter Nilsson, Dwight Perkins, Per Pettersson-Lidbom, Erik Prawitz, Nancy Qian, Martin

Rotemberg, Alex Segura, Jakob Svensson, Eric Verhoogen, Lisa Xu, and participants at the

CIFAR Institutions, Organizations & Growth meeting (2015); CSAE OXDEV Conference (2016);

Harvard Economic History Lunch (2015); Harvard Economic Development Tea (2015); National

Bureau of Economic Research Summer Institute–Development of the American Economy

meeting (2015). This project was made possible with excellent research assistance from BoSuk

Hong. I would also like to thank the staff of the Bank of Korea for data access.

9

10 CHAPTER 2. MANUFACTURING REVOLUTIONS

In 1957, Ghana and South Korea had identical national incomes, and South

Korea entered the 1960s, corrupt, unstable, and dependent on Western aid.1 By

1980, the Republic of Korea had undergone an industrial transformation that

had taken Western nations over a century to achieve (Nelson and Pack, 1998).

How did South Korea evolve from an impoverished, agrarian economy

into a modern industrial power? This paper explores Korea’s use of industrial

policy: interventions intended to shift a nation’s industrial composition to one

more favorable for growth than if the economy evolved according to static

comparative advantage [Lindbeck (1981); Chang (2003); Noland and Pack

(2003); p.10].

I consider a definitive postwar policy, South Korea’s Heavy Chemical and

Industry (HCI) drive, 1973-1979. HCI embodied the big push-style policies

imagined by development scholars, such as Rosenstein-Rodan (1943), Nurkse

(1953) and Hirschman (1958). Moreover, HCI was an infant industry policy: a

temporary (six year) intervention meant to incubate Korea’s strategic industries.

Korea’s drive was broadly representative of industrial policies used across East

Asian economies—and beyond (Vogel, 1991; Young, 1992). Korea copied their

policy from Japan, while contemporaries, such as Taiwan, pursued comparable

strategies (Cheng, 1990; Cheng, 2001). Meanwhile, Korea’s big push inspired

similar interventions in countries like Algeria, Brazil, Malaysia, and Philippines

(Kim et al., 2013; Moreira, 1994; Lall, 1995; Lall, 1996). The mixed results of

these policies have only made Korea’s big push more controversial [Kim and

Leipziger (1993); p.24].

In studying the consequences of South Korea’s big push, I make three con-

tributions. First, I estimate the effect of industrial policy on short-run industrial

development outcomes. I do so by comparing the evolution of targeted and non-

targeted manufacturing industries before and after the policy’s announcement.2

Second, I evaluate the spillovers of the intervention, tracing how the policy

propagates through the input-output network. I disentangle the effects through

forwards and backward linkages, motivating my results using a multi-sector

general equilibrium model. Finally, I test whether the effects of the drive per-

1I refer to per capita GDP (Werlin, 1991). According to the Penn World Tables, in 1960 South

Korea’s per capita national income lagged behind Cameroon, Central African Republic, Haiti,

Madagascar, Morocco, Niger, and Tanzania (Feenstra et al., 2015).2I use the terms sector and industry interchangeably in this study.

2.1. INTRODUCTION 11

sisted after the planning period, both in sectors directly targeted by policy and

in those exposed to the policy through linkages.

For the purpose of this study I construct a rich industrial dataset, combining

digitized material from archival sources with vintage machine-readable data. I

harmonize this panel with network measures from reconstructed input-output

accounts and rare trade policy data. The result is an extensive dataset spanning

South Korea’s big push episode.

External politics drove the big push in 1973 and its demise in 1979. Presi-

dent Richard Nixon’s sudden withdrawal of U.S. forces from Asia (the so-called

Nixon Doctrine) had thrown Eastern allies into a security crisis. Since World

War II, South Korea relied on the U.S. to maintain military balance against

the North. With U.S. support in doubt, the South was forced to develop their

own military-industrial capacity. Strictly speaking, the U.S. pullout prompted

a big push by executive decree, shifting the country’s policy regime from a

general export promotion strategy to one promoting a limited set of strategic

industries. Key sectors were selected based on military importance and copied

from Japan’s earlier industrial strategy (Stern et al., 1995; B.-k. Kim, 2011). Just

six years after its announcement, however, the big push died with its general:

President Park’s 1979 assassination signaled a de facto end to his cornerstone

project.

The historic context of South Korea’s big push allows me to avoid prominent

sources of bias that plague studies of industrial policy. The political nature of

industrial policy means interventions are often allocated based on elite patronage

and special interest politics rather than economic rationality. For instance,

subsidies and tariffs regularly go to declining, or “sunset,” sectors, and in the

developing world, cronyism steers resources towards projects that defy latent

comparative advantage (Harrison, 1994; Rodrik, 2005). Accordingly, empirical

studies often reveal a negative relationship between industrial policies and

industrial growth. By contrast, I argue that the big push was implemented under

the duress of a security crisis that made rational implementation paramount.

Also, I maintain HCI planning selected projects for which Korea possessed a

latent comparative advantage.

Korea’s setting suggests an intuitive estimation strategy. I compare changes

in industrial outcomes between targeted and non-targeted manufacturing indus-


tries for each year before and after the big push announcement. This flexible

differences-in-differences strategy uncovers the effect of interventions aimed at

promoting sectors in which it has latent comparative advantage. Pre-trends rep-

resent a counterfactual sectoral structure; absent HCI interventions, industries

would have evolved according to their pre-1973 specialization, or static compar-

ative advantage. The post-1973 differences reflect the efficacy of interventions—

investment subsidies and trade policy—aimed at allocating resources toward

sectors which South Korea had unrealized potential in, or latent comparative

advantage.

My preferred estimates show the big push significantly shifted economic ac-

tivity to capital-intensive industry, a shift which continued after the interventions

were retrenched. During and after the HCI-period (1973–1979), targeted sec-

tors grow significantly more than non-targeted sectors relative to pre-treatment

levels. The results are robust to various measures of growth and indicators of

industrial development. Importantly, I find evidence of significant improve-

ments in productivity during and after the big push, as shown by measures of

factor productivity, exports, and, importantly, output prices. Market entry and

employment also increase.

A key argument for industrial policy, however, is that benefits accrue to

industries outside of targeted sectors (Hirschman, 1958; Hirschman, 1968; Pack

and Westphal, 1986; Grossman, 1990). To see whether this was the case, I

estimate the network spillovers of policy by comparing the evolution of non-

targeted industries with weak linkages and those with strong linkages to targeted

sectors. I find HCI policies positively impacted forward-linked (downstream)

industry but negatively impacted backward-linked (upstream) industry. Results

suggest industrial policy surprisingly lowered the prices for downstream buyers.

On the other hand, HCI trade policies allowed targeted industries to import

intermediates and subjected upstream suppliers to import competition. Thus, I

provide new evidence that industrial policy generates pecuniary externalities,

but in ways not fully anticipated by classic developmental theory.

My study speaks to an unresolved debate on the role of industrial policy in

economic development. On one side of the debate, an influential descriptive

literature has emphasized the role of state institutions and industrial interven-

tions in postwar industrialization (including Johnson, 1982; Wade, 1990; Vogel,


1991; Amsden, 1992; Evans, 1995; Chibber, 2002; Kohli, 2004). This litera-

ture highlights the centrality of industrial policy in East Asia’s transformation.

Robert Wade (1990) and Alice Amsden (1992), in particular, emphasize that

the big push interventions were essential to Korea’s miracle.

Conversely, a large literature in economics criticizes the role industrial

policy in economic development (Baldwin, 1969; Krueger and Tuncer, 1982;

Lal, 1983; Noland and Pack, 2003). These criticisms are met with little empirical

literature on the effect of industrial policy on structural change (Herrendorf et al.,

2013).3 Accordingly, many doubt the role of these interventions in postwar East

Asia (Weinstein, 1995; Beason and Weinstein, 1996; Lawrence and Weinstein,

1999). An influential critique of postwar policies is that NICs would have grown

more in their absence (Krueger, 1995). Yoo (1990) argues this was the case

for HCI in Korea, and Lee (1996) shows evidence that policies may have been

detrimental to the industrial development of targeted sectors. (Noland, 2004)

further contends that HCI failed to target “leading industries.”

I provide one of the first econometric studies of East Asian industrial policy,

adding econometric credence to the arguments made by Robert Wade (1990)

and Alice Amsden (1992)—with important caveats for a small open economy.

In doing so, I contribute to a nascent literature on industrial policy, including

Nunn and Trefler (2010), Criscuolo et al. (2012), Aghion et al. (2015), as well

as Juhasz (2016) and Rotemberg (2015), who study the impact of industrial

policy in a development context.

My study also contributes to the literature on network economics. It draws

directly on original theories of industrialization and linkages emphasized by

Scitovsky (1954), Rasmussen (1956), Myrdal (1957), Chenery and Watanabe

(1958), and Hirschman (1958). Ciccone (2002), Jones (2008), and Jones (2013)

formalize these theories, showing that key sectors can influence aggregate

growth through input-output linkages. Similarly, (Long Jr and Plosser, 1983;

Carvalho, 2010; Acemoglu et al., 2012; Atalay, 2015), explore the influence

of sectoral shocks on the business cycle.4 My results on linkages also relate to

a development literature on the intersectoral effects of FDI (Rodriguez-Clare,

3“[T]he empirical evidence on the success of ‘big–push’ policies in particular, and industrial

policies more generally, is mixed at best,” Herrendorf et al. (2013).4Within this literature, Shea (2002), Conley and Dupor (2003), and Holly and Petrella (2012)

highlight the importance played by intersectoral factor-demand linkages.


1996; Markusen and Venables, 1999; Smarzynska Javorcik, 2004) and trade

policy (Succar, 1987; Krugman, 1998; Puga and Venables, 1999; Forslid and

Midelfart, 2005).

Finally, my study contributes to the literature on the role played by state

capacity in economic development (Besley and Persson, 2010; Besley and Pers-

son, 2011; Acemoglu et al., 2015) and the implementation of growth-enhancing

policies (Dell et al., 2016).5 Industrial policy is state action, and thus intimately

tied to the quality of government (Rodrik, 1997). Successful interventions re-

quire specific bureaucratic capabilities (Johnson, 1982; Evans, 1995; Fukuyama,

2014) and also require political incentive compatibility (Haggard, 1990; Chib-

ber, 2002; Robinson, 2010; Vu, 2010). These conditions are rarely satisfied

(Krueger, 1990). Nonetheless, Wade (1990) and Amsden (1992) suggest the

strong institutions of South Korea, Taiwan, and Japan underpinned the suc-

cessful deployment of HCI interventions. State capacity may be a necessary

ingredient for executing proper industrial development strategies and thereby

fostering economic development.

The remainder of the paper is organized as follows. Section 2.2 discusses

the historical and institutional setting of the HCI big push. Section 2.3 outlines

the effects of the policy using a multi-sector general equilibrium model. Section

2.4 describes my digitized manufacturing dataset for South Korea. Section 2.5

presents estimates of the direct effect of industrial policies on targeted industries.

Section 2.6 reports estimates of how HCI spilled over onto non-targeted sectors

through the input-output network. Finally, Section 2.7 summarizes the results

of my study.

2.2 Institutional Context

2.2.1 Drivers of the Heavy Chemical and Industry Big Push

“The enemy will hesitate to invade only when they realize that we

are equipped with strength and determined to fight to the end” –

5My related work with Melissa Dell and Pablo Querubin (Dell et al., 2016) explores the

historical effect of the Weberian state and its capacity to implement successful policy across

Asia.

2.2. INSTITUTIONAL CONTEXT 15

President Park Chung-hee6

“[Congress] may – as in the case of Vietnam – deny funds and use

of U.S. forces needed to defend Korea and even force U.S. troop

withdrawals . . . Korea’s only alternative is to achieve a degree of

self-reliance that will cushion possible loss of U.S. support before

or during conflict” – U.S. Ambassador Sneider7

This paper focuses on a period of political emergency, during which Presi-

dent Park Chung-hee declared a lifelong dictatorship in late-1972 (the Yushin

Constitution) and launched the Heavy Chemical and Industry Drive (HCI),

1973-1979.

A security crisis drove the South Korea’s heavy industrial big push (Haggard,

1990; Yoo, 1990; Stern et al., 1995; Horikane, 2005; Im, 2011; H.-A. Kim, 2011;

Moon and Jun, 2011).8 Two parallel events were at the heart of this impasse

(Kim, 1997; Kwak, 2003; Moon and Lee, 2009; Kim, 2004).9 First, a sudden

change in U.S. foreign policy towards Asia. Second, the parallel militarization

of North Korea.

In late 1969, facing domestic political pressure from the Vietnam War, Pres-

ident Nixon announced the end of U.S. military support for Asian allies, who

would now be responsible for defense against Communist aggression [Nixon

(1969); p.549]. This “Nixon Doctrine” effectively ended the Vietnam War and

preceded normalized relations with China. South Korea, an anti-Communist

stalwart that had sent 50,000 troops to South Vietnam for U.S. military commit-

ments, was outraged (Kim, 1970; Kwak, 2003).

6Kim, 2004; p.166.7Kim, 2011b; p.31.8There is no ambiguity as to the security pretext for the HCI drive. Yoo (1990), in a Korean

Development Institute report, “one of the main reasons why the government adopted the HCI

policy was the security concern” [Yoo (1990); p.18]. “When President Richard M. Nixon declared

his Guam Doctrine in 1969 to initiate U.S. military disengagement from Asia, Park’s fear of

the Americans’ departure pushed him to initiate an aggressive HCI drive to develop a defense

industry by 1973” [Moon and Jun (2011); p.119]. For a summary of HCI in the context of

building a domestic defense, see (H.-A. Kim, 2011)9Historian, James Palais: “Park was so shocked by what he perceived as the American failure

from the late 1960s to the mid-1970s to respond to North Korean provocations, to stay the course

in Vietnam, and to maintain a solid commitment to the defense of South Korea, that he decided

to institute a more determined policy to achieve the next phase of the industrial revolution by

creating a heavy and chemical industrial sector” [Kim (2004); p.xiv].


Nixon’s political shock introduced the risk of full U.S. troop withdrawal

through the 1970s. The ROK believed that they could be left to defend against a

DPRK blitzkrieg alone. A U.S. congressional subcommittee report summarized

the causal implications: a “consequence of the [troop] withdrawal was the need

for South Korea to improve its defense production capability” and needed to

play “‘catch-up ball’ with the DPRK” [U.S. House. Committee on International

Relations. Subcommittee on International Organizations. (1978); p.74].10 The

ROK feared they would become the next South Vietnam, and the U.S. could

normalize relations with the DPRK (Nam, 1986; Goh, 2004; Ostermann and

Person, 2011).11

0.00

0.01

0.02

0.03

1961

1969

Nixon's

Announcmen

t 1973

HCI 1978

Shar

e of

Art

icle

s Pu

blis

hed

A − Mentions of U.S. Troop Withdrawal from South Korea,Share of Stories in New York Times

0

2000

4000

6000

1961

1969

Nixon's

Announcmen

t 1973

HCI 1978

Tota

l Rec

ord

ed A

ctio

ns

B − Recorded North Korean ActionsAgainst Armistice

Figure 2.1: Political Events Behind the Heavy Chemical and Industry Drive

The U.S. troop withdrawal threat came in two waves. Figure 2.1, Panel A

plots the occurrences of Korean troop withdrawal stories (share of stories) in the

New York Times.12 The first shock corresponds to the spike in stories between

1970–1972. Confirmation of the U.S.’ commitment to the pull-out of ROK came

10Janne E. Nolan (1986) makes the case that the Nixon doctrine promoted similar industrial

reactions in both South Korea as well as in Taiwan, who were similarly threatened by U.S.

detente with China.11Historian Nam Joo-Hong notes that normalized Sino-American relations were a “double

loss” that strategically benefited North Korea [Nam (1986); p.126-128]. South Korean official,

Kim Dasool: “when the U.S. entered into detente with China. . . then it was a definite possibility

that the U.S. government could also enter into detente with North Korea and perhaps even

normalize its relationship with North Korea” [Ostermann and Person (2011); p.15].12Search term: South Korea + Troop Withdrawal .


in 1970 and “profoundly shocked” President Park, who expected exemptions

from the Nixon Doctrine [Rogers (1970); Nixon (1970); Kwak (2003); p.34].

That summer, U.S. Vice President Spiro Agnew unexpectedly announced the

intention of a full troop withdrawal. Immediately after Agnew’s announcement,

Korean and U.S. press first reported that—unbeknownst to Korea—the U.S.

had already scaled down their forces by 10,000 [U.S. House. Committee on

International Relations. Subcommittee on International Organizations. (1978);

p.34; Nam (1986); p.78; Kwak (2003); p.47]. The first wave of true withdrawals

occurred in 1971, when the US pulled 24,000 ground troops and three air force

battalions from the peninsula.

The threat of total U.S. withdrawal persisted through the 1970s, particularly

during the 1976 U.S. presidential campaign.13 As explained by a contempo-

raneous Asian Survey report on South Korean relations: “The Jimmy Carter

phenomenon became a veritable shock for the ROK government” [Oh (1977);

p.71]. Total withdrawal and further reduction of military assistance became a

campaign promise of the Democratic candidate, who denounced Park’s human

rights record and U.S. military support (Taylor et al., 1990).14

For South Korea, the U.S. withdrawal was ill-timed. Figure 2.1, Panel

B plots the steady escalation of “actions again the amnesty treaty” (the post

Korean War treaty) (Choi and Lee, 1989).15 Through the late-1960s, North

Korean launched a steady wave of attacks on the South, inspired by Viet Cong

tactics in Vietnam.16 As indicated by Panel B, through the 1970s the DPRK

stepped up conventional antagonism against the ROK. In late-1971, South

Korean CIA director stated, “[a]t this moment, our front-line is a half step

before crisis. A North Korean attack may come anytime. They are deploying

13“The HCI drive was also largely motivated by national security concerns, magnified bythe Carter administration’s plan to completely withdraw U.S.. [emphasis my own]” [Kim et al.

(1995); p.186].14Immediately after taking office in 1977, Carter reiterated his commitment to withdraw the

remaining U.S. troops [Han (1978); M. Y. Lee (2011); p.428]. However, the instability following

Park’s 1979 assassination meant the U.S. could not carry through with the campaign promise.15Actions against the amnesty treaty include border crossings, military exercises, and other

acts of antagonism.16“Kim Il Sung understood the power of insurgency as a lesson learned from the Vietnam

war” (Scobell and Sanford, 2007). Vietnamese-style tactics culminated in a 1968 surprise attack

on the presidential residence (the Blue House). Another assassination attempt on Park in 1974

would kill the First Lady.


units and tanks much closer to the DMZ” [Kim (2001); p.55]. A few years

later, the fall of Southern Vietnam roused South Korea’s “the worst fears” [Oh

(1976); p.78] and triggering a “near panic situation” in the Republic [Kim and

Im (2001); p.64].

The connection between the military-industrial drives and North Korean ac-

tion is illustrated by March 1974’s “Yulgok Operation,” an emergency measure

that followed DPRK attacks on Paeng’nyong Islands [Kim (2004); p.189]. The

project, which sought to upgrade ROK’s military hardware, coincided with the

establishment of a National Defense Fund, followed by a new National Defense

Tax.

North Korea was militarily and economically superior to South Korea

through the 1970s (Eberstadt, 1999; Noland et al., 2000; Eberstadt, 2007).17

Through the 1970s, the DPRK continued a non-stop military-industrial course

embarked on in 1962 [Hamm (1999); Michishita (2009); p.23]. By the early

1970s, the North had become “the most highly militarized society in the world

today” (Scalapino and Lee, 1972). Taik-young Hamm argued that during the

DPRK’s crash military build-up campaign from 1967-1971, the ROK “did (or

could) not follow suit” [Hamm (1999); p.79].

The U.S. withdrawal threat meant the South would have to militarize to

reach military balance with the North. During the first U.S. withdrawal, the

ROK had relied on dated M-1 rifles and WWII era artillery, and according

to estimates, military stocks could last for three days in the event an invasion

by the DPRK [Stern et al. (1995); p.21-22]. By the late-1970s, even after an

unprecedented military modernization the South, the military advantage lay

with North Korea – especially without U.S. troops (U.S. Senate. Committee on

Foreign Relations, 1978; Cushman, 1979; Choi, 1985; Eberstadt, 1999).18

17The exact growth rate of North Korea is mysterious. Prominent scholars of North Korea

conclude that conservatively North Korean growth dominated the Republic’s by the 1970s and,

at most, even until the 1980s [Eberstadt (2007); p.xi]. Noland et al. state, “the conventional

wisdom is that per capita income in North Korea exceeded that of South Korea well into the

1970s” [Noland et al. (2000); p.1769].18A U.S. Senate report on U.S. military withdrawal summarizes the military balance on

the peninsula in 1978: “[t]he principal advantages for the North today lie in ground weapons

(tanks, artillery, mortars), quantity of fighter aircraft and quantity of naval combat vessels” [U.S.

Senate. Committee on Foreign Relations (1978); p.2]. Lt. Gen. John Cushman concluded that

the Second Infantry would be “essential” to stop North Korea’s “superior forces in a surprise,

Blitzkrieg-Style drive to capture or threaten Seoul” [Cushman (1979); p.361]. Nick Eberstadt


2.2.2 Heavy Chemical and Industry Drive Policy

Programme and Sectoral Choice The HCI drive was announced at a New

Year’s press event, January, 12 1973, and “rapidly turned into an all-out opera-

tion for South Korea’s military modernization” [H.-A. Kim (2011); p.29].19 The

official HCI Plan was the product of executive action and covertly drawn up by

a team of technocrats (Haggard, 1990; p.131; Kim, 1997).20 To further avoid

upsetting domestic capitalist interests, as well as competing bureaucrats, ad-

ministration fell to a superagency, the Heavy Chemical and Industry Promotion

Committee (Lim, 1998; Haggard, 1990).21

Six broad “strategic” sectors were targeted by the policy: steel, non-ferrous

metals, shipbuilding, machinery, electronics, and petrochemicals (Lee, 1991;

Stern et al., 1995).22 Table 2.1 lists all 5-digit industries which fall into HCI

policy.23 Targeted industries were prioritized for ambitious investment and

growth targets and, importantly, they were to achieve a 50 percent share of

exports by the 1980s.24

The choice of HCI sectors can be boiled down to two factors: strategic

concerns and Japan’s historic experience.

echoes that by 1979 the DPRK “probably still enjoyed a military advantage over ROK [South

Korea]” [Eberstadt (1999); p.34].19The HCI Plan is was announced, June 1973. The HCI Plan is often conflated with Korea’s

Third Five Year Economic Development Plan (1972-1976), which the HCI announcement

effectively interrupted (Lee, 1991).20Alongside HCI, a secret defense program, project Yulgok, was carried out to upgrade military

weaponry (Hamm, 1999; Kim, 2004; H.-A. Kim, 2011).21“The powerful role of the planning team minimized bureaucratic conflicts and increased

effective implementation of the HCI Plan” [Lim (1998); p.81]. Planning in South Korea was

routinely used to eliminate poor candidates for industrial projects Adelman (1969).22The term “HCI” is also used to define a specific set of sectors in Korea statistical publications.

In this use of the term, HCI does not encompass the electronics industry. Hence, there is a

distinction between HCI as it is used in statistical publications and it’s specific used in the HCI

policy plans. As Suk-Chae Lee explains, the electronics industry “was one of the core industries

slate for promotion in Korea’s HCI Plan [May, 1973]; therefore any analysis of the HCI plan

should include the electronics industry” [Lee (1992); p.432].23The table lists sectors using names based on the 1970 Bank of Korea sector names, since

they were already translated. The Korea Standard Industry Classification (KSIC) are based on

1970 industry codes. Because of code harmonization through time, the exact number of industries

used in the study is slightly different.24For HCI industries to be sustainable, it was necessary for them to export. Many of HCI

industries required capacities larger than what could be sustained by the limited domestic market

in Korea [Melo and Roland-Holst (1990); p.3-5].


First, HCI sectors were required for military-industrial modernization, as

South Korea prepared for a future without U.S. assistance. It was clear to plan-

ners that heavy industry was necessary for future defense production. According

to Yumi Horikane, earlier attempts at arms manufacturing failed due to lack

of domestic input infrastructure: “the problem lay in the use of inadequate

materials and the lack of precision production. Korean policy-makers realized

the critical importance of creating a more advanced industrial base” [Horikane

(2005); p.375].

Simply put, before HCI, the South lacked the capital and technology to

develop a military-industrial base on par with the North, which received support

from the USSR and China. The official big push documentation explicitly

motivated the importance of cultivating key input sectors “with a view to

enhancing self-sufficiency in industrial raw materials” [Kim and Leipziger

(1993); p.18-19].

Steel, for example, exemplified a core input into defensive industry. Rhyu

and Lew (2011) records that Park’s preference for steel “traced its origin to both

real and perceived security threats” [Rhyu and Lew (2011); p.323].

Second, Japan’s industrial development influenced the choice of sectors

(Kong, 2000; Stern et al., 1995). Lead HCI planner, Oh Won-chol, carefully

studied the heavy industrial projects of other countries, in particular, Japan

(Perkins, 2013). The New Long-Range Economic Plan of Japan (1958-68)

was especially influential (Stern et al., 1995; Moon and Jun, 2011). Japan’s

plan presented a template of sectors–and their technologies–for which Korea

may have a potential comparative advantage. A World Bank analysis of HCI

tells that Korea used Japan to forecast their sectoral potential; government

documents from 1973 “dutifully note Japan’s export performance in 1955-71

and its composition of manufactures” [Kim and Leipziger (1993); p.18-19].

While the World Bank questions Korea’s proposal to enter into ship-building

as quixotic, Meredith Jung-En Woo argues that Korea’s belief in their latent

comparative advantage lay in Japan: “Where did the Korean government get

its confidence to push shipbuilding so massively? One of the answers was that

Korea had found in Japan’s shipbuilding industry a cynosure. . . observers noted

that the Korean strategy to promote shipbuilding was very simply a carbon copy

of Japan’s” [Woo (1991); p.137]. Similarly, Atul Kohli credits the success of


HCI’s steel push with the availability of Japan’s state-of-the-art expertise [Kohli

(2004); p.112-113].25 In other words, the proximity to Japan—institutionally

and historically—meant the sectoral choices did not defy latent comparative

advantage.

Unlike industrial policies elsewhere, copying (and partnering with) Japan

indicated a concern that HCI sectors did not contradict potential comparative

advantage. Technical requirements for erstwhile HCI projects would be ac-

quired from Japan.26 These technology transfers guaranteed reliable market for

Japanese imports (Kong, 2000; p.53-55; Westphal et al., 1981).27

Policy Levers The 1973 announcement was a distinct pivot in South Korea’s

development strategy: from industrial policies incentivizing general export

activity to a big push policy aimed at driving resources, especially capital,

toward strategic industry.

Before 1973, Park pursued total export-led industrialization. Industrial

policies had no de jure sectoral bias, and scholars argued these policies were ef-

fectively “liberal” (Krueger, 1979; Westphal and Kim, 1982; Westphal, 1990).28

The World Bank’s Larry Westphal summarized pre-HCI policy as “a virtual

free trade regime for export activity” where exporters enjoyed wide exemptions

from import controls [Nam (1980); Westphal (1990); p.44].29 In addition, ample

subsidies bolstered exporters (Cho, 1989).30

After 1973, industrial policy became surgical. HCI-era policies largely

25During HCI, Japanese lending was often contingent on purchasing Japanese inputs and

technologies [Shim and Lee (2008); p.159].26See: Korea’s Economic Miracle: The Crucial Role of Japan, Castley (1997)27Westphal et al. provide empirical evidence of many domestic Korean firms receiving foreign

technological transfer vis-a-vis direct licensing and intermediate input suppliers [Westphal et al.

(1981); p.40].28A leading World Bank study on pre-HCI industrial policy notes, with rare exception, export

incentives “were administered uniformly across all industries.” [Westphal and Kim (1982);

p.217-218]. Nevertheless, these policies likely created distortions and had a de facto biased

toward light, labor-intensive industries.29This ideas us echoed by Korean Development Institute reports on 1960s industrial policy:

“exemption of intermediate inputs and export sales from indirect taxes, and exemption from

import duties on imported inputs allowed exporters to operate under a virtual free trade regime[my emphasis]” [Nam (1980); p.9]

30Cho (1989) notes, until HCI in the early 1970, “the main thrust of directed credit programmes

was to support export ‘activity’ rather than specific industries” [Cho (1989); p.93].


consisted of two levers: investment subsidies and trade policy.

Investment subsidies were the fundamental tool of HCI (Koo, 1984; Woo,

1991; Kim, 1997).31 The National Investment Fund (NIF) opened in 1974

and became the primary means of allocating capital to key sectors.32 Between

1975-1980, the NIF mobilized over 60 percent of financing for HCI industry

equipment. In 1978 alone, at the crest of HCI policy, the NIF accounted for

67.2 percent of all HCI industry loans [Innovation and Development Network

and Kim (2012); Vittas and Wang (1991); p.30].33

The NIF provided discounted financing for equipment investment and

factory construction, and loans were provided through commercial banks and,

in particular, development banks (Koo, 1984).34 Figure 2.2 plots the value of

loans provided by the Korea Development Bank during the HCI period, the

primary lender of NIF funds.35 Grey lines correspond to non-targeted sectors

and red lines indicate targeted sectors. Clearly, after 1973 there is a remarkable

rise in credit lent by the principal NIF lender.

The tax code also shifted to subsidizing investments in HCI industries.36 Ma-

jor reforms after 1974 consolidated industry-specific laws under a new program

aimed at incentivizing investment in key sectors (Kwack, 1984; Kim, 1990;

Trela and Whalley, 1990; Stern et al., 1995). By 1975, the Korean corporate

tax code included a menu of generous investment tax credits and depreciation

allowances for HCI sectors.37

31“Allocation of loanable funds has been one of the most powerful tools to affect patterns of

industrial development in Korea” (Koo, 1984). For overview of state financing of HCI, see Raceto the Swift: State and Finance in Korean Industrialization, Woo (1991)

32“Financial support for heavy and chemical industries may be said to have started with

introduction of the National Investment Fund in 1974” [Kim (2005); p.18-19]. A 1984 Korean

Development Institute study prepared for the U.S. Trade Commission notes the NIF was “the

major source of long-term financing for so-called strategic industries” [Koo (1984); p.36].33NIF was funded primarily through bond sales to banks and to public non-banking institutions

(e.g. pensions). Byung-kook Kim notes the “NIF was an outright forced savings program,” funded

in part by requiring public non-banking institutions to purchase NIF bonds and then requiring 8

percent of wage income to be levied into pensions [B.-k. Kim (2011); p.226].34By the end of HCI, long-term NIF interest rates were about 5 percent lower than conventional

commercial bank loans.35The Korea Development Bank lent 62 percent of all NIF funds through 1981 [OECD (2012);

p.39].36The World Bank reported that “export tax incentives no longer played a central role compared

to that played by [the] industry incentive scheme,” which aimed to concentrate investment in “a

relatively small numbers of industries” [Trela and Whalley (1990); p.19]37In particular, these incentives were provided under the “Special Tax Treatment for Key


support heavy industrial sectors after 1973.

HCI industrial policies did not last. I use 1979 as the de facto end date

for the big push; that year on October 26, President Park was assassinated

by Korean Central Intelligence Agency director, Kim Jae-kyu.43 The murder

signaled a shocking close to the Park’s Yushin dictatorship and the garrison

state’s core policy agenda (Cho and Kim, 1995; p.19; N.-y. Lee, 2011).44

HCI was dismantled in the transition following the assassination.45 In 1980,

Oh Won-chol, the lead HCI planner, was arrested and banned from government

work [Kim (2004); p.8-9]. Between 1981-1983, the commercial banking system

was liberalized. The share of total government loans to manufacturing shrank,

and interest rates between strategic and non-strategic sectors converged (Cho

and Cole, 1986; Nam, 1992 ).46 Between 1979-1980, the transitional govern-

ment implemented multiple rounds of “investment adjustment” for targeted

sectors [Kim (1994); p.349] as trade liberalization progressed in earnest (Kim,

1988; Kim and Leipziger, 1993). The import liberalization ratio, as calculated

by the Ministry of Trade and Industry, climbed from 68.6 in 1979 to 76.6 by

1982.47 Starting in 1982 and again in 1984, maximum import tariff exemptions

for domestic industries were reduced.

43For contemporaneous overview of the Park assassination and its political implications see

South Korea 1979: Confrontation, Assassination, and Transition (Lee, 1980).44Earlier that year, the government had announced the “Comprehensive Stabilization Program,”

in efforts to address the apparent macroeconomic instability brought on by turbulent world

economic conditions and HCI’s imbalances. Nonetheless, the death of Park truly opened the

door for wide-scale liberalization—economic and political.45“[W]ith the death of Park the state’s policy orientation changed fundamentally in the early

1980s, with the EPB-led proponents of economic stabilization and liberalization replacing the

nationalistic mercantilist bureaucrats like O Won-chol in key decision-making positions” [N.-y.

Lee (2011); p.318].46Similarly, in 1981 public finance reforms limited the “special tax treatment for key industries.”

By 1982 the gap in effective corporate tax rates between strategic and non-strategic industries

was closed [Kwack and Lee (1992); Nam (1992); p.7].47In general, though, average import liberalization ratios gradually climbed through the HCI

period 1973-1979. KDI’s Young Soogil writes that import liberalization was only seriously

discussed in 1978, but economic instability in 1979-1980 postponed until the post-Yushin era

[Kim (1988); pg.1].


2.3 Theoretical Framework

Section 2.2.2 described the details of South Korea’s industrial policy, which

used capital subsidies and trade policy to shift economic activity toward targeted

sectors. Below I use a multi-sector model by Long Jr and Plosser (1983), and

revisited by (Jones, 2008, Acemoglu et al. (2012), and Acemoglu et al. (2016)),

to illustrate the general equilibrium effects of the big push. The following

section reviews key elements and predictions of this theoretical framework,

emphasizing externalities generated by industrial policy to forward-linked

(downstream) and backward-linked (upstream) sectors. This framework yields

four simple predictions which I later use to motive my empirical findings.

I model Korea’s industrial policy by considering two factor market distor-

tions, or “wedges,” which planners remove for key industries.48 In the words

of Alice Amsden (1992), planners “get prices wrong” so as to steer resources

toward HCI sectors49 The first distortion, (1 + τMi ) resembles a tax on imported

inputs; the second, (1 + τRi ), a tax on investment.50 Removing (1 + τR

i ) and

(1 + τMi ) leads to growth in targeted sectors. This expansion of supply ben-

efits forward-linked (downstream) sectors, but may be positive or negative

for backward-linked suppliers, depending on whether targeted sectors import

competing intermediate inputs.

Consider an N industry economy. In each industry i, a representative firm

manufactures a single good in a perfectly competitive market with a constant

returns to scale technology. The production function of a representative firm

has the following Cobb-Douglas form:

yi � Ai kαk

ii lαl

ii

N∏j�1

xaj→i

j→i

N∏j�1

mbj→i

j→i . (2.1)

where Ai is productivity, ki is capital, and li is labor. Following the constant

48In a similar spirit, Cheremukhin et al. (2013) consider Stalin’s structural change policies

as the shifting of factor and product market wedges across different sectors. My discussion of

wedges in a general equilibrium Long-Plosser model follows Leal (2016). Rotemberg (2015)

frames Indian capital subsidies in terms of the removal of capital market distortions.49See: “Wrong” Prices, Right Direction? in Amsden (1992).50One could also imagine that industrial policy directly impacts the productivity of targeted

industries. Recent work by Itskhoki and Moll (2016) conceptualizes industrial policy as interven-

tions promoting the revenue productivity of industries with a latent comparative advantage.

2.3. THEORETICAL FRAMEWORK 27

returns to scale assumption with αl , αk > 0, and a j→i , b j→i ≥ 0: αli + αk

i +∑Nı�1 a j→i +

∑Nı�1 b j→i � 1. The subscript, j → i demarcates the direction of

transactions from sector j to sector i, for example a j→i is the cost share of input

j used by industry i.In (2.1), production of good i requires products from other industries, j:

xj→i . With Cobb-Douglas production and perfect competition, the coefficient

a j→i corresponds to entries from the (domestic) input-output matrix, capturing

the share of good j used in the total intermediate input bundle of industry i.Similarly, b j→i corresponds to entries in an input-output matrix for imported

intermediates.51 For now, I assume the two types of inputs are distinct and not

substitutable.

The market clearing condition for industry i includes output sold to other

industries as intermediates, xi→ j , and output consumed as final goods, ci:

yi � ci +

N∑j�1

xi→ j ,∀i. (2.2)

A representative household has Cobb-Douglas preferences u (c1 , ..., cN ) �∏Ni�1 cβi

i , where βi ∈ (0,1) represents the weight of good i in the household’s

preferences, normalized such that∑N

i βi � 1. The household finances consump-

tion through capital and labor income, C �∑N

i ci pi � rK + wL. For simplicity,

I ignore state transfers and ignore trade balance: C � Y. The household’s maxi-

mization problem yields the conditions,pi ciβi

�pj c jβ j,∀i , j, and pi �

βici

Y,∀i. In

other words, consumption shares are constant, each equal to the coefficient

weight in the household’s utility function.

For each industry i, a representative firm’s maximization problem is the

following

max{xj→i }nj�1

,{mj→i }nj�1,ki ,li

��pi yi −wli − (1 + τRi )rki −

N∑j�1

pj x j→i −N∑

j�1

(1 + τMj )p j mj→i

��(2.3)

where p are exogenous world prices for imported intermediate inputs, and

51Due to data limitations, the empirical side of this study is restriction to total input shares:

where Korean input-output matrices combine foreign and domestic input shares.


(1+τRi ) and (1+τM

j ) are distortions on investment and imported intermediates,

respectively.

The firm’s problem (2.3) yields a competitive supply curve for good i as a

function of factor prices and output prices. Accordingly, log-linearized supply

is increasing in productivity (∂ ln yi∂Ai> 0), and decreasing in both the domestic

price of intermediates and the price of imported intermediates (∂ ln yi∂pj,∂ ln yi∂p j< 0).

Differentiating the supply curve with respect to changes in capital taxes (1+τRi )

or intermediate input tariffs (1 + τMj ) yields,

∂ ln yi

∂(1 + τMj )

� −b j→i (2.4)

∂ ln yi

∂(1 + τRi )

� −αki . (2.5)

Prediction 1: Removing import restrictions (lowering (1 + τMj ))

and increasing capital subsidies (lowering investment wedge (1 +

τRi )) promotes real output growth in targeted industries.

It is also useful to consider the effect of industrial policy on prices. Assum-

ing zero profits, industry i’s unit cost function is equal to industry prices. Hence

industry i’s Cobb-Douglas price index is,

pi � κi[(1 + τR

i )r]αk

i wαli

N∏j�1

paj→i

j

N∏j�1

[(1 + τM

j )p j] b j→i

(2.6)

where

κi � ��1

αli

��αl

i ��1

αki

��αk

i N∏j�1

(1

a j→i

) a j→i N∏j�1

(1

b j→i

) b j→i

. (2.7)

In this context, prices are completely pinned down by the supply-side of the

economy. Prices for good i are increasing in domestic and imported intermediate

input prices:∂ ln pi∂pj,∂ ln pi∂p j> 0. Importantly, i’s prices are also increasing in

the size of the intermediate import wedges∂ ln pi

∂(1+τMj )

� b j→i , as well as the

investment wedge∂ ln pi

∂(1+τRi )

� αki . In other words, prices for i are decreasing with

2.3. THEORETICAL FRAMEWORK 29

the industrial policy:

Prediction 2: Industrial policy—removing (1 + τM) and (1 + τR)

for targeted industries—decreases prices in targeted industries.

This framework also illustrates how the expansion of targeted sectors affects

forward-linked (downstream) and backward-linked (upstream) industries. The

combination of Cobb-Douglas preferences and production, guarantees that

supply shocks and demand shocks propagate through the input-output network

in predictable ways (Acemoglu et al., 2016).

First, consider the effect of industrial policy on forward-linked sectors.

Prediction 1 and Prediction 2 show that industrial policies increase the supply

of targeted industry goods. Growth in industry j’s output, yj , and a decline

in j’s output price, pj , are beneficial for downstream industries. To see this,

consider a manipulation of the (2.1); plugging in the first order conditions from

the firm’s optimization problem, and total differentiating after log-linearization:

ln yi varies positively with∑N

j�1 a j→i ln yj .

Moreover, as seen from industry i’s price index (2.6), a decline in the

targeted sector’s price, pj , leads to a decline in the output price pi .52 Hence, the

effect of industrial policy on forward-linked sectors can be summarized as,

Prediction 3: Successful industrial policy confers benefits to forward-

linked (downstream): output increases in purchasing industries and

prices decline.

The expansion of targeted sectors also affects backward-linked industries—

domestic industries that supply goods to targeted sectors. Suppose industry i is

an industry selling goods to targeted industry j. Intuitively, growth in targeted

sector j translates into increased demand for intermediate products produced by

i, xi→ j . Production in industry i increases to meet higher demand for its output.

Moreover, demand shocks do not impact prices, as in this framework prices are

wholly determined by the supply-side of the economy.

To see how industrial policy creates demand shocks for upstream suppliers,

consider the market clearing condition (2.2) for a backward-linked industry i.52Similar downstream effects of industrial policy (specifically, subsidies), are shown by Forslid

and Midelfart (2005).


Total differentiating (2.2), inserting the firm’s first order conditions, and lever-

aging that consumption levels do not change, yieldsd(yi pi )

yi pi�∑N

j�1 ai→ jd(yj p j )

yi pi.

With constant prices, this expression simplifies to dyi �∑N

j�1 ai→ j dyj . Output

of the backward-linked industry, yi , increases with the output of the targeted

sector yj .

Realistically, however, targeted sectors use imported inputs that may com-

pete with domestic industries, in which case industrial policy has negative

effects through backward-linkages (Autor et al., 2013; Acemoglu et al., 2015).

Let mi→ j be an intermediate import used by targeted sector j; this good com-

petes with a domestically supplied good xi→ j . Since the policy lowers the

price of intermediate imports for treated sectors, j imports more mi→ j . The

detrimental effect of import competition can be incorporated into the model in

a reduced form way, incorporating a competing import into industry i’s market

clearing condition (2.2): yi � ci +∑N

j�1 xi→ j −mi→ j .53 Clearly, an increase the

competing import mi→ j reduces i’s output, yi .

Prediction 4: For targeted sectors, industrial policy lowers the cost

of importing intermediate inputs. If intermediate imports compete

with domestic suppliers operating in the same market, then indus-

trial policy creates a negative demand shock for backward-linked

industries and their output declines.

2.4 Data

Digitized Manufacturing Dataset Though South Korea’s moderniza-

tion was a relatively recent historical event, there are few sources of disaggre-

gated, machine-readable data. For my study I created a new dataset on South

Korean manufacturing industries that encompasses the period of rapid industri-

alization.54 To create this dataset, I have digitized and combined materials from

a number of archival sources.55

53Acemoglu et al. (2015) similarly examines the reduced form impact of intermediate imports

on a competing domestic industry by using the market clearing condition.54In South Korea, this include the mining sector as well.55Unless specified, this study does encompass the non-table or agricultural sectors.

2.4. DATA 31

The main source of industrial data were digitized from records published by

the Economic Planning Board’s (EPB) Mining and Manufacturing Surveys and

Census (MMS), 1970-1986.56 The industrial census records were published

approximately every five years from 1970 onward, and intercensal statistics were

published as individual survey volumes. Importantly, the unit of enumeration

for each MMS is the establishment-level. With rare exception, variables are

consistent across MMS publications, allowing me to construct a panel dataset

from digitized materials.

The digitized MMS dataset reports economic statistics at the lowest level of

disaggregation, the 5-digit industry level.57 To illustrate this level of aggregation

consider two samem sectors: 35291, Manufactures of adhesives and gelatin

products, and 35292, Manufactures of explosives and pyrotechnic products. In

other words, MMS industrial data is at a suitable level of variation.

A second source of MMS data come from tape data sold by the EPB in

the 1980s and spans the years 1977-1986. The MMS mainframe data also

reports annual industrial statistics at 5-digit level. However, this data spans a

more limited set of variables relative to those published in the digitized MMS

volumes. Variables includes (nominal) value of shipments, employment, wage

bill, total fixed capital formation and total capital disposals. Data extracted from

these tapes was cleaned using OpenRefine and converted to a contemporary

data format.

The digitized MMS data was combined with the mainframe tape data to

create a single harmonized panel. Table 2.2 reports pre-1973 averages and

standard deviations for major industrial variables used in this study. Two data

transformations are used for both dependent and independent variables: log

normalization (with a small constant) and inverse hyperbolic sine (IHS) normal-

ization. Since many variables, such as capital acquisition variables, have many

0s, the IHS transformation is preferred. While IHS approximates log, estimated

coefficients are not as readily interpretable. Since in almost all cases log and

IHS estimates are essentially equivalent, log-normalized interpretations appear

56The Economic Planning Board is also the historic predecessor to Statistics Korea.57Firm-level data from the period is not available in published or machine readable format.

To my knowledge, early firm- or establishment-level data is unavailable for most of the study

period. However, product-level data and data by firm-size bin × industry data have also been

digitized and compiled for my database.


in the text and IHS estimates appear in tables.

Harmonization and Crosswalk Schemas My analysis requires industrial

and product definitions that are consistent through time. For the MMS industrial

publications, the EPB used codes based on the International Standard Indus-

trial Classification (ISIC) system. Nonetheless, South Korean industrial codes

were updated repeatedly (1970, 1975, and 1984), requiring multiple crosswalk

schemas to build a harmonized industry panel. The crosswalk schemas — algo-

rithms for harmonizing across many industrial coding schemes — were created

with the help of concordance tables digitized from Economic Planning Board

publications. These crosswalks allowed me to map sector definition “splits” to

time-consistent industry identifiers.

For the main MMS industrial census dataset, the crosswalk schemes were

used to map sector “splits” back to their original code format. For example,

consider an example from the non-metallic minerals sector. In 1975 the indus-

tries (36994) Manufacture of Asbestos Products and (36995) Manufacture of

Mineral Wools were split from the 1970 industry (36996) Manufacture of Stone

Texture. My crosswalk schema aggregates the two 1975 sector codes back to

their original 1970 code.

Conversely, some Korean industry codes were merged through time.58 For

example the 1975 sector (32163) Manufacture of Man-made Fibre Fabrics was

merged from two distinct 1970 industry codes: (32172) Manufacture of Silk

Fabrics and (32176) Manufacture of Fabrics of Man-made Fibers. In the case

of aggregation of sectors through time, the two 1970 industries are aggregated

into a larger synthetic sector, instead of splitting the 1975 industry into two

separate industries.

The preceding harmonization process was performed for all Korean industry

code changes for revision years 1970, 1975, and 1984. After harmonization, the

1970-1986 industrial panel is a bit more aggregated than each individual cross

section, yielding 268 consistent industry codes for the main MMS dataset.

In addition to harmonizing digitized manufacturing data through time, man-

ufacturing, price, trade, and input-output panels each use their own coding

58Clearly, accounting for simple renaming of sector codes is a trivial problem.

2.4. DATA 33

system.59 Thus, further crosswalk schemas were used to harmonize datasets

across coding schemes. Thus, over a dozen harmonization algorithms were

required to create the main 5-digit industrial panel used below.

Input-Output Network Data Intersectoral linkage data comes from South

Korea’s 1970 basic input-output (IO) tables, published by the Bank of Korea.

The 1970 IO tables were translated from Korean into English and then digitized

into a machine-readable format.60 Machine readable input-output tables for

later periods (1975, 1980, 1983, and 1985) were graciously provided by the

Bank of Korea.61

Trade Policy and Trade Data A panel of South Korean trade data has

been constructed using the World Bank’s World Integrated Trade Solution

(WITS) database, 1962-1987. Trade data analysis is conducted at the 4-digit

ISIC (Revision 2) level.

Detailed measures of quantitative restrictions (QRs) and tariffs were digi-

tized from Luedde-Neurath (1986) and are available at the product-level (Cus-

toms Commodity Code Number, or CCCN, product-level). Luedde-Neurath

(1986)’s dataset is used because it is the most complete and disaggregated

available.62

The digitized trade policy data was then merged with the 1970-1986 MMS

industry panel. Average tariffs (QRs) on output were calculated for each 5-

digit KSIC industry. Input tariffs (QRs) are calculated as the weighted sum of

59Manufacturing data: Korean Standard Industrial Classification; prices: current (as of 2015)

Bank of Korea industry classifications; trade: ISIC (Rev. 2); and input-output data: historic Bank

of Korea sector codes.60The basic input-output tables for 1970, which encompass 320 sectors, was not available from

the Bank of Korea in machine readable format. Unlike later years, the 1970 tables report totalvalues of flows between industries and does not differentiate between domestic and imported

values, as later publications do.61Once again, all IO data was harmonized into consistent sectoral definitions using a crosswalk

schema and concordance definitions digitized from IO table publications. Since IO tables use a

separate industrial classification system from the industrial census/surveys, a crosswalk schema

is used to combine the datasets.62Westphal (1990) notes it is the most extensive source for. Alternative studies of South

Korean tariff structure are often highly aggregated; make strong assumptions with the intention

of measuring effective rates of protection; and focus mostly on period of 1960s export-oriented

industrial policies.


average tariff (QR) exposure for each input into industry production using the

1970 input-output tables. Following Amiti and Konings (2007) and Amiti and

Davis (2012), the input tariff (and QR) exposure is defined as input-tariffi �∑j α j→i × output-tariff j , where α j→i are estimated cost-shares for industry i

from the input-output accounts.

2.5 Direct Effects of Industrial Policy

In this subsection I estimate the direct effect of the HCI big push on in-

dustrial development. Before turning to the core development estimates, I first

discuss sources of endogeneity and motivate the estimation framework. Next, I

show that measures associated with industrial policy change differentially for

targeted and non-targeted sectors, as modeled by policy wedges in my theoreti-

cal framework (Section 2.3). Finally, I confirm Prediction 1 and Prediction 2 of

my model and show that targeting was associated with the development of HCI

industries.

2.5.1 Direct Effects: Empirical Framework

Identification I contend the Korean HCI context is a natural experiment in

that (1) targeting was orthogonal to traditional sources of bias, and (2) industrial

policy conformed to notions of latent or dynamic comparative advantage.

Estimating the (direct) effect of industrial policy on industrial development

is often problematic. Industrial policy is state action, and thus policies are allo-

cated according to politics (Grossman and Helpman, 1994; Goldberg and Maggi,

1999; Baldwin and Robert-Nicoud, 2007). Such political-economy factors can

be both unobserved and negatively correlated with industry fundamentals. Un-

surprisingly, many empirical studies report a negative relationship between

the effect of protection on growth or productivity (Harrison, 1994; Harrison

and Rodriguez-Clare, 2009; Rodriguez and Rodrik, 2001). Moreover, unlike

many economic policies, research designs based on the random allocation of

policies may be uninformative (Rodrik, 2004). Industrial policy are systematic

interventions to promote industries with a latent comparative advantage (Noland

and Pack, 2003; Lin and Chang, 2009).

2.5. DIRECT EFFECTS OF INDUSTRIAL POLICY 35

Two sources of political bias translate into a negative relationship between

industrial development and interventions.

First, policies often benefit declining, or “sunset,” sectors.63 For example,

Japan’s Ministry of International Trade and Industry (MITI) notably intervened

in troubled manufacturing sectors and similar policies have been widely docu-

mented around the developing world.64

Second, around the world cronyism shapes the allocation of interventions,

which frequently defy notions of comparative advantage (Rodrik, 2005; Lin

and Chang, 2009; Lin, 2012). For example, Tommy Suharto, son of Indonesia’s

General Suharto, received gracious subsidies to develop a national automobile

industry—without any prior experience or skill in automobile manufacturing

(Eklof, 2002; Fisman and Miguel, 2010). Ferdinand Marcos, Park Chung-hee’s

contemporary in the Philippines, used ambitious, capital-intensive industrial

projects as a vehicle for pure clientalism rather than industrial development

(Boyce, 1993; Kang, 2002; White, 2009).65

In South Korea, targeted industries were not chosen because of unobserved

and/or anticipated declines in economic conditions, nor were they chosen due

to political criteria that defied latent comparative advantage. Why did HCI cut

across critical sources of unobserved endogeneity?

To begin with, many industries targeted by South Korea, such as shipbuild-

ing, simply did not exist in that country, and so could certainly not have been

sunset industries. To argue that unobserved negative trends guided policy — neg-

ative or otherwise — is moot. The chemical industry was similarly minuscule

and had to be built from scratch (Woo, 1991).66

63A theoretical literature has long discussed optimal policies to declining industries (Gray,

1973; Hillman, 1982; Flam et al., 1983).64For example the U.K.’s National Enterprise Board, buffered a failing automotive industry in

the 1970s (Hindley and Richardson, 1983; Sawyer, 1992). U.S. presidential candidate Richard

Nixon wooed southern constituents with protection for textile sectors facing declining compar-

ative advantage (Cox and Skidmore-Hess, 1999). Supports for declining industry defined U.S.

industrial policy debates in the Reagan-era (Congressional Budget Office, 1983).65For example, Marcos forced U.S. auto parts manufacturers out of the Philippine market,

granting monopoly rights and industrial subsidies to crony, Ricardo Silverio, who promptly

mismanaged nearly a billion pesos in liabilities before bankruptcy in 1984 (Kang, 2002; p.140;

White, 2009).66Woo-Cummings notes during HCI, “[t]he chemical industry in Korea was built on practically

nothing, unlike other industries that had some vested enterprises to start from. Korean dependence

on imports of fertilizers from 1955-1961 was an amazing 100 percent” [Woo (1991); p.139].


Institutionally, the political environment of South Korea meant that policy

was guided by strategic criteria rather than the cronyism. A binding security

crisis provoked a shift in national industrial strategy with little political inter-

ference. Park’s sudden consolidation of power allowed for the creation of a

technocratic Heavy Chemical and Industry Planning Board that superseded

competing political actors. Planning conformed to what Peter Evans called

“embedded autonomy:” a bureaucracy insulated from special interest politics and

administered by specialists with knowledge of environment they are operating

in (Evans, 1995).67

A core criterion for successful industrial policy is that targeted industries

possess dynamic, or latent, comparative advantage. Though Korea did not have

static comparative advantage in HCI industries, targeted sectors did not grossly

defy latent comparative advantage as with industrial policy of other countries.

In section 2.2.2, I explain that Japan’s earlier heavy industrial targeting reflected

the potential comparative advantage of Korean industries. Moreover, profes-

sional bureaucratic guidance minimized the potential of choosing sectors that

contradicted notions of comparative advantage.68

The dynamic differences-in-differences framework I introduce below maps

naturally into a notion of latent, or dynamic, comparative advantage. The thrust

of industrial policy is that the state is selectively intervening in sectors to

produce industrial development that would have not occurred had the economy

expanded according to static comparative advantage [Noland and Pack (2003);

p.10]. This dovetails with assumption of differences-in-differences estimation:

without policy interventions, the economy would have evolved according to the

pretrends — that is, according to static comparative advantage.

Estimation Framework The first estimating equation explores the relation-

ship between industrial targeting and industrial development during the big

push. This framework estimates the year-specific differences between targeted

and non-targeted industries relative to a 1972 baseline, the year before the an-

67The South Korean developmental bureaucracy, specifically, is a representative of Evan’s

embedded autonomy concept.68Stern et al. (1995) notes the use of technical and scale feasibility studies used by HCI

planners to constrain the choice of industries [Stern et al. (1995); p.23-25]. For instance the

construction of jet engines was seen as beyond the technical capability of South Korea.


nouncement of the industrial policy drive. Concretely, I estimate the following

specification:

Yit �

1986∑j�1970

β j ·(Targetedi ×Year

jt

)+

∑i�n

αn · Ini +

1986∑j�1970

λ j ·Yearjt +

1986∑j�1970

X′iYearjtΩ j + εit

(2.8)

where Y is an industrial development or policy-related outcome, i indexes 5-

digit industries, and t indexes the years 1970–1986. The variable Targeted is an

indicator equal to one if a sector is targeted by the Heavy Chemical and Industry

committee, zero otherwise; Year are time period indicators. Specification 2.8

contains industry-level fixed effects∑

n In and time period effects∑

j Year j .

Preferred specifications include a rich set of pre-treatment variables—and

their trends—to control for unobserved productivity. Controls include average

establishment size, average wages, raw material costs, employment, fixed capital

investment, and labor productivity. Each baseline control (trend) is interacted

with time period indicators:∑1986

j�1970 X′iYearjtΩ j .

The coefficient of interest in equation 2.8, β j , gives the estimated difference

between targeted and untargeted sectors in year j relative to 1972, the year

preceding the big push announcement. The set of estimated coefficients give a

sense of the differential evolution of targeted industries through time. Before the

policy, I expect no difference between targeted and untargeted sectors: β1970 ≈β1971 ≈ β1972 ≈ 0. After the 1973 policy announcement, I expect increasing

differences between the two types of sectors, β1974 ≤ β1975 ≤ ... ≤ β1979, until

1979, when Park Chung-hee was assassinated and the dissolution of HCI was

binding. For years after 1979, we may expect that the estimated coefficients

decline after subsidies are removed: β1979 ≥ β1980 ≥ β1981... ≥ β1986.69

While estimates from the flexible specification in 2.8 convey the pattern

of the policy roll-out, it is useful to get a sense of the total average impact of

industrial targeting before and after 1972. Here the conventional differences-

in-differences is useful. I ascertain the average effect of targeting on indus-

69For a similar discussion, see: Nunn and Qian (2011).


trial development by interacting the Tar geted sector indicator with a post-

announcement indicator:

Yit �β · (Targetedi ×Postt)

+

∑i�n

αn · Ini +

1986∑j�1970

λ j ·Yearjt +

1986∑j�1970

X′iYearjtΩ j + εit

(2.9)

Substantively, the estimated coefficient of interest, β, captures the average

growth in treated industries before-after the policy announcement. The Targetedi×Postt interaction is the only difference between the difference-in-differences

equation (2.9) and the flexible regression in equation (2.8).

2.5.2 Results: Targeting & Policy Mixtures

I now confirm that industrial policy packages significantly changed for

targeted relative to non-targeted sectors. First, I study the impact of subsidies

by examining whether investment activity in targeted industries change signif-

icantly over the HCI period (1973-1979), relative to non-targeted industries.

How did the relaxation of credit constraints affect fixed and variable costs?

Given that many subsidies were intended for capital accumulation, I examine

measures of gross fixed capital formation. I then turn to the effects of HCI

on (real) capital investment across different assets. Credit also financed the

purchase of other advanced intermediates. Thus, I also examine changes in

(real) materials expenditure, following Banerjee and Duflo (2014) and Manova

et al. (2015).

Next, I turn to protectionism. HCI policies were long associated with trade

policy in the form of output protection and import protection. Exemptions

from tariffs and non-tariff barriers (quantitative restrictions) were given to the

purchasers of imported inputs and protective measures (purportedly) sheltered

domestic industry from international competition. Thus, in addition to subsidy

variables, I analyze changes of trade policies over the planning period.

Responses to Targeted Subsidies Figure 2.4 conveys the relative changes

in (gross) fixed investment measures and materials investment for the periods


1970-1986, relative to a 1972 baseline. Panels A and B plot the flexible coef-

ficient estimates of equation (2.8) for each year. Figure 2.4 Panels C and D

examine differences in targeted versus non-targeted industry capital acquisitions

for two types of assets: equipment and buildings, respectively. Because state

lending, especially from Korea’s National Investment Fund (see Section 2.2.2),

emphasized the financing of equipment purchases and factory expansions for

HCI firms. All specifications include both 5-digit industry fixed effects, period

fixed effects, and include baseline covariates and pre-trends, both interacted

with period fixed effects. Data for disaggregated capital acquisitions is only

available until 1982 and does not include acquisitions for the census year 1973.

The light gray bands represent standard errors for each coefficient, clustered at

the 5-digit industry-level.

Figure 2.4 illuminates four points. First, a robust pattern confirms that,

conditional on controls, targeted and non-targeted sector outcomes are not

significantly different before the policy announcement. There is no sign of

significant anticipatory investment activity. Second, there is a conspicuous

divergence in purchases of total intermediate inputs and fixed capital—both in

aggregate capital and across all asset classes. Third, this divergence wanes after

Park’s 1979 assassination and the subsequent liberalization of the economy.

For all outcomes, estimated differences decline relative to their 1979 peak,

corresponding to the liberalization of state lending in the early 1980s.70 Finally,

plots for disaggregated capital investment are consistent with the investment

pattern incentivized by state-lending policy, which favored equipment and

construction investment (Yoo, 1990; p.39-41; World Bank, 1987).71

While Figure 2.4 presents the pattern of estimates for (2.8), it is also infor-

mative to estimate the average effect over the same period.

Table 2.3 shows average estimates of HCI targeting on total value of (real)

gross capital formation and total (real) value of intermediate input purchases.

Columns (1)-(3) report estimates for capital acquisitions; columns (4)-(6), mate-

70The second oil crisis also corresponds to the year 1979. While the oil crisis should nega-

tively impact HCI industry, the plots reveal a sustained dip in differences through the 1980s.

Moreover, the first global oil shock (1973-1974) coincided with the beginning of the policy, and

a commensurate dip does not appear in the estimates for the period.71The pattern also indicates the source of worries of growing excess capacity prior in the early

1980s (Kim, 1994).


indicators of trade restrictiveness (Mason, 1980; Nam, 1995). While the HCI

period is associated with highly interventionist policy, in fact the South Korea

was actively liberalizing its trade policy since the late 1960s. From 1970-1980,

import controls dropped. Though after the post-1979 liberalization episode,

some of the import controls for targeted industries remained, as is evident from

the output tariff/QR panels of Figure 2.5, and liberalization of trade policy

occurred mostly after 1982, the end of the sample (Yoo, 1993). Moreover, im-

port controls are significantly lower for only a few periods for tariffs and QR

estimates, since import restrictions were generally falling over the period.

Table 2.5 simplifies the flexible regression analysis, showing average es-

timated changes in trade outcomes after 1973. Columns (1)-(6) report esti-

mates for average output protection; columns (7)-(12), average input measures.

Columns (1), (4), (7), and (10) include only time and industry fixed effects.

Columns (2), (5), (8), and (11) include baseline control averages (with period

interaction). Columns (3), (6), (9), and (12) add pretrend controls. Importantly,

differences in average output protection for targeted industry is insignificant

and the estimates straddle zero.

Input protection measures, however, declines significantly for targeted

industries and results are robust across specifications. Point estimates for QRs

for preferred specifications are -.045 (5 percent level) . Estimates for average

import tariffs are more negative: -.192 (1 percent level), translating into an

average of 21 percent lower input tariff exposures for targeted industries relative

to non-targeted after 1973.


2.5.3 Results: Manufacturing Growth and Industrialization.

Having confirmed that industrial policies, especially responses to directed

credit, vary as expected over the big push period, I turn to industrial growth and

industrial development outcomes.

Growth (Prediction 1) Figure 2.6, Panel A plots estimates from equation

(2.8) for industrial output (real value shipped). Estimated coefficients include

time and year fixed effects, as well as time-varying baseline controls and

associated pretrends. The estimates illustrate a distinct pattern similar to that of

the industrial policy plots in Section 2.5.2, in particular the results for capital

subsidies.

The industrial growth results in Figure 2.6, Panel A convey three key

insights. First, conditional on controls, the plots show no pre-pretreatment

differences between targeted and non-targeted industries prior to the 1973

policy announcement. Second, after 1973, estimated differences between treated

and non-treated industries widen markedly. Finally, following Park Chung-

hee’s assassination and the retrenchment of interventions in 1979, estimated

differences in output declines a bit but nonetheless remain significantly higher

than their 1972 level relative to non-targeted sectors.

For completeness, Table 2.6 column (3) shows the estimates associated with

Figure 2.6 Panel A, along with two other measures of output: gross output (4)-

(6); and value added (7)-(9). Models in columns (3), (6), (9) report estimates for

models with the full set of controls. Columns (2), (5), and (8) exclude pretrends.

Specifications with only year and industry fixed effects correspond to columns

(1), (4), and (7). The table confirms that the plotted coefficients presented in

Figure 2.6, Panel A are robust across various measures of output and controls.

Table 2.7 presents estimates of the average effect of targeting on industrial

growth for periods after 1973. Preferred estimates for (real) value shipped in

column (3) indicate average changes of 0.614 at 1 percent level of significance.

These estimates translate into a nearly 85 percent difference in industrial growth

between treated and untreated industries. Similar estimates for gross output (6)

and value added outcomes (9) show a, respective, 81 percent (5 percent signifi-

cance) and 77 percent (1 percent significance) difference in growth between the


targeted and non-targeted sectors for the same period.

Factor Productivity and Prices (Prediction 2) Figure 2.6, Panel B visu-

alizes the pattern of coefficient estimates for labor productivity, measured as

(real) gross output per worker. The pattern for labor productivity reveal the

same pattern for the levels of output in Panel A.

Table 2.8 reports average estimates for labor productivity. Columns (1)-

(3) show estimates for value added labor productivity; columns (4)-(6), gross

output labor productivity. The preferred specifications for estimates of industrial

productivity appear in columns (3) and (6) correspond to an average relative

growth in labor productivity of 3 percent (5 percent significance) and 9 percent

(1 percent significance), respectively, for value added and gross output-based

measures.

Figure 2.6, Panel C reveals the relative fall in output prices for targeted

sectors. While labor productivity (Panel B) is an incomplete measure of produc-

tivity, the strong relative decline in prices during and after the HCI planning

period are telling, as well as highly significant. Table 2.9, column (3) suggests

output prices fell 11 percent more in targeted relative to non-targeted sectors (1

percent level of significance). Estimates for price outcomes results are robust

across specifications.

Structural Change, Entry, and Labor The big push aimed to reallocate

manufacturing activity from low value added, light industries to HCI sectors.

Figure 2.7 reports standard structural change outcomes: Panel A, share of

manufacturing output and Panel B, share of manufacturing employment. The

figures reveal that HCI effectively reallocated manufacturing activity to strategic

industries. More so, even after the retrenchment of HCI policies starting in

1979, the average share of activity in strategic sectors continued to grow more

than other manufacturing sectors, relative to 1972 levels.

In other words, Figure 2.7 makes the case that HCI policy induced structural

change towards strategic industry. Table 2.9 reports the average relative rise

share of manufacturing employment (Column 15) and share of manufacturing

output (Column 18). These estimates suggest that the share of manufacturing

employment for HCI industries rose over 40 percent (10 percent significance).


2.6 Network Externalities

The case for industrial policy has typically been motivated by the existence

of positive spillovers beyond treated sectors (Krueger and Tuncer, 1982; Pack

and Westphal, 1986; Grossman, 1990; Krugman, 1993). A classic literature

in development highlighted the importance of linkages in justifying industrial

interventions: notably Scitovsky (1954), Rasmussen (1956), and Hirschman

(1958). South Korea’s Heavy Chemical and Industry drive exemplified a big

push policy targeting key intermediate goods sectors. Having shown the sudden

growth of HCI sectors (Section 2.5.3), I examine how this growth impacted

non-targeted sectors through the input-output network.

Accordingly, I use the traditional language of development economics

(“linkages”) to discuss the network externalities. Effects of HCI interventions

propagates through backward linkages—to downstream firms selling goods to

targeted sectors, or through forward linkages—to upstream firms purchasing

goods from targeted sectors.

The network graphs shown in Figures 2.9 and 2.10 illustrate the (pre-

treatment) variation in linkages for the South Korean economy. Both plots

visualize input-output accounts (aggregated 153 × 153 sector) for 1970, in-

cluding both tradable and non-tradable sectors.73 Red nodes correspond to

targeted (HCI) industries; gray nodes, non-targeted. The relative size of each

node is weighted by its number of raw connections (“degrees”, in the language

of network theory).74

Figure 2.9 gives a sense of the distribution of forward links (“out degrees”)

from IO sectors, and figure 2.10 shows the distribution of backward links (“in

degrees”) to IO sectors. I use the Kamada-Kawai algorithm (1989) to determine

the graph layout, and nodes for industries with more links appear closer to one

another. The targeted nodes vary considerably in terms of inward links and

outward links. Moreover, targeted industries are not the most central nodes, nor

are they weakly connected nodes on the periphery.

73While the study uses 320 × 320 sectors, I use the “medium” 153 × 153 input-output accounts

for visual clarity. Summary “sectors,” such as employee remuneration, and scrap sectors are

excluded.74Note: The number of “treated” HCI nodes in the graph differs from number used in the

industrial census dataset, since input-output data is presented at a different level of aggregation.

2.6. NETWORK EXTERNALITIES 51

2.6.1 Measures of Network Exposure

To estimate the impact of industrial policy through intersectoral linkages, I

construct two measures of network exposure to industrial policy. First, I focus

on the direct exposure to policy by using the total weighted share of sales

(purchases) to (from) targeted sectors. However, sectors two degrees away from

a targeted sector may also be exposed indirectly to the policy. Thus, I introduce

a second measure of network exposure that captures total exposure to targeted

sectors. To do so, I utilize a measure based on the famous Leontief inverse. As is

well known, the Leontief inverse measure captures not only first-degree linkage

effects between sectors, but also second, third, fourth, etc., degree relationships

to (from) targeted sectors.

Direct Linkages Direct (first-degree) measures of network exposure are

calculated in the following way.

Consider industrial policy effects that propagate through backward linkages.

Let i be non-targeted industry.75 A single backward link is defined as a connec-

tion between industry i and industries purchasing their output, indexed by j.This relationship is denoted by the subscript i→ j.

The backward linkage measure is defined as the weighted sum of all links

between industry i and their buyers:

Backward Linkagei �∑

j

αi→ j (2.10)

where the linkage weight αi→ j is defined as

αi→ j �Sales i→ j∑j′ Salesi→ j′

(2.11)

The linkage weight (2.11) is the value of i’s sales to j, divided by the total sales

of i to all purchasing industries j′.76 Following traditional input-output analysis,

75The description of the first-degree connections and their calculation follow the language of

Acemoglu et al. (2015) and Acemoglu et al. (2016).76Note, I do not count i’s sales to itself; this amounts to excluding “diagonals” in the input-


the denominator of equation 2.11 is equivalent to summing over industry i′stotal sales across all industries—tradable and non-tradable alike—including i’soutput sold as final products.77 Notice that weight αi→ j is the very weight used

in j’s Cobb-Douglass production functions (Section 2.3, Equation 2.1).

We are interested in how industry i may be exposed to HCI policy vis-

a-vis their total collection of backward (forward) linkages to (from) targeted

industries only. Equation 2.12 captures the policy exposure by summing the

share of sales (αi→ j) only to targeted industries:

Backward HCI Linkagesi �∑

j∈HCI

αi→ j (2.12)

In other words, (2.12) measures only linkages between i and targeted sectors

( j ∈ HCI, where HCI is the set of all targeted industries).

The preceding calculations were shown for backward linkages. The forward

linkage versions of equation 2.12 are calculated in a similar manner: measure

Forward Linkagesi is equal to∑

j α j→i and Forward HCI Linkagesi is equal to∑j∈HCI α j→i . Similarly, a Forward non-HCI Linkagesi captures these forward

linkages to non-HCI manufacturing sectors. Thus, the forward linkage measure

reflects the extent to which industry i’s intermediate inputs are purchased from

targeted industries j.

Total Linkages The measures calculated in equation 2.12 capture only direct

spillovers from industrial policy. By appealing to the Leontief inverse, however,

I construct a complete linkage measure that accounts for the n-degree effects of

industrial policy through backward (forward) linkages.

Define the technical coefficient matrix AAA as a matrix of the weights defined

in equation 2.10. An entry of AAA, ai→ j , captures the share of sales from industry

i sold to industry j.

output table, i.e. αi→i � 0. In the parlance of network/graph theory, I do not count “loops.”77See: Acemoglu et al. (2016)’s calculations.


AAA ≡

⎡⎢⎢⎢⎢⎢⎢⎢⎢⎢⎢⎣

a1→1 a1→2 . . . a1→ j

a2→1 a2→2 . . . a2→ j...

.... . .

...

a j→1 a j→2 . . . a j→ j

⎤⎥⎥⎥⎥⎥⎥⎥⎥⎥⎥⎦(2.13)

The Leontief inverse is calculated by taking the inverse of the technical

coefficient matrix AAA, LLL ≡ (I−AAA)−1:

LLL ≡

⎡⎢⎢⎢⎢⎢⎢⎢⎢⎢⎢⎣

1→1 1→2 . . . 1→ j

2→1 2→2 . . . 2→ j...

.... . .

...

j→1 j→2 . . . j→ j

⎤⎥⎥⎥⎥⎥⎥⎥⎥⎥⎥⎦(2.14)

Consider a single entry,i→ j , from the Leontief matrix LLL in 2.14. These Leontief

coefficients represent how much a 1 percent increase in sector j’s output raises

sector i’s output.78 If i→ j � 1.2, a 1 percent rise in industry k leads to a 1.2

percent rise in i, accounting for all of j’s first, second, third, etc.., degree effects

on i’s output.

I calculate the total backward linkage effects of industrial policy using the

following measure:

Total Backward HCI Linkagei �∑

j∈HCI

i→ j (2.15)

The measure in equation 2.15 adds industry i’s Leontief coefficients for pur-

chasing sectors, j, but only for j’s targeted by the HCI big push.79 In other

words, for an industry row i, I add together column-wise entries j for j’s in the

set of targeted industries.

One can think of Total Backward HCI Linkagei as being the n-degree ana-

logue of the direct backward linkage measure (equation 2.12). Substantively,

Total Backward HCI Linkagei captures the total exposure of industry i vis-a-vis

targeted industries purchasing i’s output.

78In this method of input-output economics, more precisely, the entry refers to a rise in i’sfinal demand.

79As with the direct linkage calculations, I do not count on-diagonal Leontief coefficients.

E.g.: i→i .


The preceding calculations were shown for total backward linkage effects

of industrial policy. The Total Forward HCI Linkagei measure is calculated in

a similar way. However, instead of summing across columns for row i, I sum

across rows, indexed by j, for column i. These row-wise sums are restricted to

suppliers k in the set of targeted industries.

It is helpful to get an intuition for the types of sectors with strong connec-

tions to treated industries. Figure 2.11 lists non-targeted sectors with the highest

direct connections to targeted sectors—measured by Backward HCI Linkagesi

and Forward HCI Linkagesi , Equation (2.12).80 The left-hand side shows the

top-20, 5-digit manufacturing industries with the highest share of inputs sourced

from targeted sectors. These sectors include Jewelry & related articles and Plas-

tic products, with over 60 percent of intermediate inputs coming from targeted

industries. Qualitatively, many of the products with high forward linkages from

HCI sectors are more downstream industries.

On the right-hand side, I list the top 20 industries with the highest direct,

backward-links to targeted sectors. Unsurprisingly, many of the sectors supply-

ing a large share of output to targeted industries are raw material sectors, such

as processed ores and various non-metallic mineral products. Many of these

industries send over 50 percent of output to HCI industries.

2.6.2 Network Economies: Empirical Strategy

The proceeding analysis focuses on the spillover effects from targeted in-

dustries to external industries. Figure 2.12 shows a simple bivariate relationship

between log growth (1972-1982) and the strength of (first-degree) 1970 link-

ages (Equation 2.12) from/to treated sectors. Grey dots represent non-targeted

industries; red, targeted. Regression slopes are shown for non-targeted and

targeted observations, though neither are significantly different.

The empirical pattern displayed in Figure 2.12 encapsulate the patterns I

will explore in depth below. The left-hand panel shows a positive relationship

between forward linkages from targeted sectors and (real) growth in value of

80Names of the sectors reflect both the harmonization of industry names through time, as

well as the matching of input-output tables to 5-digit industry codes. Industry names may not

be literally interpretable and are meant to convey a general, qualitative pattern to the reader.

Measures Backward HCI Linkagesi and Forward HCI Linkagesi are presented in raw formats.


Weaving & spinning industry, NECOther fibre yarn 2Other fibre yarn 1

Other transportation equipment 1Other transportation equipment 2

Rubber footwearString & processed string goods, NEC

Rope & fishing nets 1Rope & fishing nets 2Bicycles & cycle carts

Worked hair & postiches 1Worked hair & postiches 2

Flat glassManganese ores

LimestoneMiscellaneous manufactured articles

Abrasive productsDyeing & finishingSynthetic fibre yarn

Synthetic fibre fabricsPlastic products

Jewelry & related articles

0.00 0.25 0.50 0.75

Share Forward Linkages

Indu

stry

Nam

e

Packaging container & related productsPaper containers & other products 2

Pulp, NEC, species & cardboard manufacturingPaper containers & other products 1

Other leather products 2Other leather products 1

Abrasive productsStone products

Crude saltGraphite

TalcAnimal & vegetable oils & fats

Miscellaneous non−metallic minerals 1Miscellaneous non−metallic minerals 2

Rubble collected industries, NECSilica sand

Manganese oresGold & silver ores

Moribuden miningCopper ores

Tungsten oresMiscellaneous non−ferrous metal ores

Zinc ores

0.00 0.25 0.50 0.75 1.00

Share Backward Linkages

Figure 2.11: Top 20 Non-HCI Sectors with Highest Forward and Backward

(Direct) Linkages to Targeted Industry, 1970.

output shipped, 1972-1982. The coefficient for the combined regression is β �

1.8350 (t � 3.110). Panel A indicates a potentially strong positive relationship

output growth and the strength of forward connections from targeted sectors.

On the other hand, the right-hand panel of Figure 2.12 shows a negative, weak

relationship between backward linkages and industrial growth over the same

period: β � −0.9871 (t � −1.63).

I estimate the effect of the HCI big push on backward (forward) linked

industries, regressing industrial development outcomes on my first-degree (and

also total) linkages measures defined above. These continuous measures are

interacted with time period indicators to convey the dynamic pattern of changes

for backward (forward) linked industries.

Specifically, I estimate the following flexible specification:


Yit �

1986∑j�1970

γj ·(Backward HCI Linkagesi ×Year

jt

)+

1986∑j�1970

β j ·(Targetedi ×Year

jt

)+

∑i�n

αn · Ini +

1986∑j�1970

λ j ·Yearjt + εit

(2.16)

The parameters of interest are the estimated γjs, which show the growth of

linked sectors versus unlinked sectors, relative to the pre-treatment levels. Sub-

stantively, these coefficients represent the estimated changes in linked, relative

to changes in less-linked sectors. Prior to 1972, the estimated effect ought to be

0, indicating no anticipatory effect of the policy on linked industries. Estimates

after 1972 should increase gradually, until at least the 1979-1982 period, when

the policies were taken away. Estimates for the post-liberalization period indi-

cator long-run effects of the policy (if coefficients continue to be greater than or

equal to earlier estimates) or temporary-policy effects (if coefficients decline

for periods after the policy).

I control for the direct effects of targeting using the time-varying interaction

term: Targeted×Year. As in the direct effect analysis, I include industry controls∑n In time period fixed,

∑j Year j . Standard errors are clustered at the 5-digit

industry level.

The identifying assumption is that, conditional on industry and year fixed

effects, the difference in industrial growth between backward (forward) linked

and non-linked industry would have changed similarly in the absence of the HCI

industrial policy. Section 2.5.1 explained the HCI interventions were orthogonal

to conventional sources of bias. For the current empirical exercise, I take the pre-

determined input-output network (1970) to be exogenous to the rapid growth of

targeted sectors.


Panels A (direct forward linkage effects) and Panel B (total forward linkage

effects) indicate industries that purchased larger shares of input from treated

sectors grow more than other industries, relative to pre-treatment levels. Esti-

mates for both models indicate industries with strong upstream connections

benefited from the policy during the 1973-1979 period. Moreover, estimated

differences using the direct linkage measure diminish after 1979 (Panel A).

However, the post-1979 effects are stronger when accounting for total forward

linkage exposure (Panel B).

Similarities between the two measures indicate that the major effect occurs

for industries most directly connected to targeted sectors and rapidly dissipate.82

These findings are consistent with Prediction 3 of the multi-sectoral model.

Table 2.11 reports the average effect for direct, forward linkages before

and after the policy announcement. These estimates correspond to a simple

differences-in-differences version of the dynamic specification, Equation .

Columns (1), shows estimated spillover effects using the entire sample of indus-

tries. The estimates are substantial and significant, 1.15 (10 percent). Columns

(2) estimates the model using only non-targeted industries; and column (3), es-

timates spillovers for only targeted sectors. The results for the restricted sample

are similar in positive and similar in magnitude, though only significant for the

model restricted to targeted sectors.

Table 2.12 presents estimates from a similar differences-in-differences

specification to Table 2.11 but using a total (Leontief) forward linkage measure.

Forward linkage effects (columns 1-3) are much stronger than the direct effects

of Table 2.11. In particular, the estimated effect of total forward linkages

(column 1) is stronger, 1.354 (5 percent level significance), than direct linkage

effects. When restricting the model to only non-targeted sectors, the effect is

much stronger and highly significant: 3.742 (1 percent significance), compared

to the much weaker effect of direct linkages on non-targeted sectors.

Table 2.13 reports estimates for other industrial growth outcomes, such

as employment and entry. Column (1) shows that strong forward linkages are

significantly tied to the entry of new establishments: 1.203 at 1 percent level

of significance. Column (3) shows a corresponding 1.694 estimate (1 percent

82For example, estimates for second-degree effects (not shown) are about half the size of

direct effects and insignificant.


significance) for employment.

Forward-Linkages, Prices, and Mechanisms Prediction 3 also suggests

that a supply shock in targeted industries also decreases the output price of

downstream sectors. Table 2.14 shows the relative output prices of forward-

linked industry fall significantly during the HCI period. Column (1) shows

conventional differences-in-difference estimates for the effect of forward link-

ages from targeted sectors. Sectors with strong forward linkages experience

a significant decline in the price of their output, relative to sectors with weak

linkages: a point estimate of -.43 (1 percent significance). Estimates are stronger

and significant if I use a total forward linkage measure.

If HCI policy positively affected downstream industries, it should have

done so by providing cheaper domestic intermediate inputs. One indication

of this effect, would be to see increased purchases of intermediate goods by

forward-link industries.

Accordingly, Table 2.14, columns (3) and (4) corroborate the mechanisms

behind the positive downstream spillovers. Indeed, forward linked sectors

appear to purchase more intermediate materials and capital goods than sectors

less reliant on HCI intermediates. Point estimates for material cost growth and

capital investment growth are both 1.2 and highly significant (1 percent level).

Inventory investments, both for semi-finished products (column 5) and raw

materials (column 6) also increase significantly more for forward-link sectors.

Together, the preliminary analysis of mechanisms hints to the potential

pecuniary externalities highlighted by Murphy et al. (1989) and Ciccone (2002),

as well as big push scholars (Hirschman, 1958). The relationship between

equipment investment and growth is one of the strongest relationships in the

cross-country growth literature (Sala-I-Martin, 1997; Hsieh, 2001). Specifically,

DeLong and Summers (1991), DeLong and Summers (1993), and Bond et

al. (2010) point to the role of equipment investment and growth. Focusing on

relative prices, complementary studies by Jones (1994), Jovanovic and Rob

(1997), and Restuccia and Urrutia (2001) show a negative relationship between

equipment prices and growth.

Backward-Linkages and Growth (Prediction 4) Since Hirschman (1958),


following the 1979 assassination of Park.

Table 2.11 columns (4-6) illustrate the potential negative effect of HCI

policy on direct upstream suppliers. As before, these tables present the average

linkage effect from a standard differences-in-differences version of the dynamic

specification in Equation (2.6.3). Columns (6) reports a strong negative average

effect of backward linkages using the full sample of industries (and controlling

for targeted and non-targeted sectors): -1.322 (10 percent). While the estimate

is stable when restricting the sample to non-targeted industries (columns 8), the

spillover effect is positive and insignificant for targeted industries (column 9).

Accounting for total backward linkages, Table 2.12 columns (4)-(6) also re-

ports a negative effect of HCI on sectors with strong backward linkages, relative

to sectors with weak links. All estimates are insignificant. Point estimates using

the entire sample (column 4) are much weaker, but nonetheless negative: -0.245.

Restricting the sample to non-targeted industry only, the effect of backward

linkages is stronger (-0.486), though insignificant.

The negative effects of HCI on domestic suppliers is also reflected in

differences-in-differences estimates using other industrial development out-

comes. For instance, Table 2.13 column (2) shows a large relative decline in

employment, -0.975, though the effect is insignificant.

Backward-Linkages and Import Competition The preceding results show

evidence that domestic suppliers with strong connections to targeted sectors

shrank relative to those with weak connections. One possible reason, suggested

by the HCI policy context and my model, is that the big push allowed targeted

sectors to import inputs, which may have negatively affected domestic industry.

Figure 2.15 illustrates why HCI may have negatively impacted backward-

linked producers. For 1962-1973 and 1973-1986, I show the simple bivari-

ate relationship between the value of imports and the strength of backward

connections from non-targeted to targeted industry. Before 1973, there is no

relationship between manufacturing industries with backward linkages and the

value of imports. The estimated coefficient is slightly negative and insignificant:

β � −1.8619 (t � −1.161). After 1973, however, there is a positive and signifi-

cant relationship between industries with connections to targeted industries and

the value of imports: β � 4.828 (t � 4.118). The pattern is consistent with a


Together, these findings are consistent with a story that domestic suppliers

were negatively impacted by import competition. Interestingly, Table 2.15, col-

umn (1) also shows that sectors with strong forward linkages increased exports,

relative to unconnected industries. This evidence contrasts with the general

findings of Blonigen (2016), who shows cross-country evidence that industrial

policy, specifically interventions targeting steel, hurt the export performance of

downstream (forward-linked) industry.

In summary, it is far from conclusive that industrial policies like HCI,

which require the importation of foreign inputs, (relatively) benefited upstream

suppliers. There is evidence that HCI may have sacrificed upstream sectors

to the benefit of downstream producers, by virtue of enabling key sectors to

liberally import key inputs from abroad.

At face value, negative results for backward-linked industry seems counter-

intuitive. Scholars like Albert Hirschman, stressed the importance of backward

linkages in industrial development.83 In the HCI context, however, targeted

firms were allowed to freely import many raw materials and intermediate

goods. In a small (relatively) open economy setting like South Korean setting,

instead of receiving a positive demand shock from targeted industries, upstream

sectors were subjected to increased competition as targeted sectors expanded

and increased their use of imported intermediates.

Direct Effects with Linkages Section 2.5.3 showed that HCI sectors directly

targeted by the big push grew significantly more than other sectors, relative to

pre-policy levels. Does accounting for either forward or backward spillovers

alter estimates of the direct effects—e.g. estimates from specification 2.12?84

The grey points ( grey confidence bands ) in Figure 2.16 plot estimates of

Targeted×Time from the main flexible differences-in-differences specification

for direct effects; the red points ( pink bands ) plot this same model, but

including both the Forward HCI Linkage and the Backward HCI Linkage

measures in specification 2.12.

Side-by-side, Figure 2.16 shows estimates from the two models are strik-

83See Backward Linkages at Work [Hirschman (1958); p.109-113].84The existence of either forward or backward spillovers from the industrial policy may alter

the differences-in-differences assumption: that the targeted treatment is contained only to treated

sectors.

2.7. CONCLUSION 65

ingly similar. The implication: accounting for first-order linkage effects does

not significantly change the pattern of the direct effects. Estimates from the

specification with linkages are only slightly lower for most years and generally

less precise. Nonetheless, accounting for first-degree linkage effects—the domi-

nant spillover–does not fundamentally modify the results for the direct effect of

HCI on industrial growth.

One reason for the similarity may be that the (positive) forward linkage

effects and (negative) backward linkage effects cancel out, in which case the

control group direct effect estimates is not polluted by spillovers from the

treated sector.

2.7 Conclusion

In this paper, I study a seminal event in post-war economic development,

South Korea’s rapid industrialization. Specifically, I explore Park Chung Hee’s

Heavy Chemical and Industry big push (HCI, 1973-1979), a large-scale in-

dustrial policy that attempted to shift Korea from a light exporting economy

to a modernized industrial power capable of domestic arms production. This

paper shows that the ambitious intervention promoted industrial development in

manufacturing sectors targeted by the policy. In addition, I show the industrial

intervention had widescale ramifications. First, the big push created positive

effects in treated industries long after major elements of the policy were re-

trenched. Moreover, the regime’s policy mix created winners and losers in

sectors differentially linked to treated industries.

The role of industrial policy in the East Asian growth miracle has long been

debated by economists (Rodrik, 1995; Lal, 1983; Krueger, 1995). My study

provides some of the first estimates of the impact of infant industry policy on

industrial development.85 For example, real output of industries targeted by

the HCI big push grew 80 percent more relative to non-targeted manufacturing

industries during the policy period. Not only did Korea’s interventions promote

growth in real output, but also a permanent reallocation of economic activity

from light to heavy industrial sectors. This transformation of the Korean econ-

85Recent work by Juhasz (2016) provides some of the first causal estimates of industrial policy

using historic French data.

2.7. CONCLUSION 67

omy delivered a nearly 11 percent decline in the relative price of output in

treated sectors. I document that, contrary to popular wisdom, Korea relied on

capital subsidies and subsidies to imported intermediate inputs, rather than the

differential protection of treated industries. Finally, I show most of the direct

effects of industrial policy persist long after the de facto end date of the policy,

when South Korea began the process of liberalization.

Targeted industries impacted external industries through the input-output

network. Guided by the predictions of a multi-sectoral general equilibrium

model (Long Jr and Plosser, 1983; Acemoglu et al., 2016), I show the rel-

ative decline in the output price of treated sectors benefited forward linked,

or downstream, sectors. Specifically, downstream buyers with strong links to

treated sectors grow relatively more in terms of output, establishment entry, and

employment, than downstream industries with weak links. The relative price of

output in downstream sectors also decreased significantly for linked versus un-

linked sectors. Accordingly, I also provide evidence that these forward-linked

sectors invested more in capital and increased their purchases of intermedi-

ate goods. The combined results indicate that HCI industrial policy generate

positive pecuniary externalities to forward-linked sectors. These conclusions

agree with earlier theoretical studies of big push development policy (Murphy

et al., 1989) and research highlighting the potential spillovers from equipment

investment (DeLong and Summers, 1991).

Development scholars, such as Albert Hirschman, have long highlighted

the role of linkages in promoting industrialization, emphasizing the role of

backward linkages in producing demand for upstream producers. I find, however,

that HCI industrial policies had mixed effects on backward linked sectors. In

particular, upstream suppliers with strong links to targeted industry – e.g. raw

material producers – decline relative to those with weak links. I show the decline

of upstream industry arose from industrial policies that benefited targeted

industries, such intermediate import subsidies. I thus provide evidence that the

negative effects that HCI had on upstream industry resulted from increased

import competition, indicated by a marked rise in imports of intermediate goods

used by treated sectors. In other words, South Korean industrial policy sacrificed

more upstream sectors for the benefit of downstream sectors.

Together, this study unpacks the effects of South Korea’s influential heavy


industrial big push. My study’s findings correspond to rich qualitative arguments

posed by Wade (1990) and Amsden (1992), who argued that industrial policies

promoted post-war industrialization. Moreover, I also show that industrial

policies may have heterogeneous impacts on other industries through the input-

output network. These results update earlier work by Hirschman (1958) and

others, indicating that the effects of traditional policy prescriptions may be more

complex in a highly globalized economy. Nonetheless, the results of my study

should be interpreted with caution. While my study highlights the effects of

industrial policy on a multitude of industrial development outcomes — such

as output prices, output growth, and the reallocation of manufacturing activity

— I have not delved into issues of total factor productivity, which I investigate

deeper in an upcoming analysis. Similarly, a next step for future research would

be to fully account for the effects of industrial policy on aggregate welfare and

factor misallocation.

2.7. CONCLUSION 69

Acemoglu, D., Akcigit, U., & Kerr, W. (2016). Networks and the Macroe-

conomy: An Empirical Exploration. In M. Eichenbaum & J. Parker (Eds.),

NBER macroeconomics annual 2015, volume 30 (pp. 276–335). Chicago,

Illinois: University of Chicago Press.

Acemoglu, D., Autor, D., Dorn, D., Hanson, G. H., & Price, B. (2015). Import

Competition and the Great US Employment Sag of the 2000s. Journal of

Labor Economics, 34(S1), S141–S198.

Acemoglu, D., Carvalho, V. M., Ozdaglar, A., & Tahbaz-Salehi, A. (2012). The

Network Origins of Aggregate Fluctuations. Econometrica, 80(5), 1977–

2016. https://doi.org/10.3982/ECTA9623

Adelman, I. (1969). Practical Approaches to Development Planning: Korea’s

Second Five-Year Plan. (I. Adelman, Ed.) (p. 306). Baltimore, Maryland:

Johns Hopkins Press.

Aghion, P., Dewatripont, M., Du, L., Harrison, A. E., & Legros, P. (2015).

Industrial Policy and Competition. American Economic Journal: Macroeco-

nomics, 7(4), 1–32.

Amiti, M., & Davis, D. R. (2012). Trade, Firms, and Wages: Theory and

Evidence. The Review of Economic Studies, 79(1), 1–36.

Amiti, M., & Konings, J. (2007). Trade Liberalization, Intermediate Inputs, and

Productivity: Evidence from Indonesia. American Economic Review, 97(5),

1611–1638.

Amsden, A. (1992). Asia’s Next Giant: South Korea and Late Industrialization

(Second). New York, New York: Oxford University Press.

Atalay, E. (2015). How Important Are Sectoral Shocks? University of Wisconsin

- Madison.

Autor, D. H., Dorn, D., & Hanson, G. H. (2013). The China Syndrome: Local

Labor Market Effects of Import Competition in the United States. American

Economic Review, 103(6), 2121–2168.

Baldwin, R. E. (1969). The Case against Infant Industry Tariff Protection.


Journal of Political Economy, 77(3), 295–305.

Baldwin, R. E., & Robert-Nicoud, F. (2007). Entry and Asymmetric Lobby-

ing: Why Governments Pick Losers. Journal of the European Economic

Association, 5(5), 1064–1093.

Banerjee, A. V., & Duflo, E. (2014). Do Firms Want to Borrow More? Test-

ing Credit Constraints Using a Directed Lending Program. The Review of

Economic Studies, 81(2), 572–607. https://doi.org/10.1093/restud/

rdt046

Beason, R., & Weinstein, D. E. (1996). Growth, Economies of Scale, and

Targeting in Japan (1955-1990). The Review of Economics and Statistics,

78(2), 286–295.

Besley, T., & Persson, T. (2010). State Capacity, Conflict, and Development.

Econometrica, 78(1), 1–34.

Besley, T., & Persson, T. (2011). Pillars of Prosperity: The Political Economics

of Development Clusters. Princeton, New Jersey: Princeton University Press.

Blonigen, B. A. (2016). Industrial Policy and Downstream Export Performance.

The Economic Journal, (Forthcoming).

Bond, S., Leblebicioglu, A., & Schiantarelli, F. (2010). Capital accumulation

and growth: a new look at the empirical evidence. Journal of Applied Econo-

metrics, 25(7), 1073–1099. https://doi.org/10.1002/jae.1163

Boyce, J. K. (1993). The Philippines: The Political Economy of Growth and

Impoverishment in the Marcos Era. Honolulu, Hawaii: University of Hawaii

Press; MacMillan Press Ltd.

Carvalho, V. M. (2010). Aggregate Fluctuations and the Network Structure

of Intersectoral Trade. Centre de Recerca en Economia Internacional -

Universitat Pompeu Fabra.

Castley, R. (1997). Korea’s Economic Miracle: The Crucial Role of Japan

2.7. CONCLUSION 71

(First). London; Basingstoke, U.K.: Palgrave Macmillan UK.

Chang, H.-J. (2003). Globalization, Economic Development and the Role of the

State. London, UK; New York, NY: Zed Books.

Chenery, H. B., & Watanabe, T. (1958). International Comparisons of the

Structure of Production. Econometrica, 26(4), 487–521.

Cheng, T.-j. (1990). Political Regimes and Development Strategies. In G. Gereffi

& D. L. Wyman (Eds.), Manufacturing miracles: Paths of industrialization in

latin america and east asia (pp. 139–178). Princeton, New Jersey: Princeton

University Press.

Cheng, T.-j. (2001). Transforming Taiwan’s Economic Structure in the 20th

Century. In R. L. Edmonds & S. M. Goldstein (Eds.), Taiwan in the twentieth

century: A retrospective view (pp. 19–36). Cambridge, England: Cambridge

University Press.

Cheremukhin, A., Golosov, M., Guriev, S., & Tsyvinski, A. (2013). Was Stalin

Necessary for Russia’s Economic Development. National Bureau of Eco-

nomic Research.

Chibber, V. (2002). Bureaucratic Rationality and the Developmental State.

American Journal of Sociology, 107(4), 951–989.

Cho, Y. J. (1989). Finance and Development: The Korean Approach. Oxford

Review of Economic Policy, 5(4), 88–102.

Cho, Y.-J., & Cole, D. C. (1986). The Role of the Financial Sector in Korea’s

Structural Adjustment. Harvard Institute for International Development,

Harvard University.

Cho, Y.-J., & Kim, J.-K. (1995). Credit Policies and the Industrialization of

Korea. Washington, D.C.: The World Bank.

Choi, J. J. (1990). Labor and the Authoritarian State: Labor Unions in South

Korean Manufacturing Industries, 1961-1980. Honolulu, Hawaii: University


of Hawaii Press.

Choi, T. Y., & Lee, S. G. (1989). Effect analysis of US military aid to the Repub-

lic of Korea (Declassified). Naval Postgraduate School; Naval Postgraduate

School Archive.

Choi, Y. (1985). The North Korean Military Buildup and Its Impact on North

Korean Military Strategy in the 1980s. Asian Survey, 25(3), 341–355.

Ciccone, A. (2002). Input Chains and Industrialization. The Review of Economic

Studies, 69(3), 565–587.

Congressional Budget Office. (1983). The Industrial Policy Debate (CBO stud-

ies). Washington, D.C.: The Congress of the United States, Congressional

Budget Office.

Conley, T., & Dupor, B. (2003). A Spatial Analysis of Sectoral Complementarity.

Journal of Political Economy, 111(2), 311–352.

Cox, R. W., & Skidmore-Hess, D. (1999). US Politics and the Global Economy:

Corporate Power, Conservative Shift. Lynne Rienner Publishers.

Criscuolo, C., Martin, R., Overman, H., & Reenen, J. V. (2012). The Causal

Effects of an Industrial Policy. National Bureau of Economic Research.

https://doi.org/10.3386/w17842

Cushman, J. H. (1979). The Military Balance in Korea. Asian Affairs: An

American Review, 6(6), 359–369.

Dell, M., Lane, N., & Querubin, P. (2016). State Capacity, Local Governance,

and Economic Development.

DeLong, J. B., & Summers, L. H. (1991). Equipment Investment and Economic

Growth. Quarterly Journal of Economics, 106(2), 445–502.

DeLong, J. B., & Summers, L. H. (1993). How Strongly do Developing

Economies Benefit from Equipment Investment? Journal of Monetary Eco-

nomics, 32(3), 395–415.

Eberstadt, N. (1999). The End of North Korea. Washington, D.C.: The AEI

2.7. CONCLUSION 73

Press.

Eberstadt, N. (2007). The North Korean Economy: Between Crisis and Catas-

trophe. New Brunswick, New Jersey; London, U.K.: Transaction Publishers.

Eklof, S. (2002). Politics, Business, and Democratrization in Indonesia. In E. T.

Gomez (Ed.), Political business in east asia (First, pp. 216–249). London,

England; New York, New York: Routledge; Taylor & Francis Group.

Evans, P. B. (1995). Embedded Autonomy: States and Industrial Transformation.

Princeton, New Jersey: Princeton University Press.

Feenstra, R. C., Inklaar, R., & Timmer, M. P. (2015). The Next Generation of

the Penn World Table. American Economic Review, 105(10), 3150–3182.

Fisman, R., & Miguel, E. (2010). Economic Gangsters: Corruption, Violence,

and the Poverty of Nations. Princeton, New Jersey: Princeton University

Press.

Flam, H., Persson, T., & Svensson, L. (1983). Optimal Subsidies to Declin-

ing Industries: Efficiency and Equity Considerations. Journal of Public

Economics, 22, 327–345.

Forslid, R., & Midelfart, K. H. (2005). Internationalisation, industrial policy

and clusters. Journal of International Economics, 66(1), 197–213.

Fukuyama, F. (2014). Political Order and Political Decay, from the industrial

revolution to the Globalization of Democracy (First, pp. 1–139). New York,

New York: Farrar, Straus,; Giroux.

General Agreement on Tariffs and Trade: Balance-of-Payments Committee.

(1978). GATT: 1978 Consulation Under Article XVIII:12(b) with Korea.

General Agreement of Tariffs; Trade.

Goh, E. (2004). Constructing the US Rapprochement with China, 1961–1974:

From ’Red Menace’ to ’Tacit Ally’. Cambridge University Press.

Goldberg, P. K., & Maggi, G. (1999). Protection for Sale: An Empirical Investi-


gation. The American Economic Review, 89(5), 1135–1155.

Gray, H. P. (1973). Senile Industry Protection: A Proposal. Southern Economic

Journal, 39(4), 569–574.

Grossman, G. M. (1990). Promoting New Industrial Activities: A Survey Of

Recent Arguments And Evidence. OECD Journal: Economic Studies, (14),

87–125.

Grossman, G. M., & Helpman, E. (1994). Protection for Sale. The American

Economic Review, 84(4), 833–850. Retrieved from +

Haggard, S. (1990). The Politics of Growth in the Newly Industrializing Coun-

tries (First). Ithaca, New York; London, U.K.: Cornell University Press.

Hamm, T.-y. (1999). Arming the Two Koreas: State, Capital and Military Power

(Vol. 10). London, England; New York, New York: Routledge.

Han, S. (1978). South Korea 1977: Preparing for Self-Reliance. Asian Survey,

18(1), 45–57.

Harrison, A. E. (1994). An Empirical Test of the Infant Industry Argument:

Comment. The American Economic Review, 84(4), 1090–1095.

Harrison, A. E., & Rodriguez-Clare, A. (2009). Trade, Foreign Investment, and

Industrial Policy for Developing Countries. In D. Rodrik & Mark Rosen-

zweig (Eds.), Handbook of development economics, volume 5 (pp. 4039–

4214). Amsterdam, The Netherlands: Elsevier B.V.; North-Holland.

Herrendorf, B., Rogerson, R., & Valentinyi, A. (2013). Growth and structural

transformation. National Bureau of Economic Research.

Hillman, A. L. (1982). Declining Industries and Political-Support Protectionist

Motives. American Economic Review, 72(5), 1180–1187.

Hindley, B., & Richardson, R. (1983). United Kingdom: Pulling Dragon’s Teeth

— the National Enterprise Board. In Brian Hindley (Ed.), State investment

companies in western europe (First, pp. 263–281). London; Basingstoke,

2.7. CONCLUSION 75

U.K.: MacMillan Press Ltd./Trade Policy Research Centre.

Hirschman, A. O. (1958). The Strategy of Economic Development (Third). New

Haven, Connecticut: Yale University Press.

Hirschman, A. O. (1968). The Political Economy of Import-Substituting Indus-

trialization in Latin America. The Quarterly Journal of Economics, 82(1),

1–32.

Holly, S., & Petrella, I. (2012). Factor Demand Linkages, Technology Shocks

and the Business Cycle. The Review of Economics and Statistics, 94(4),

948–963.

Hong, W. (1987). Export-Oriented Growth and Trade Patterns of Korea. In J.

Colin I. Bradford & W. H. Branson (Eds.), Trade and structural change in

pacific asia (pp. 273–306). Chicago, Illinois: University of Chicago Press.

Hong, W. (1992). Trade Policies in Korea. In D. Salvatore (Ed.), National trade

policies (pp. 401–422). Westport, CT: North-Holland.

Horikane, Y. (2005). The Political Economy of Heavy Industrialisation: The

Heavy and Chemical Industry (HCI) Push in South Korea in the 1970s.

Modern Asian Studies, 39(2), 369–397.

Hsieh, C. T. (2001). Trade Policy and Economic Growth: A Skeptic’s Guide to

the Cross-National Evidence (comment). NBER Macroeconomics Annual

2000, Volume 15, 15(January), 261–338.

Im, H. B. (2011). The Origins of the Yushin Regime: Machiavelli Unveiled. In E.

F. Vogel & B.-k. Kim (Eds.), The park chung hee era: The transformation of

south korea (pp. 233–261). Cambridge, Massachusetts: Harvard University

Press.

Innovation and Development Network, & Kim, S.-e. (2012). The National Fund

for Industry Development in South Korea. Seoul: Innovation; Development

Network (IDN), Seoul National University.

Itskhoki, O., & Moll, B. (2016). Optimal Development Policies with Financial


Frictions. Econometrica, (Forthcoming).

Janne E. Nolan. (1986). Military Industry in Taiwan and South Korea. London,

England: Palgrave Macmillan UK.



Jones, C. I. (1994). Economic growth and the relative price of capital. Journal

of Monetary Economics, 34(3), 359–382.

Jones, C. I. (2008). Intermediate Goods, Weak Links, and Superstars: A Theory

of Economic Development. NBER Working Paper Series, 13834(March),

1–48.

Jones, C. I. (2013). Misallocation, Economic Growth, and Input-Output Eco-

nomics. In D. Acemoglu, M. Arellano, & E. Dekel (Eds.), Advances in

economics and econometrics, tenth world congress of the econometric soci-

ety, volume ii. New York, New York: Cambridge University Press.

Jovanovic, B., & Rob, R. (1997). Solow vs. Solow: Machine Prices and Devel-

opment. National Bureau of Economic Research Working.

Juhasz, R. (2016). Temporary Protection and Technology Adoption: Evidence

from the Napoleonic Blockade.

Kamada, T., & Kawai, S. (1989). An Algorithm for Drawing General Undirected

Graphs. Information Processing Letters, 31(1), 7–15.


Korea and the Philippines (First). Cambridge, U.K.: Cambridge University

Press.

Kim, B.-k. (2011). The Leviathan: Economic Bureaucracy under Park. In B.-k.

Kim & E. Vogel (Eds.), The park chung hee era: The transformation of

south korea (pp. 200–233). Cambridge, Massachusetts; London, England:

Harvard University Press.

Kim, B.-k., & Im, H.-b. (2001). ’Crony Capitalism’in South Korea, Thailand

2.7. CONCLUSION 77

and Taiwan: Myth and Reality. Journal of East Asian Studies, 1(1), 5–52.

Kim, D., Oh, Y. S., Kim, K. H., Abrenca, M. J. V., Pinheiro, M. C., & Yun, C.

T. (2013). International Comparisons of Processes and Policies for Devel-

oping Heavy and Chemical Industries. Seoul: Korea Institute for Industrial

Economics; Trade.

Kim, E. M. (1988). From Dominance to Symbiosis: State and Chaebol in Korea.

Pacific Focus, 3(2), 105–121.

Kim, E. M. (1997). Big Business, Strong State: Collusion and Conflict in South

Korean Development, 1960-1990. Albany, New York: State University of

New York Press.

Kim, H.-A. (2004). Korea’s Development Under Park Chung Hee: Rapid Indus-

trialization, 1961-1979. Abingdon, UK; New York, NY: Routledge; Taylor

& Francis Group.

Kim, H.-A. (2011). Heavy and Chemical Industrialization, 1973-1979: South

Korea’s Homeland Security Measures. Reassessing the Park Chung Hee

Era, 1961-1979; Development, Political Thought, Democracy, & Cultural

Influence.

Kim, J. H. (1990). Korean Industrial Policy in the 1970’s: The Heavy and

Chemical Industry Drive. Seoul: Korea Development Institute.

Kim, J.-K., Shim, S. D., & Kim, J.-I. (1995). The Role of Government in

Promoting Industrialization and Human Capital Accumulation in Korea. In

T. Ito & A. O. Krueger (Eds.), Growth theories in light of the east asian

experience, nber-ease volume 4 (pp. 181–200). Chicago, Illinois: University

of Chicago Press.

Kim, K. S. (1994). Trade and Industrialization Policies in Korea: An overview.

In G. K. Helleiner (Ed.), Trade policy and industrialization in turbulent

times (First, pp. 317–363). New York, New York: Routledge.

Kim, K., & Leipziger, D. M. (1993). Korea: a Case of Government-Led Devel-


opment (Third). Washington, D.C.: World Bank Publications.

Kim, K.-D. (2005). Financing the Economic Development: The Role of Govern-

ment, Banks, Capital Market. In KIEP-university of sao paulo joint seminar.

Sao Paulo, Brazil.

Kim, S. J. (1970). South Korea’s Involvement in Vietnam and Its Economic and

Political Impact. Asian Survey, 10(6), 519–532.

Kim, S.-s. (2003). Labor Policy and Industrial Relations in the Park Chung-hee

Era. In B.-c. Lee (Ed.), Developmental dictatorship and the park chung-

hee era: The shaping of modernity in the republic of korea (pp. 153–184).

Paramus, New Jersey: Homa & Sekey Books.

Kim, S.-Y. (2001). Security, nationalism and the pursuit of nuclear weapons

and missiles: The South Korean case, 1970–82. Diplomacy and Statecraft,

12(4), 53–80.

Kohli, A. (2004). State-Directed Development: Political Power and Industrial-

ization in the Global Periphery. Cambridge, U.K.: Cambridge University

Press.

Kong, T. Y. (2000). The Politics of Economic Reform in South Korea: A Fragile

Miracle. London, England; New York, New York: Routledge; Taylor &

Francis Group.

Koo, B.-Y. (1984). The Role of the Government in Korea’s Industrial Devel-

opment: A paper submitted to the U.S.I.T.C. for its Study on ’Industrial

Targeting Practices in Korea’. Seoul: Korea Development Institute.

Krueger, A. O. (1979). The Developmental Role of the Foreign Sector and Aid

(p. 256). Cambridge, Massachusetts; London, England: Council on East

Asian Studies, Harvard University; Harvard University Press.

Krueger, A. O. (1990). Government Failures in Development. The Journal of

Economic Perspectives, 4(3), 9–23.

Krueger, A. O. (1995). East Asian Experience and Endogenous Growth Theory.

In Growth theories in light of the east asian experience, nber-ease volume 4

2.7. CONCLUSION 79

(pp. 9–36). Chicago, Illinois: University of Chicago Press.

Krueger, A. O., & Tuncer, B. (1982). An Empirical Test of the Infant Industry

Argument. American Economic Review, 72(5), 1142–1152.

Krugman, P. (1998). What’s new about the new economic geography? Oxford

Review of Economic Policy, 14(2), 7–17.

Krugman, P. R. (1993). The Current Case for Industrial Policy. In D. Salvatore

(Ed.), Protectionism and world welfare (pp. 160–179). Cambridge, U.K;


Kwack, T. (1984). Industrial Restructuring Experience and Policies in Korea in

the 1970s. Seoul: Korean Development Institute.

Kwack, T. (1985). Depreciation and Taxation of Income from Capital. Seoul,

Korea: Korean Development Institute.

Kwack, T., & Lee, K.-S. (1992). Tax Reform in Korea. In T. Ito & A. O.

Krueger (Eds.), The political economy of tax reform, nber-east asia seminar

on economics volume 1 (First, pp. 117–136). Chicago, Illinois; London,

England: University of Chicago Press.

Kwak, T. Y. (2003). The Nixon Doctrine and the Yusin Reforms: American

Foreign Policy, the Vietnam War, and the Rise of Authoritarianism in Korea,

1968–1973. Journal of American-East Asian Relations, 12(1), 33–34.

Lal, D. (1983). The Poverty of Development Economics. London, England:

Institute of Economics Affairs.

Lall, S. (1995). Malaysia: Industrial success and the role of the government.

Journal of International Development, 7(5), 759–773.

Lall, S. (1996). ‘The East Asian Miracle’ Study: Does the Bell Toll for Indus-

trial Strategy? In S. Lall (Ed.), Learning from the asian tigers: Studies in

technology and industrial policy (pp. 107–123). London, England: Palgrave

Macmillan UK.

Lawrence, R. Z., & Weinstein, D. E. (1999). Trade and growth: import-led or


export-led? Evidence from Japan and Korea. National bureau of economic

research.

Leal, J. (2016). Cross-Sector Misallocation with Sector-Specific Distortions.

The World Bank Economic Review, 30(2), 1–15.

Lee, C.-S. (1980). South Korea 1979: Confrontation, Assassination, and Transi-

tion. Asian Survey, 20(1), 63–76. https://doi.org/10.2307/2644008

Lee, J.-W. (1996). Government Interventions and Productivity Growth. Journal

of Economic Growth, 1(3), 391–414.

Lee, K. T. (1992). Korea’s Economic Development; Market-Pulled or Government-

Pushed? (pp. 1–60). Seoul: Korea Institute for Industrial Economics; Trade.

Lee, M. Y. (2011). The Vietnam War: South Korea’s Search for National Se-

curity. In B.-k. Kim & E. F. Vogel (Eds.), The park chung hee era: The

transformation of south korea (pp. 403–429). Cambridge, Massachusetts;

London, England: Harvard University Press.

Lee, N.-y. (2011). The Automobile Industry. In B.-K. Kim & E. F. Vogel (Eds.),

The park chung hee era: The transformation of south korea (pp. 295–321).

Cambridge, Massachusetts; London, England: Harvard University Press.

Lee, S.-C. (1991). The Heavy and Chemical Industries Promotion Plan (1973-

1979). In L.-J. Cho & Y. H. Kim (Eds.), Economic development in the

republic of korea: A policy perspective (pp. 431–471). Honolulu, Hawaii:

East-West Center, University of Hawaii Press.

Lim, H. (1998). Korea’s Growth and Industrial Transformation. New York,

New York: Macmillan Press; St. Partin’s Press, Inc.

Lin, J. Y. (2012). New Structural Economics: A Framework for Rethinking

Development and Policy. World Bank Research Observer, 26(2), 193–221.

Lin, J., & Chang, H. (2009). Should Industrial Policy in Developing Countries

Conform to Comparative Advantage or Defy it? A Debate Between Justin

2.7. CONCLUSION 81

Lin and Ha Joon Chang. Development Policy Review.

Lindbeck, A. (1981). Industrial Policy as an Issue in the Economic Environment.

The World Economy, 4(4), 391–406.

Long Jr, J. B., & Plosser, C. I. (1983). Real Business Cycles. The Journal of

Political Economy, 91(1), 39–69.

Luedde-Neurath, R. (1986). Import controls and Export-Oriented Development:

a Reassessment of the South Korean case. Boulder, Colorado; Londong,

England: Westview Press.

Manova, K., Wei, S.-J., & Zhang, Z. (2015). Firm exports and multinational

activity under credit constraints. Review of Economics and Statistics, 97(3),

574–588.

Markusen, J. R., & Venables, A. J. (1999). Foreign direct investment as a

catalyst for industrial development. European Economic Review, 43(2),

335–356.

Mason, E. S. (1980). The Economic and Social Modernization of the Republic of

Korea (Third). Council on East Asian Studies, Harvard University; Harvard

University Press.

Melo, J. de, & Roland-Holst, D. (1990). Industrial Organization and Trade

Liberalization: Evidence From Korea. Washington, D.C.: The World Bank.

Michishita, N. (2009). North Korea’s Military-Diplomatic Campaigns, 1966-

2008. London, England; New York, New York: Taylor & Francis.

Moon, C.-i., & Jun, B.-j. (2011). Modernization Strategy: Ideas and Influences.

In B.-K. Kim & E. F. Vogel (Eds.), The park chung hee era: The transfor-

mation of south korea (pp. 115–139). Cambridge, Massachusetts; London,


Moon, C.-i., & Lee, S. (2009). Military Spending and the Arms Race on the

Korean Peninsula. Asian Perspective, 33(4), 69–99.

Moreira, M. M. (1994). Industrialization, Trade and Market Failures: the Role of


Government Intervention in Brazil. Revista Brasileira de Economia, 48(3),

295–324.

Murphy, K. M., Shleifer, A., & Vishny, R. W. (1989). Industrialization and the

Big Push. Journal of Political Economy, 97(5), 1003–1026.

Myrdal, G. (1957). Economic Theory and Under-developed Regions. London,

England: G. Duckworth.

Nam, C.-H. (1980). Trade and Industrial Policies, and the Structure of Protec-

tion in Korea. Seoul: Korea Development Institute.

Nam, C.-H. (1995). The Role of Trade and Exchange Rate Policy in Korea’s

Growth. In T. Ito & A. O. Krueger (Eds.), Growth theories in light of the

east asian experience, nber-ease volume 4 (pp. 153–179). Chicago, Illinois:

National Bureau of Economic Research; University of Chicago Press.

Nam, J.-H. (1986). America’s Commitment to South Korea: The First Decade

of the Nixon Doctrine. Cambridge, England: Cambridge University Press.

Nam, S.-W. (1992). Korea’s financial reform since the early 1980s. In Gerard

Caprio, I. Atiyas, & James A. Hanson (Eds.), Financial reform: Theory and

experience (pp. 184–222). New York, New York: Cambridge University

Press.

Nelson, R. R., & Pack, H. (1998). The Asian Miracle and Modern Growth

Theory. The Economic Journal, 109(1881), 416–436.

Nixon, R. M. (1969). Public Papers of the Presidents of the United States -

Richard Nixon. Washington, D.C.: Office of the Federal Register, Govern-

ment Printing Office.

Nixon, R. M. (1970). Letter From President Nixon to Korean President Park. In

E. C. Keefer & C. Yee (Eds.), Foreign relations of the united states, 1969-

1976 (2010th ed., Vol. XIX, pp. 152–154). Washington, D.C.: United States

Government Printing Office.

Noland, M. (2004). Selective Intervention and Growth: The Case of Korea. In

M. G. Plummer (Ed.), Empirical methods in international trade: Essays in

2.7. CONCLUSION 83

honor of mordechai kreinin (pp. 229–246). Cheltenham, UK; Northampton,

MA: Edward Elgar.

Noland, M., & Pack, H. (2003). Industrial Policy in an Era of Globalization:

Lessons from Asia. Washington, D.C.: Institute for International Economics.

Noland, M., Robinson, S., & Wang, T. (2000). Rigorous Speculation: The

Collapse and Revival of the North Korean Economy. World Development,

28(10), 1767–1787.

Nunn, N., & Qian, N. (2011). The Potato’s Contribution to Population and Ur-

banization: Evidence From A Historical Experiment. The Quarterly Journal

of Economics, 126(2), 593–650.

Nunn, N., & Trefler, D. (2010). The Structure of Tariffs and Long-Term Growth.

American Economic Journal: Macroeconomics, 2(4), 158–194.

Nurkse, R. (1953). Problems of Capital Formation in Underdeveloped Countries.

Oxford, UK: Oxford University Press.

OECD. (2012). Industrial Policy and Territorial Development - Lessons from

Korea. Development Centre Studies, OECD Publishing.

Oh, J. K. C. (1976). South Korea 1975: A Permanent Emergency. Asian Survey,

16(1 (A Survey of Asia in 1975: Part I)), 72–81.

Oh, J. K. C. (1977). South Korea 1976: The Continuing Uncertainties. Asian

Survey, 17(1), 71–80.

Ohashi, H. (2005). Learning by doing, export subsidies, and industry growth:

Japanese steel in the 1950s and 1960s. Journal of International Economics,

66(2), 297–323.

Ostermann, C. F., & Person, J. F. (2011). The rise and fall of Detente on the

Korean Peninsula, 1970-1974. In C. F. Ostermann & J. F. Person (Eds.),

History and public policy program critical oral history conference series.

Washington, D.C.: Woodrow Wilson International Center for Scholars.

Pack, H., & Westphal, L. E. (1986). Industrial strategy and technological change:


Theory versus reality. Journal of Development Economics, 22, 87–128.

Park, T.-H. (1977). A Study on the Industry Tariff System of Korea - with refer

to Heavy & Chemical Industry Development, 4(5), 173–214.

Perkins, D. H. (2013). East Asian Development: Foundations and Strategies

(The Edwin O. Reischauer Lectures). Cambridge, Massachusetts; London,


Puga, D., & Venables, A. J. (1999). Agglomeration and Economic Develop-

ment: Import Substitution vs. Trade Liberalisation. The Economic Journal,

109(455), 292–311.

Rasmussen, P. N. (1956). Studies in Inter-Sectoral Relations (First). (Køben-

havn) Copenhagen, Denmark: E. Harck.

Restuccia, D., & Urrutia, C. (2001). Relative prices and investment rates. Jour-

nal of Monetary Economics, 47(1), 93–121.

Rhyu, S.-y., & Lew, S.-j. (2011). Pohang Iron & Steel Company. In The park

chung hee era: The transformation of south korea (pp. 322–344). Cambridge,

Massachusetts; London, England: Harvard University Press.

Robinson, J. A. (2010). Industrial Policy and Development: A Political Econ-

omy Perspective. In J. Y. Lin & B. Pleskovic (Eds.), Annual world bank

conference on development economics 2010 - global: Lessons from east asia

and the global financial crisis (pp. 61–79). Washington, D.C.: The World

Bank.

Rodriguez, F., & Rodrik, D. (2001). Trade Policy and Economic Growth: A

Skeptic’s Guide to the Cross-National Evidence. NBER Macroeconomics

Annual 2000, Volume 15, 15(January), 261–338.

Rodrik, D. (1995). Getting Interventions Right: How South Korea and Taiwan

Grew Rich. Economic Policy, 10(20), 53–107.

Rodrik, D. (1997). The ’paradoxes’ of the successful state. European Economic

2.7. CONCLUSION 85

Review, 41(3-5), 411–442.

Rodrik, D. (2004). Industrial Policy for the Twenty-First Century. John F.

Kennedy School of Government.

Rodrik, D. (2005). Growth Strategies. In P. Aghion & S. N. Durlauf (Eds.),

Handbook of economic growth (Vol. 1, pp. 967–1014). Amsterdam, The

Netherlands: Elsevier B.V.; North-Holland.

Rodriguez-Clare, A. (1996). Multinationals, Linkages, and Economic Develop-

ment. The American Economic Review, 86(4), 852–873.

Rogers, W. P. (1970). Telegram From the Department of State to the Embassy

in Korea. In E. C. Keefer & C. Yee (Eds.), Foreign relations of the united

states, 1969-1976 (2010th ed., Vol. XIX, pp. 150–152). Washington, D.C.:

United States Government Printing Office.

Rosenstein-Rodan, P. N. (1943). Problems of Industrialisation of Eastern and

South-Eastern Europe. The Economic Journal, 53(210/211), 202–211.

Rotemberg, M. (2015). Equilibrium Effects of Firm Subsidies.

Sala-I-Martin, X. X. (1997). I Just Ran Two Million Regressions. The American

Economic Review Papers & Preceedings, 87(2), 178–183.

Sawyer, M. (1992). Britain. In K. Cowling & R. Sugden (Eds.), Current is-

sues in industrial economic strategy. Manchester, U.K.; New York, N.Y.:

Manchester University Press.

Scalapino, R. A., & Lee, C.-S. (1972). Communism in Korea: The movement.

University of California Press.

Schmitz Jr, J. A. (2001). Government production of investment goods and

aggregate labor productivity. Journal of Monetary Economics, 47(1), 163–

187.

Scitovsky, T. (1954). Two Concepts of External Economies. Journal of Political

Economy, 62(2), 143–151.

Scobell, A., & Sanford, J. M. (2007). North Korea’s Military Threat: Py-


ongyang’s Conventional Forces, Weapons of Mass Destruction, and Ballistic

Missiles. Carlisle Barracks, Pennsylvania: U.S. Army War College Strategic

Studies Institute.

Shea, J. (2002). Complementarities and Comovements. Journal of Money,

Credit, and Banking, 34(2), 412–433.

Shim, J.-S., & Lee, M. (2008). The Korean Economic System: Governments,

Big Business and Financial Institutions. London, U.K.; New York, N.Y.:

Routledge; Taylor & Francis Group.

Smarzynska Javorcik, B. (2004). Does Foreign Direct Investment Increase the

Productivity of Domestic Firms? In Search of Spillovers Through Backward

Linkages. The American Economic Review, 94(3), 605–627.

Smith, H. (2000). Industry Policy in Taiwan and Korea in the 1980s: Winning

with the Market. Cheltenham, UK; Northampton, MA: Edward Elgar.

Stern, J. J., Kim, J.-h., Perkins, D. H., & Yoo, J.-h. (1995). Industrialization

and the State: The Korean Heavy and Chemical Industry Drive. Cambridge,

Massachusetts: Harvard Institute for International Development.

Succar, P. (1987). The Need for Industrial Policy in LDC’s - A Restatement

of the Infant Industry Argument. International Economic Review, 28(2),

521–534.

Suh, J. (2007). Overview of Korea’s Development Process until 1997. In J. Suh

& D. H. Chen (Eds.), Korea as a knowledge economy: Evolutionary process

and lessons learned (pp. 17–46). Washington, D.C.: The World Bank.

Taylor, W. J., Smith, J. A., & Mazarr, M. J. (1990). US Troop Reductions from

Korea, 1970–1990. The Journal of East Asian Affairs, 4(2), 256–286.

Trela, I., & Whalley, J. (1990). Taxes, outward orientation, and growth perfor-

mance in Korea. Washington, D.C.: The World Bank.

U.S. House. Committee on International Relations. Subcommittee on Interna-

tional Organizations. (1978). Investigation of Korean-American Relations:

Report of the Subcommittee on International Organizations of the Commit-

2.7. CONCLUSION 87

tee on International Relations, US House of Representatives. Washington,

D.C.: US Government Printing Office.

U.S. Senate. Committee on Foreign Relations. (1978). US Troop Withdrawal

From the Republic of Korea. Washington, D.C.: U.S. Government Printing

Office.

Vittas, D., & Wang, B. (1991). Credit Policies in Japan and Korea: A Review of

the Literature. Washington, D.C.: The World Bank.

Vogel, E. (1991). The Four Little Dragons: The Spread of Industrialization in

East Asia. Cambridge, Massachusetts; London, England: Harvard University

Press.

Vu, T. (2010). Paths to Development in Asia South Korea, Vietnam, China, and

Indonesia. New York, NY: Cambridge University Press.

Wade, R. H. (1990). Governing the Market: Economic Theory and the Role

of Government in East Asian Industrialization (Second). Princeton, New

Jersey: Princeton University Press.

Weinstein, D. E. (1995). Evaluating administrative guidance and cartels in

Japan (1957-1988). Journal of the Japanese and International Economies,

9, 200–223.

Werlin, H. (1991). Ghana and South Korea: Lessons from World Bank case

studies. Public Administration and Development, 11(3), 245–255.

Westphal, L. E. (1990). Industrial Policy in an Export Propelled Economy:

Lessons from South Korea’s Experience. Journal of Economic Perspectives,

4(3), 41–59.

Westphal, L. E., & Kim, K. S. (1982). Forstering Technological Mastery by

Means of Selective Infant Industry Protection. In B. Balassa (Ed.), Devel-

opment strategies in semi-industrial economies (pp. 212–279). Baltimore,

Maryland: Johns Hopkins University Press.

Westphal, L. E., Rhee, Y. W., & Pursell, G. (1981). Korean Industrial Compe-


tence: Where It Came From. Washington, D.C.: The World Bank.

White, L. T. (2009). Political Booms: Local Money and Power in Taiwan,

East China, Thailand, and the Philippines. Singapore: World Scientific

Publishing.

Woo, J.-e. (1991). Race to the Swift: State and Finance in Korean Industrializa-

tion. New York, New York: Columbia University Press.

World Bank. (1987). Korea - Managing the industrial transition : The conduct

of industrial policy (p. 204). Washington, D.C.: The World Bank.

Yoo, J.-h. (1990). The Industrial Policy of the 1970s and the Evolution of the

Manufacturing Sector in Korea (No. 9017). Seoul, Korea: Korea Develop-

ment Institute; Korea Development Institute.

Yoo, J.-h. (1993). The Political Economy of Protection Structure in Korea. In T.

Ito & A. O. Krueger (Eds.), Trade and protectionism, nber-ease volume 2

(pp. 361–385). Chicago, Illinois: University of Chicago Press.

Young, A. (1992). A Tale of Two Cities: Factor Accumulation and Technical

Change in Hong Kong and Singapore. In O. J. Blanchard & S. Fischer

(Eds.), NBER macroeconomics annual 1992 (pp. 13–64). Cambridge, Mas-

sachusetts: MIT Press.

2.7. CONCLUSION 89

Tabl

e2.

1:T

reat

edD

isag

gre

gat

edIn

dust

ries

,U

sing

(5D

igit

)1970

Indust

ryC

odes

and

Nam

es

Indust

ryN

ames

(K)S

ICIn

dust

ryN

ames

(K)S

ICIn

dust

ryN

ames

(K)S

IC

Cal

cium

carb

ide

35111

Photo

chem

ical

and

sensi

tize

dm

ater

ials

35296

Boil

ers

38212

Cau

stic

soda

35111

Pri

nti

ng

inks

35297

Far

mm

achin

ery

38220

Hydro

chlo

ric

acid

35111

Mis

cell

aneo

us

chem

ical

pro

duct

s35299

Mac

hin

eto

ols

for

work

ing

met

als

38231

Oth

erso

diu

mpro

duct

s35111

Gas

oli

ne

35301

Met

alw

ork

ing

mac

hin

ery

38234

Soda

ash

35111

Nap

hth

a35301

Min

ing

and

const

ruct

ion

mac

hin

ery

38241

Sulf

uri

cac

id35111

Fuel

oil

35302

Tex

tile

mac

hin

ery

38242

Anhydro

us

amm

onia

35112

Lubri

cati

ng

oil

san

dgre

ases

35302

Food

pro

duct

sm

achin

ery

38243

Oth

erin

dust

rial

com

pre

ssed

gas

es35112

Oth

erpet

role

um

pro

duct

s35309

Oth

ersp

ecia

lin

dust

rym

achin

ery

38249

Bas

icpet

roch

emic

alpro

duct

s35113

Bri

quet

tes

35401

Offi

cean

dse

rvic

ein

dust

rym

achin

es38250

Form

alin

35114

Dry

dis

till

ated

coal

pro

duct

s35402

Gen

eral

indust

rial

mac

hin

ery

38291

Oth

erac

ycl

icin

term

edia

tes

35114

Fer

roal

loys

37101

Gen

eral

mac

hin

ery

par

ts38292

Cycl

icin

term

edia

tes

35115

Pig

iron

37101

Ref

riger

ators

and

oth

erhouse

hold

appli

ance

s38293

Pig

men

ts35117

Raw

stee

l37101

Sew

ing

mac

hin

es38294

Synth

etic

dyes

tuff

s35117

Oth

erst

eel

roll

ing

and

dra

win

g37102

Gen

erat

ors

and

moto

rs38311

Oth

erin

org

anic

chem

ical

s35118

Ste

elbar

s37102

Tra

nsf

orm

ers

38312

Mis

cell

aneo

us

org

anic

chem

ical

s35119

Ste

elpla

tes

and

shee

ts37102

Oth

erel

ectr

ictr

ansm

issi

on

and

dis

trib

uti

on

equip

mnet

38313

Pro

cess

edoil

san

dfa

tspro

duct

s35119

Ste

elsh

apes

and

sect

ions

37102

Oth

erel

ectr

ical

indust

rial

appar

atus

38319

Nit

rogen

ous

fert

iliz

ers

35121

Ste

eltu

bes

and

pip

es37102

Com

munic

atio

ns

equip

men

t38324

Phosp

hat

icfe

rtil

izer

s35121

Cas

tir

on

tubes

and

pip

es37103

Ele

ctro

nic

com

ponen

ts38329

Cal

cium

cyan

amid

e35122

Iron

and

stee

l-ca

stin

gs

37103

Rad

ioan

dte

levis

ion

sets

38329

Agri

cult

ura

lch

emic

als

35126

Gal

van

ized

stee

lpro

duct

s37109

House

hold

elec

tric

appli

ance

s38330

Oth

erch

emic

alfe

rtil

izer

s35126

Ste

elfo

rgin

gs

37109

Insu

late

dw

ire

and

cable

38391

Pet

role

um

synth

etic

resi

ns

35131

Copper

37201

Ele

ctri

cla

mps

38392

Poly

vin

yl

chlo

rides

35131

Gold

and

silv

erin

gots

37201

Sto

rage

and

pri

mar

ybat

teri

es38394

Ther

mose

ttin

gre

sins

35131

Oth

ernon-f

erro

us

met

alin

gots

37201

Oth

erel

ectr

ical

equip

men

tan

dsu

ppli

es38399

Chem

ical

fibre

s35133

Nonfe

rrous

roll

ing

and

dra

win

g37203

Ship

s,N

EC

38413

Pai

nts

and

alli

edpro

duct

s35210

Nonfe

rrous

cast

ings

37204

Ste

elsh

ips

38414

Soap

and

acti

ve

agen

ts35232

House

hold

met

alpro

duct

s38111

Rai

lroad

tran

sport

atio

neq

uip

men

t38421

Cosm

etic

san

dto

oth

pas

tean

dpow

der

35233

Tools

38112

Moto

rveh

icle

s38431

Per

fum

es35233

Met

alfu

rnit

ure

38120

Auto

mobil

ere

pai

r38432

Adhes

ives

35291

Str

uct

ura

lm

etal

pro

duct

s38130

Moto

rveh

icle

par

ts38432

Explo

sives

and

pro

duct

s35292

Mis

cell

aneo

us

met

alpro

duct

s38197

Mea

suri

ng

and

scie

nti

fic

inst

rum

ents

38512

Mat

ches

35293

Pri

me

mover

s38211


Table 2.2: Pre-1973 Industry Statistics, Non-HCI v. HCI

Variable HCI Mean St.Dev. Min Max Obs.

A. Industrial Statistics (Ln)Costs Non-Targeted 1.75 2.37 0.00 7.81 3009

Costs Targeted 1.84 2.59 0.00 8.73 1547

Establishments Non-Targeted 1.78 3.52 0.00 8.37 3009

Establishments Targeted 1.66 3.41 0.00 7.48 1547

Gross Output Non-Targeted 2.65 5.59 0.00 10.80 3009

Gross Output Targeted 2.80 5.76 0.00 12.60 1547

Prices Non-Targeted 0.67 3.36 1.10 5.33 3009

Prices Targeted 0.81 3.60 1.01 5.88 1547

Labor Productivity Non-Targeted 0.14 0.12 -0.03 1.50 3009

Labor Productivity Targeted 0.25 0.15 0.00 2.45 1547

Inventory Non-Targeted 3.36 2.31 0.00 11.89 3009

Inventory Targeted 3.61 2.51 0.00 12.82 1547

Average Size Non-Targeted 0.03 0.03 0.00 0.61 3009

Average Size Targeted 0.02 0.02 0.00 0.14 1547

Shipments Non-Targeted 2.67 5.55 0.00 10.79 3009

Shipments Targeted 2.81 5.73 0.00 12.60 1547

Investment Non-Targeted 2.05 2.47 0.00 7.84 3009

Investment Targeted 2.24 2.89 0.00 9.71 1547

Value Added Non-Targeted 2.44 4.85 0.00 10.55 3009

Value Added Targeted 2.52 4.96 0.00 10.95 1547

Average Wages Non-Targeted 0.01 0.00 0.00 0.37 3009

Average Wages Targeted 0.01 0.00 0.00 0.18 1547

Workers Non-Targeted 2.76 6.97 0.00 12.39 3009

Workers Targeted 2.77 6.96 0.00 12.36 1547

B. LinkagesBackward Linkage, From Targeted Non-Targeted 0.17 0.80 0.13 1.01 3009

Backward Linkage, From Targeted Targeted 0.20 0.45 0.22 0.98 1547

Backward Linkage, From Targeted Non-Targeted 0.14 0.17 0.00 0.87 3009

Backward Linkage, From Targeted Targeted 0.21 0.49 0.02 0.76 1547

Forward Linkage, To Targeted Non-Targeted 0.24 0.84 0.00 1.00 3009

Forward Linkage, To Targeted Targeted 0.23 0.74 0.00 1.00 1547

Forward Linkage, To Targeted Non-Targeted 0.20 0.09 0.00 1.00 3009

Forward Linkage, To Targeted Targeted 0.21 0.19 0.00 0.92 1547

C. Trade Statistics (Ln)Value Exports (Sitc4 Products) Non-Targeted 7.03 2.82 0.69 14.49 10738

Value Exports (Sitc4 Products) Targeted 6.48 2.34 0.69 12.64 468

Value Imports (Sitc4 Products) Non-Targeted 7.43 2.58 0.69 15.67 10787

Value Imports (Sitc4 Products) Targeted 7.73 2.55 0.69 13.05 463

Quantitative Restrictions Output Non-Targeted 0.37 0.51 0.00 1.10 3009

Quantitative Restrictions Output Targeted 0.25 0.37 0.00 1.10 1547

Tariff Output Non-Targeted 0.54 3.81 2.40 5.02 3009

Tariff Output Targeted 0.45 3.33 1.52 4.45 1547

2.7. CONCLUSION 91

Tabl

e2.

3:D

iffe

ren

ces

inT

ota

lG

ross

Cap

ital

Inves

tmen

t&

Co

sts,

Bef

ore

-Aft

er1

97

3,

19

70

-19

86

Dep

end

ent

Var

iab

le(I

HS

):

To

tal

Cap

ital

Fo

rmat

ion

To

tal

Cap

ital

Fo

rmat

ion

To

tal

Cap

ital

Fo

rmat

ion

To

tal

Inp

ut

Co

sts

To

tal

Inp

ut

Co

sts

To

tal

Inp

ut

Co

sts

(1)

(2)

(3)

(4)

(5)

(6)

Tar

get

edX

Po

st0

.59

4*

**

0.6

67

**

*0

.68

3*

**

0.5

68

**

*0

.49

6*

**

0.4

93

**

*

(0.1

64

)(0

.16

2)

(0.1

64

)(0

.14

1)

(0.1

37

)(0

.13

6)

Co

nst

ant

1.7

41

2.1

54

2.1

19

2.6

46

2.0

08

2.0

04

(0.0

71

)(0

.33

8)

(0.3

51

)(0

.05

8)

(0.2

61

)(0

.27

0)

Ind

ust

ryF

ixed

Eff

ects

XX

XX

XX

Yea

rF

ixed

Eff

ects

XX

XX

XX

Bas

elin

eC

on

tro

lsX

XX

X

Tre

nd

sB

asel

ine

XX

R-S

qu

ared

0.8

14

0.8

21

0.8

27

0.8

71

0.8

82

0.8

90

Ob

serv

atio

ns

42

88

42

88

42

88

42

88

42

88

42

88

Clu

ster

s2

68

26

82

68

26

82

68

26

8

Not

e:D

iffe

rence

s-in

-Dif

fere

nce

ses

tim

ates

of

the

effe

ctof

Hea

vy

Chem

ical

and

Indust

ryin

dust

rial

targ

etin

gon

tota

lval

ue

of

gro

ssca

pit

alfo

rmat

ion

and

tota

l

val

ue

of

inte

rmed

iate

mat

eria

lspurc

has

es.

All

capit

aloutc

om

esar

edefl

ated

usi

ng

thei

rre

spec

tive

whole

sale

pri

cein

dex

.C

olu

mns

(1)-

(3)

report

esti

mat

esfo

r

capit

alac

quis

itio

ns;

colu

mns

(4)-

(6),

mat

eria

lco

sts.

All

spec

ifica

tions

incl

ude

indust

ryan

dyea

rfi

xed

effe

cts.

Colu

mns

(1)

and

(4)

corr

espond

toes

tim

ates

from

spec

ifica

tio

ns

wit

ho

ut

add

itio

nal

.C

olu

mn

s(2

)an

d(5

)in

clu

de

bas

elin

eco

ntr

ols

:p

re-1

97

3av

erag

esfo

r(I

HS

)em

plo

ym

ent,

lab

or

pro

du

ctiv

ity,

aver

age

wag

e,av

erag

e

cost

,av

erag

ees

tabli

shm

ent

size

,an

dav

erag

efi

xed

inves

tmen

t,ea

chin

tera

cted

flex

ibly

wit

hper

iod

effe

cts.

Inad

dit

ion,co

lum

ns

(3)

and

(6),

incl

ude

pre

-tre

nds

in

bas

elin

eco

ntr

ol

var

iab

les,

each

inte

ract

ion

wit

ha

per

iod

effe

cts.

Yea

ref

fect

sab

sorb

the

po

stp

erio

din

dic

ato

r;in

div

idu

alin

du

stry

fixed

affe

cts

abso

rbth

eTa

rget

eddum

my

var

iable

.R

egre

ssio

nlo

gsp

ecifi

cati

ons

are

esse

nti

ally

iden

tica

lan

dar

ein

cluded

inth

eA

ppen

dix

.R

obust

stan

dar

der

rors

are

clust

ered

on

the

5-d

igit

ind

ust

ry-l

evel

.S

tan

dar

der

rors

inp

aren

thes

es:

*p

<0

.05

,*

*p

<0

.01

,*

**

p<

0.0

01

.

Sour

ce:

Min

ing

and

Man

ufa

ctu

rin

gS

urv

ey&

Min

ing

and

Man

ufa

ctu

rin

gC

ensu

s:1

97

0-1

98

7.


Table 2.4: Differences in Investment Across Asset Class, Before & After 1973, 1970-1986

Dependent Variable (IHS) :

Acquisitions Building Acquisitions Machinery Acquisitions Land Acquisitions Vehicle

(1) (2) (3) (4)

Targeted X Post 0.485*** 0.631*** 0.335** 0.244*

(0.141) (0.152) (0.116) (0.106)

Constant 1.855 2.274 1.326 1.283

(0.210) (0.275) (0.147) (0.175)

Industry Fixed Effects X X X X

Year Fixed Effects X X X X

Baseline Controls X X X X

Trends Baseline X X X X

R-Squared 0.776 0.809 0.679 0.786

Observations 2680 2680 2680 2680

Clusters 268 268 268 268

Note: Differences-in-Differences estimates of the effect of Heavy Chemical and Industry industrial targeting on

different capital asset acquisitions. All variables and controls use an IHS transformation. Column (1) report estimates

for building and structural acquisitions; columns (2), equipment and machinery acquisitions; (3) land acquisitions;

and (4) vehicle acquisitions. Each have been deflated using a capital goods price index (2010 baseline values). All

regressions include period and 5-digit industry fixed effects. In additions all regression include the standard baseline

pre-treatment averages and pretrends interacted with time period effects. Regression log specifications are essentially

identical and are included in the Appendix. Robust standard errors are clustered on the 5-digit industry-level. Standard

errors in parentheses: *p<0.05, ** p<0.01, *** p<0.001.

Source: Mining & Manufacturing Survey and Mining & Manufacturing Census: 1970-1987. National Input-Output

Accounts, Bank of Korea, 1970.

2.7. CONCLUSION 93

Tabl

e2.

5:D

iffe

ren

ces

inP

rote

ctio

nP

oli

cy,

Bef

ore

-Aft

er1

97

3,

19

70

-19

82

Dep

enden

tV

aria

ble

(IH

S)

:

QR

Outp

ut

QR

Outp

ut

QR

Outp

ut

Tar

iff

Outp

ut

Tar

iff

Outp

ut

Tar

iff

Outp

ut

QR

Input

QR

Input

QR

Input

Tar

iff

Input

Tar

iff

Input

Tar

iff

Input

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

(9)

(10)

(11)

(12)

Tar

get

edX

Post

0.0

39

0.0

29

0.0

34

0.0

28

0.0

17

0.0

10

-0.0

45**

-0.0

44**

-0.0

41**

-0.2

16***

-0.2

03***

-0.2

01***

(0.0

47)

(0.0

47)

(0.0

48)

(0.0

28)

(0.0

27)

(0.0

27)

(0.0

14)

(0.0

14)

(0.0

14)

(0.0

43)

(0.0

41)

(0.0

40)

Const

ant

0.7

01

0.6

50

0.6

60

4.5

36

4.5

20

4.5

48

0.3

91

0.3

60

0.3

62

3.7

19

3.6

59

3.6

60

(0.0

19)

(0.0

83)

(0.0

85)

(0.0

10)

(0.0

37)

(0.0

36)

(0.0

06)

(0.0

24)

(0.0

24)

(0.0

12)

(0.0

24)

(0.0

24)

Indust

ryF

ixed

Eff

ects

XX

XX

XX

XX

XX

XX

Yea

rF

ixed

Eff

ects

XX

XX

XX

XX

XX

XX

Bas

elin

eC

ontr

ols

XX

XX

XX

XX

Tre

nds

Bas

elin

eX

XX

X

R-S

quar

ed0.7

74

0.7

81

0.7

86

0.9

59

0.9

61

0.9

63

0.8

81

0.8

85

0.8

93

0.9

74

0.9

77

0.9

78

Obse

rvat

ions

1340

1340

1340

1340

1340

1340

1340

1340

1340

1340

1340

1340

Clu

ster

s268

268

268

268

268

268

268

268

268

268

268

268

Not

e:D

iffe

ren

ces-

in-D

iffe

ren

ces

esti

mat

eso

fth

eef

fect

of

Hea

vy

Ch

emic

alan

dIn

du

stry

ind

ust

rial

targ

etin

go

nin

du

stri

alo

utp

ut.

All

ou

tco

mes

are

dafl

ecte

db

yin

du

stry

-lev

elp

rice

ind

ices

and

refl

ect

real

val

ues

.C

olu

mns

(1)-

(3)

report

resu

lts

for

val

ue

of

ship

men

ts;

colu

mns

(4)-

(6),

for

gro

ssoutp

ut;

colu

mns

(7)-

(9),

for

val

ue

added

.A

llsp

ecifi

cati

ons

incl

ude

indust

ryan

dyea

rfi

xed

effe

cts;

the

yea

ref

fect

sab

sorb

sth

epost

per

iod

indic

ator.

Colu

mns

(2),

(5),

and

(8)

incl

ude

pre

-1973

aver

ages

for

(IH

S)

emplo

ym

ent,

labor

pro

duct

ivit

y,av

erag

ew

age,

aver

age

cost

,av

erag

ees

tabli

shm

ent

size

,

and

aver

age

fixed

inves

tmen

t,ea

chin

tera

cted

flex

ibly

wit

hper

iod

effe

cts.

Colu

mns

(3),

(6),

and

(9)

incl

ude

pre

-tre

nds

inth

eaf

ore

men

tioned

bas

elin

eco

ntr

ol

vari

able

s,ea

chin

tera

ctio

nw

ith

aper

iod

du

mm

y.R

egre

ssio

nlo

gsp

ecifi

cati

on

sar

ees

sen

tial

lyid

enti

cal

and

are

incl

ud

edin

the

Ap

pen

dix

.R

obu

stst

and

ard

erro

rsar

ecl

ust

ered

on

the

5-d

igit

ind

ust

ry-l

evel

.S

tan

dar

der

rors

inp

aren

thes

es:

*p<

0.0

5,**

p<

0.0

1,***

p<

0.0

01.

Sour

ce:

Min

ing

&M

anufa

cturi

ng

Surv

eyan

dM

inin

g&

Man

ufa

cturi

ng

Cen

sus:

1970-1

987.T

arif

fsan

dP

rote

ctio

n,L

ued

de-

Neu

rath

,1986.


Table 2.6: Differences in Industrial Growth Relative to 1972, 1970-1986


Value Shipments Value Shipments Value Shipments Gross Output Gross Output Gross Output Value Added Value Added Value Added

(1) (2) (3) (4) (5) (6) (7) (8) (9)

Targeted X 1970 -0.041 -0.051 0.038 0.027 0.033 0.114 -0.002 0.005 0.095

(0.122) (0.124) (0.045) (0.127) (0.132) (0.066) (0.118) (0.123) (0.064)

Targeted X 1971 0.046 0.024 0.028 0.117 0.103 0.117 0.059 0.056 0.080

(0.127) (0.129) (0.097) (0.127) (0.130) (0.098) (0.106) (0.107) (0.089)

Targeted X 1972 - - - - - - - - -

Targeted X 1973 0.233 0.237 0.237 0.263* 0.268* 0.279* 0.255 0.320** 0.314**

(0.127) (0.125) (0.120) (0.125) (0.124) (0.119) (0.130) (0.122) (0.116)

Targeted X 1974 0.322** 0.327** 0.286* 0.298* 0.302* 0.266* 0.240* 0.243* 0.224

(0.122) (0.120) (0.120) (0.121) (0.117) (0.117) (0.118) (0.116) (0.120)

Targeted X 1975 0.351 0.246 0.300 0.234 0.037 0.063 0.165 0.004 0.018

(0.200) (0.196) (0.205) (0.233) (0.212) (0.213) (0.204) (0.191) (0.194)

Targeted X 1976 0.554* 0.402 0.429 0.576* 0.431 0.461 0.509* 0.395 0.432*

(0.242) (0.227) (0.235) (0.244) (0.232) (0.241) (0.214) (0.207) (0.216)

Targeted X 1977 0.607* 0.441 0.491* 0.630* 0.472* 0.525* 0.491* 0.371 0.427

(0.248) (0.227) (0.241) (0.247) (0.228) (0.242) (0.217) (0.204) (0.218)

Targeted X 1978 0.757** 0.618* 0.682** 0.794** 0.662** 0.730** 0.657** 0.559* 0.624**

(0.249) (0.239) (0.250) (0.251) (0.242) (0.254) (0.228) (0.223) (0.234)

Targeted X 1979 1.108*** 0.943*** 0.987*** 1.131*** 0.972*** 1.020*** 0.926*** 0.811*** 0.863***

(0.265) (0.237) (0.256) (0.266) (0.241) (0.259) (0.237) (0.221) (0.238)

Targeted X 1980 0.783** 0.619** 0.636** 0.806** 0.649** 0.670** 0.694** 0.578** 0.609**

(0.254) (0.238) (0.241) (0.252) (0.238) (0.242) (0.228) (0.220) (0.224)

Targeted X 1981 0.774** 0.608** 0.680** 0.792** 0.634** 0.707** 0.697** 0.581** 0.648**

(0.248) (0.232) (0.245) (0.249) (0.235) (0.247) (0.224) (0.216) (0.227)

Targeted X 1982 0.695** 0.525* 0.587* 0.721** 0.559* 0.619* 0.603* 0.479* 0.538*

(0.264) (0.247) (0.259) (0.263) (0.247) (0.259) (0.238) (0.227) (0.238)

Targeted X 1983 0.874** 0.726** 0.712** 0.892*** 0.751** 0.739** 0.719** 0.619** 0.610**

(0.264) (0.244) (0.243) (0.267) (0.251) (0.250) (0.241) (0.232) (0.232)

Targeted X 1984 0.945*** 0.807** 0.797** 0.968*** 0.837** 0.829** 0.853*** 0.758** 0.755**

(0.271) (0.253) (0.251) (0.274) (0.259) (0.257) (0.250) (0.239) (0.239)

Targeted X 1985 0.983*** 0.824** 0.797** 0.997*** 0.844** 0.820** 0.870** 0.760** 0.743**

(0.290) (0.271) (0.273) (0.293) (0.277) (0.279) (0.265) (0.256) (0.258)

Targeted X 1986 0.976** 0.816** 0.834** 0.991** 0.839** 0.860** 0.886** 0.776** 0.797**

(0.296) (0.275) (0.276) (0.299) (0.281) (0.282) (0.272) (0.260) (0.262)

Constant 4.989 3.079 3.046 5.011 3.191 3.159 4.278 2.911 2.867

(0.081) (0.440) (0.454) (0.082) (0.471) (0.487) (0.073) (0.418) (0.432)

Industry Fixed Effects X X X X X X X X X

Year Fixed Effects X X X X X X X X X

Baseline Controls X X X X X X

Trends Baseline X X X

R-Squared 0.841 0.858 0.864 0.829 0.848 0.854 0.831 0.849 0.856

Observations 4556 4556 4556 4556 4556 4556 4556 4556 4556

Clusters 268 268 268 268 268 268 268 268 268

Note: ’Fully-flexible’ differences-in-differences estimates of the effect of Heavy Chemical and Industry industrial targeting on industrial output, relative to 1972 baseline levels.

All outcomes are daflected by industry-level price indices and reflect real values. Columns (1)-(3) report results for value of shipments; columns (4)-(6), for gross output; columns

(7)-(9), for value added. All specifications include 5-digit industry and year fixed effects; the industry-level fixed effects absorb the targeted dummy variable. Columns (2), (5), and

(8) include pre-1973 averages for (IHS) employment, labor productivity, average wage, average cost, average establishment size, and average fixed investment, each interacted

flexibly with period effects. Columns (3), (6), and (9) include pre-trends in the aforementioned baseline control variables, each interaction with a period dummy variabl. These

estimates appear in the corresponding visualization figure. Regression log specifications are essentially identical and are included in the Appendix. Robust standard errors are

clustered on the 5-digit industry-level. Standard errors in parentheses: *p<0.05, ** p<0.01, *** p<0.001.

Source: Mining & Manufacturing Survey and Mining & Manufacturing Census: 1970-1987. Tariffs and Protection, Luedde-Neurath, 1986.

2.7. CONCLUSION 95

Tabl

e2.

7:D

iffe

ren

ces

inIn

du

stri

alG

row

th,

Bef

ore

-Aft

er1

97

3,

19

70

-19

86

Dep

end

ent

Var

iab

le(I

HS

):

Val

ue

Sh

ipm

ents

Val

ue

Sh

ipm

ents

Val

ue

Sh

ipm

ents

Gro

ssO

utp

ut

Gro

ssO

utp

ut

Gro

ssO

utp

ut

Val

ue

Ad

ded

Val

ue

Ad

ded

Val

ue

Ad

ded

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

(9)

Tar

get

edX

Po

st0

.71

0*

**

0.6

03

**

*0

.59

6*

*0

.67

3*

**

0.5

62

**

0.5

51

**

0.5

93

**

0.5

30

**

0.5

04

**

(0.1

91

)(0

.18

0)

(0.1

83

)(0

.19

7)

(0.1

85

)(0

.18

7)

(0.1

79

)(0

.17

3)

(0.1

73

)

Co

nst

ant

4.6

80

3.0

68

2.9

66

4.6

62

3.0

40

2.9

84

3.9

49

2.7

60

2.7

21

(0.0

86

)(0

.44

6)

(0.4

56

)(0

.09

3)

(0.4

72

)(0

.48

5)

(0.0

85

)(0

.41

9)

(0.4

31

)

Ind

ust

ryF

ixed

Eff

ects

XX

XX

XX

XX

X

Yea

rF

ixed

Eff

ects

XX

XX

XX

XX

X

Bas

elin

eC

on

tro

lsX

XX

XX

X

Tre

nd

sB

asel

ine

XX

X

R-S

qu

ared

0.8

39

0.8

58

0.8

65

0.8

27

0.8

47

0.8

54

0.8

29

0.8

49

0.8

56

Ob

serv

atio

ns

45

56

45

56

45

56

45

56

45

56

45

56

45

56

45

56

45

56

Clu

ster

s2

68

26

82

68

26

82

68

26

82

68

26

82

68

Not

e:D

iffe

rence

s-in

-Dif

fere

nce

ses

tim

ates

of

the

effe

ctof

Hea

vy

Chem

ical

and

Indust

ryin

dust

rial

targ

etin

gon

indust

rial

outp

ut.

All

outc

om

esar

edefl

ated

by

indust

ry-l

evel

pri

ce

indic

esan

dre

flec

tre

alva

lues

.C

olu

mns

(1)-

(3)

report

resu

lts

for

valu

eof

ship

men

ts;

colu

mns

(4)-

(6),

for

gro

ssoutp

ut;

colu

mns

(7)-

(9),

for

valu

ead

ded

.A

llsp

ecifi

cati

ons

incl

ude

ind

ust

ryan

dy

ear

fixed

effe

cts;

the

yea

ref

fect

sab

sorb

sth

ep

ost

per

iod

ind

icat

or.

Co

lum

ns

(2),

(5),

and

(8)

incl

ud

ep

re-1

97

3av

erag

esfo

r(I

HS

)em

plo

ym

ent,

lab

or

pro

du

ctiv

ity,

aver

age

wag

e,av

erag

eco

st,

aver

age

esta

bli

shm

ent

size

,an

dav

erag

efi

xed

inves

tmen

t,ea

chin

tera

cted

flex

ibly

wit

hper

iod

dum

my.

Colu

mns

(3),

(6),

and

(9)

incl

ude

pre

-tre

nds

in

the

afore

men

tioned

bas

elin

eco

ntr

ol

var

iable

s,ea

chin

tera

cted

wit

ha

per

iod

dum

my.

Reg

ress

ion

log

spec

ifica

tions

are

esse

nti

ally

iden

tica

lan

dar

ein

cluded

inth

eA

ppen

dix

.

Ro

bu

stst

and

ard

erro

rsar

ecl

ust

ered

on

the

5-d

igit

ind

ust

ry-l

evel

.S

tan

dar

der

rors

inp

aren

thes

es:

*p

<0

.05

,*

*p

<0

.01

,*

**

p<

0.0

01

.

Sour

ce:

Min

ing

and

Man

ufa

ctu

rin

gS

urv

ey&

Min

ing

and

Man

ufa

ctu

rin

gC

ensu

s:1

97

0-1

98

7.


Table 2.8: Differences in Labor Productivity, Before-After 1973, 1970-1986


Value Added Gross Output

Labor Prod. Labor Prod. Labor Prod. Labor Prod. Labor Prod. Labor Prod.

(1) (2) (3) (4) (5) (6)

Targeted X Post 0.025 0.029* 0.028* 0.092** 0.084** 0.084***

(0.015) (0.014) (0.012) (0.031) (0.028) (0.025)

Constant 0.081 0.080 0.095 0.170 0.177 0.207

(0.007) (0.022) (0.020) (0.012) (0.053) (0.049)

Industry Fixed Effects X X X X X X

Year Fixed Effects X X X X X X


Trends Baseline X X

R-Squared 0.808 0.836 0.856 0.825 0.854 0.866

Observations 4556 4556 4556 4556 4556 4556

Clusters 268 268 268 268 268 268

Note: Differences-in-Differences estimates of the effect of Heavy Chemical and Industry industrial targeting on

industrial labor productivity. All outcomes are daflected by industry-level price indices and reflect real values.

Columns (1)-(3) report estimates for value added labor productivity. Alternatively, columns (4)-(6) report gross

output labor productivity. All specifications include industry and year fixed effects; the year effects absorbs the

post period indicator. Columns (2), (5), and (8) include baseline controls. Columns (3), (6), and (9) include

pre-trends in the aforementioned baseline control variables, each interacted with a period dummy. Regression

log specifications are essentially identical and are included in the Appendix. Robust standard errors are clustered

on the 5-digit industry-level. Standard errors in parentheses: *p<0.05, ** p<0.01, *** p<0.001.

Source: Mining and Manufacturing Survey & Mining and Manufacturing Census: 1970-1987.

2.7. CONCLUSION 97

Tabl

e2.

9:D

iffe

ren

ces

inIn

du

stri

alO

utc

om

es,

Bef

ore

-Aft

er1

97

3,

19

70

-19

86

Dep

enden

tV

aria

ble

(IH

S)

:

Pri

ces

Pri

ces

Pri

ces

Avg.W

ages

Avg.W

ages

Avg.W

ages

Entr

yE

ntr

yE

ntr

yE

mplo

ym

ent

Em

plo

ym

ent

Em

plo

ym

ent

Lab

or

Shar

eL

abor

Shar

eL

abor

Shar

eS

har

eof

Outp

ut

Shar

eof

Outp

ut

Shar

eof

Outp

ut

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

(9)

(10)

(11)

(12)

(13)

(14)

(15)

(16)

(17)

(18)

Tar

get

edX

Post

-0.1

726***

-0.1

681***

-0.1

667***

0.0

008

0.0

001

0.0

002

0.3

241*

0.1

861

0.1

897

0.5

800*

0.3

783

0.3

786

0.0

758*

0.0

675*

0.0

632*

0.0

916**

0.0

839**

0.0

803*

(0.0

389)

(0.0

335)

(0.0

329)

(0.0

024)

(0.0

002)

(0.0

002)

(0.1

502)

(0.1

306)

(0.1

316)

(0.2

530)

(0.2

151)

(0.2

201)

(0.0

301)

(0.0

293)

(0.0

297)

(0.0

316)

(0.0

318)

(0.0

312)

Const

ant

3.3

223

3.4

422

3.4

315

0.0

057

0.0

004

0.0

007

3.6

454

2.2

696

2.2

978

6.7

478

4.4

773

4.4

653

0.2

929

0.1

650

0.1

653

0.2

145

0.1

274

0.1

314

(0.0

153)

(0.0

447)

(0.0

443)

(0.0

016)

(0.0

006)

(0.0

002)

(0.0

586)

(0.1

966)

(0.1

982)

(0.1

020)

(0.3

718)

(0.3

747)

(0.0

122)

(0.0

282)

(0.0

286)

(0.0

121)

(0.0

402)

(0.0

407)

Indust

ryF

ixed

Eff

ects

XX

XX

XX

XX

XX

XX

XX

XX

XX

Yea

rF

ixed

Eff

ects

XX

XX

XX

XX

XX

XX

XX

XX

XX

Bas

elin

eC

ontr

ols

XX

XX

XX

XX

XX

XX

Tre

nds

Bas

elin

eX

XX

XX

X

R-S

quar

ed0.9

44

0.9

53

0.9

57

0.2

71

0.9

01

0.9

45

0.8

57

0.8

84

0.8

87

0.7

92

0.8

25

0.8

29

0.8

97

0.9

05

0.9

08

0.8

93

0.9

01

0.9

07

Obse

rvat

ions

4552

4552

4552

4556

4556

4556

4556

4556

4556

4556

4556

4556

4556

4556

4556

4556

4556

4556

Clu

ster

s268

268

268

268

268

268

268

268

268

268

268

268

268

268

268

268

268

268

Not

e:D

iffe

rence

s-in

-Dif

fere

nce

ses

tim

ates

of

the

effe

ctof

Hea

vy

Ch

emic

alan

dIn

du

stry

ind

ust

rial

targ

etin

go

nin

du

stri

alla

bo

rp

rod

uct

ivit

y.A

llo

utc

om

esar

ed

aflec

ted

by

ind

ust

ry-l

evel

pri

cein

dic

esan

dre

flec

tre

alva

lues

.C

olu

mn

s(1

)-(3

)re

po

rtes

tim

ates

for

ou

tpu

tp

rice

s.C

olu

mn

s(4

)-(6

)re

port

aver

age

wag

es,

or

the

tota

l(r

eal)

wag

ebil

ld

ivid

edb

yin

du

stry

emp

loy

men

t.C

olu

mn

s(7

)-(9

)ar

efo

ren

try,

asm

easu

red

by

esta

bli

shm

ent

entr

y.C

olu

mn

s(1

0)-

(12

)ar

eto

tal

ind

ust

ryem

plo

ym

ent

esti

mat

es.

Co

lum

ns

(13

)-(1

5)

refl

ect

lab

or

stru

ctu

ral

chan

ge:

the

ind

ust

ryem

plo

ym

ent

asa

shar

eo

fto

tal

man

ufa

ctu

rin

gem

plo

ym

ent.

Sim

ilar

ly,

colu

mn

s(1

6)-

(18

)re

flec

to

utp

ut

stru

ctu

ral

chan

ge,

refl

ecte

das

real

gro

ssin

du

stry

ou

tpu

tas

shar

eo

fto

tal

man

ufa

ctu

rin

go

utp

ut.

All

spec

ifica

tio

ns

incl

ud

ein

du

stry

and

yea

rfi

xed

effe

cts.

Reg

ress

ion

log

spec

ifica

tio

ns

are

esse

nti

ally

iden

tica

lan

dar

ein

clu

ded

in

the

Appen

dix

.R

obust

stan

dar

der

rors

are

clust

ered

on

the

5-d

igit

indust

ry-l

evel

.S

tandar

der

rors

inpar

enth

eses

:*p<

0.0

5,**

p<

0.0

1,***

p<

0.0

01.

Sour

ce:

Min

ing

and

Man

ufa

cturi

ng

Surv

ey&

Min

ing

and

Man

ufa

cturi

ng

Cen

sus:

1970-1

987.


Table 2.10: Differences in Exports and Imports, Before-After 1973, 1970-1986


Import Value Export Value

(1) (2) (3) (4) (5) (6)

Targeted X Broadpost -0.4832 -0.2089 -0.2284 0.8070 1.0416* 1.0604*

(0.2706) (0.3350) (0.3327) (0.4420) (0.4954) (0.5017)

Constant 11.8400 8.9995 9.3343 11.3009 7.0224 6.6820

(0.0859) (0.6243) (0.7368) (0.1291) (1.2135) (1.6588)




Trends Baseline X X

R-Squared 0.891 0.900 0.901 0.856 0.878 0.880

Observations 2044 2044 2044 2044 2044 2044

Clusters 85 85 85 85 85 85

Note: Differences-in-Differences estimates of the effect of Heavy Chemical and Industry

industrial targeting on industrial labor productivity. All outcomes are daflected by industry-

level price indices and reflect real values. Columns (1)-(3) report estimates for value added

labor productivity. Alternatively, columns (4)-(6) report gross output labor productivity. All

specifications include industry and year fixed effects; the year effects absorbs the post period

indicator. Columns (2), (5), and (8) include baseline controls. Columns (3), (6), and (9)

include pre-trends in the aforementioned baseline control variables, each interacted with a

period dummy. Regression log specifications are essentially identical and are included in the

Appendix. Robust standard errors are clustered on the 5-digit industry-level. Standard errors

in parentheses: *p<0.05, ** p<0.01, *** p<0.001.

Source: Mining and Manufacturing Survey & Mining and Manufacturing Census: 1970-

1987.

2.7. CONCLUSION 99

Table 2.11: Impact of Direct Linkages on Industrial Growth, 1970-1986

Dependent Variable (IHS) Shipments

(1) (2) (3) (4) (5) (6)

Post X Forward HCI Linkage 1.051* 0.895 1.315*

(0.507) (0.736) (0.582)

Post X Backward HCI Linkage -1.224* -1.553* -0.492

(0.479) (0.611) (0.648)

Constant 4.989 4.833 4.381 4.989 4.833 4.381

(0.081) (0.111) (0.135) (0.080) (0.109) (0.135)



Sample Full Sample Non-Targeted Targeted Full Sample Non-Targeted Targeted

R-Squared 0.841 0.826 0.868 0.842 0.828 0.867

Observations 4556 3009 1547 4556 3009 1547

Clusters 268 177 91 268 177 91

Note: Shipments are the (real) value of shipments for each industry in a census year. Columns (1) and (4) estimate

the spillover effects on the entire sample–including but treated and non-treated sectors. Columns (2) and (5), examine

spillover effects for only non-targeted industries. Likewise, columns (3) and (6), do so for only targeted industries.

All specification include year and 5-digit industry fixed effects. Linkage measures are from pre-treatment (1970)

input-output accounts. The Forward HCI Linkage variable measures the total weighted share of intermediate inputs

purchased from treated sectors; Forward HCI Linkage, similarly captures the total weighted share of intermediates

sourced from non-treated sectors. Backward HCI Linkage measures the total weighted share of output sold to treated

sectors; Forward Non-HCI Linkage, similarly captures the total weighted share of intermediates sold to non-treated

sectors. Regression log specifications are essentially identical and are included in the Appendix. Robust standard errors

are clustered on the 5-digit industry-level. Standard errors in parentheses: *p<0.05, ** p<0.01, *** p<0.001.

Source: Mining and Manufacturing Survey & Mining and Manufacturing Census: 1970-1987. Bank of Korea, Input-

Output Accounts, 1970.


Table 2.12: Impact of Total (Leontief) Linkages to Policy on Industrial Growth, 1970-1986

Dependent Variable (IHS) Shipments

(1) (2) (3) (4) (5) (6)

Post X Leontief HCI Forward Linkage 1.354** 3.742*** 0.410

(0.417) (0.930) (0.389)

Post X Leontief HCI Backward Linkage -0.245 -0.486 0.302

(0.365) (0.504) (0.383)

Constant 4.989 4.833 4.381 4.989 4.833 4.381

(0.080) (0.107) (0.135) (0.081) (0.110) (0.134)



Sample Full Sample Non-Targeted Targeted Full Sample Non-Targeted Targeted

RSquared 0.842 0.829 0.867 0.841 0.826 0.867

Observations 4556 3009 1547 4556 3009 1547

Clusters 268 177 91 268 177 91

Note: Shipments are the (real) value of shipments for each industry in a census year. Each model is estimated using the full sample

of 5-digit industries.Total linkages measures are calculated from pre-treatment (1970) input-output accounts. The Leontief-based

linkage measures capture the total linkage effect of targeted or non-targeted sector output shifts on the output of other sectors,

accounting for N-order effects. The Leontief Forward HCI Linkage for an industry refers to row sums of the Leontief inverse matrix,

excluding non-targeted linkages. Leontief Forward Non-HCI Linkage refers to row sums of the Leontief inverse matrix, but only

for non-targeted industries. Leontief Backward HCI Linkage refers to column sums of the Leontief matrix, excluding non-targeted

linkages; Leontief Forward Non-HCI Linkage, includes only non-targeted industries. Regression log specifications are essentially

identical and are included in the Appendix. Robust standard errors are clustered on the 5-digit industry-level. Standard errors in

parentheses: *p< 0.05, ** p< 0.01, *** p< 0.001.

Source: Mining and Manufacturing Survey & Mining and Manufacturing Census: 1970-1987. Bank of Korea, Input-Output

Accounts, 1970.

2.7. CONCLUSION 101

Tabl

e2.

13:I

mp

act

of

Dir

ect

Lin

kag

eso

nIn

du

stri

alD

evel

op

men

tO

utc

om

es,

19

70

-19

86

Dep

end

ent

Var

iab

le(I

HS

):

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

En

try

En

try

Em

plo

ym

ent

Em

plo

ym

ent

Av

gW

ages

Av

gW

ages

Av

gS

ize

Av

gS

ize

Po

stX

Fo

rwar

dH

CI

Lin

kag

e1

.32

7*

**

1.5

14

*0

.01

50

.00

5

(0.3

63

)(0

.59

2)

(0.0

11

)(0

.00

8)

Po

stX

Bac

kw

ard

HC

IL

ink

age

-0.3

82

-1.1

84

*-0

.00

60

.01

3*

(0.3

05

)(0

.59

4)

(0.0

04

)(0

.00

6)

Co

nst

ant

3.6

19

3.6

19

6.8

07

6.8

07

0.0

04

0.0

04

0.0

31

0.0

31

(0.0

62

)(0

.06

2)

(0.1

02

)(0

.10

1)

(0.0

01

)(0

.00

1)

(0.0

01

)(0

.00

1)

Ind

ust

ryF

ixed

Eff

ects

XX

XX

XX

XX

Yea

rF

ixed

Eff

ects

XX

XX

XX

XX

Su

bsa

mp

leF

ull

Sam

ple

Fu

llS

amp

leF

ull

Sam

ple

Fu

llS

amp

leF

ull

Sam

ple

Fu

llS

amp

leF

ull

Sam

ple

Fu

llS

amp

le

R-S

qu

ared

0.8

59

0.8

58

0.7

93

0.7

93

0.2

79

0.2

74

0.5

25

0.5

26

Ob

serv

atio

ns

45

56

45

56

45

56

45

56

45

56

45

56

45

56

45

56

Clu

ster

s2

68

26

82

68

26

82

68

26

82

68

26

8


Tabl

e2.

14:L

ink

ages

and

(Mo

re)

Ind

ust

rial

Dev

elo

pm

ent,

Bef

ore

-Aft

er1

97

3,

19

70

-19

86

Dep

end

ent

Var

iab

le(I

HS

):

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

(9)

(10

)

Pri

ces

Pri

ces

Co

sts

Co

sts

Cap

ital

Acq

uis

itio

ns

Cap

ital

Acq

uis

itio

ns

Inven

tory

Ou

tpu

tIn

ven

tory

Ou

tpu

tIn

ven

tory

Inp

uts

Inven

tory

Inp

uts

Po

stX

Fo

rwar

dH

CI

Lin

kag

e-0

.31

0*

0.7

17

0.5

1.3

32

*1

.73

0*

*

(0.1

31

)(0

.36

9)

(0.4

43

)(0

.61

1)

(0.6

02

)

Po

stX

Bac

kw

ard

HC

IL

ink

age

0.5

17

**

*-0

.7*

*-0

.82

6*

-0.6

27

-0.2

44

(0.0

71

)(0

.28

5)

(0.3

31

)(0

.49

5)

(0.3

36

)

Co

nst

ant

3.1

83

3.1

84

2.4

60

2.4

60

1.6

55

1.6

55

3.1

91

3.1

91

2.6

95

2.6

95

(0.0

14

)(0

.01

4)

(0.0

64

)(0

.06

4)

(0.0

72

)(0

.07

2)

(0.1

06

)(0

.10

6)

(0.0

89

)(0

.09

0)

Ind

ust

ryF

ixed

Eff

ects

XX

XX

XX

XX

XX

Yea

rF

ixed

Eff

ects

XX

XX

XX

XX

XX

Su

bsa

mp

leF

ull

Sam

ple

Fu

llS

amp

leF

ull

Sam

ple

Fu

llS

amp

leF

ull

Sam

ple

Fu

llS

amp

leF

ull

Sam

ple

Fu

llS

amp

leF

ull

Sam

ple

Fu

llS

amp

le

R-S

qu

ared

0.9

47

0.9

49

0.8

69

0.8

69

0.8

02

0.8

02

0.5

35

0.5

35

0.4

90

0.4

89

Ob

serv

atio

ns

45

52

45

52

45

56

45

56

45

56

45

56

45

56

45

56

45

56

45

56

Clu

ster

s2

68

26

82

68

26

82

68

26

82

68

26

82

68

26

8

Not

e:P

rice

ou

tco

mes

are

ind

ust

ry-l

evel

pro

du

cer

pri

cein

dic

es,

har

mo

niz

edto

acco

un

tfo

rh

isto

ric

chan

ges

inin

du

stry

defi

nit

ion

s.A

llva

riab

les

inth

ese

model

suse

anin

ver

sehyper

boli

csi

ne

(IH

S)

tran

sform

atio

n.

The

cost

ou

tco

me

refl

ects

the

(rea

l)to

tal

cost

of

mat

eria

lin

pu

ts.S

imil

arly

,(r

eal)

tota

lin

ves

tmen

tre

flec

tth

eval

ue

of

val

ue

of

tota

lca

pit

alac

qu

isit

ion

sd

uri

ng

ace

nsu

sy

ear.

All

inven

tory

var

iab

les

are

refl

ect

chan

ge

in

inven

tori

es.

Ou

tpu

tin

ven

tori

esar

ech

ang

esin

un

ship

ped

fin

ish

edo

rse

mi-

fin

ish

edp

rod

uct

s;li

kew

ise,

mat

eria

lsin

ven

tori

esco

rres

po

nd

chan

ges

inin

term

edia

tein

pu

tst

ock

.E

ach

mo

del

ises

tim

ated

usi

ng

the

full

sam

ple

of

5-d

igit

ind

ust

ries

.Lin

kag

em

easu

res

are

fro

mp

re-t

reat

men

t,1

97

0in

pu

t-o

utp

ut

acco

un

ts.

Th

eFo

rwar

dH

CIL

inka

geva

riab

lem

easu

res

the

tota

lw

eig

hte

dsh

are

of

inp

ut

pu

rch

ased

fro

mta

rget

edse

cto

rs;

the

Bac

kwar

dH

CIL

inka

gevar

iab

les,

the

shar

eo

fto

tal

wei

gh

ted

sale

sto

targ

eted

sect

ors

.R

egre

ssio

nlo

gsp

ecifi

cati

on

sar

ees

sen

tial

lyid

enti

cal

and

are

incl

ud

edin

the

Ap

pen

dix

.R

obu

stst

and

ard

erro

rsar

ecl

ust

ered

on

the

5-d

igit

ind

ust

ry-l

evel

.S

tan

dar

der

rors

inp

aren

thes

es:

*p

<0

.05

,*

*p

<0

.01

,*

**

p<

0.0

01

.

Sour

ce:

Min

ing

and

Man

ufa

ctu

rin

gS

urv

ey&

Min

ing

and

Man

ufa

ctu

rin

gC

ensu

s:1

97

0-1

98

7.

Ban

ko

fK

ore

a,In

pu

t-O

utp

ut

Acc

ou

nts

,1

97

0.

2.7. CONCLUSION 103

Table 2.15: Linkages and Trade, Before-After 1973, 1962-1986


(1) (2) (3) (4)

Export Value Export Value Import Value Import Value

Post X Forward HCI Linkage 0.013 0.257

(1.095) (0.715)

Post X Backward HCI Linkage -2.911*** 2.475***

(0.592) (0.689)

Constant 2.313 2.368 8.394 8.373

(1.111) (1.025) (1.094) (1.016)

Industry Fixed Effects X X X X

Year Fixed Effects X X X X

Subsample Full Sample Full Sample Full Sample Full Sample

R-Squared 0.882 0.886 0.901 0.906

Observations 2044 2044 2044 2044

Clusters 85 85 85 85

Note: Differences-in-differences estimates of backward (forward) linkages from (to) targeted

industries. The cost outcome reflects the (real) total cost of material inputs on trade outcomes.

Columns (1)-(2) correspond to average estimates of linkages before-after HCI on the (real) value of

exports; columns (3) and (4) correspond to (real) value of imports. Columns (1) and (3) estimate

average effects of forward linkages to targeted industry; columns (2) and (4), backward linkages

from targeted industry.Linkage measures are from pre-treatment, 1970 input-output accounts. The

Forward HCI Linkage variable measures the total weighted share of input purchased from targeted

sectors; the Backward HCI Linkage variables, the share of total weights sales to targeted sectors.

Regression log specifications are essentially identical and are included in the Appendix. Robust

standard errors are clustered on the 5-digit industry-level. Standard errors in parentheses: *p<0.05,

** p<0.01, *** p<0.001.

Source: Mining and Manufacturing Survey & Mining and Manufacturing Census: 1970-1987. Bank

of Korea, Input-Output Accounts, 1970.

3. Waiting for the Great Leap

Forward - The Green Revolution

and Structural Change in the

Philippines *

3.1 Introduction

Revolutions–whether political or technological–often produce unintended

consequences. The green revolution, the march of biological crop innovations

following World War II, transformed agricultural productivity.1 Economists

long believed such agricultural productivity growth was essential to the ascent

of modern economic sectors–i.e. structural change (Gilboy, 1932; Johnston and

Mellor, 1961; Nurkse, 1953; Rostow, 1960).2 Nonetheless, despite large gains

to agricultural productivity around the world, the impact of the green revolution

on structural change has been mixed. Ironically, these ambiguities are most

prominent in the home of the green revolution: the Philippines.

This paper studies how the green revolution affected structural change in

*This paper has benefited from discussions with Melissa Dell, Suresh Naidu, Nathan Nunn,

and James Robinson1Evenson and Gollin (2003) understand the green revolution as a successive wave of innova-

tions starting in the 1960s (for rice), rather than a singular shock to productivity. The modern

varieties introduced in this era account for 21 percent of agricultural productivity growth (yields)

for 1961-1980, and 40 percent for 1981-2000 (i.e. early versus late green revolution periods).

Estimates of the effect of the green revolution on growth have been staggering: Gollin, Hansen,

and Wingender (2016) show that a 10 percentage points increase in HYV adoption increased

GDP by 15 percent.2Specifically, the interaction between agricultural productivity growth and the subsequent

demand for industrial goods by laborers in a closed economy. This chain of general equilibrium

effects has been referred to as the Mellor Hypothesis, starting with Johnston and Mellor (1961).

This view undoubtedly is related to the significant rise in British agricultural yields preceding

the industrial revolution (Allen, 2000; Clark, 1987,9).

105

106 Waiting for the Great Leap Forward

its home country. I trace out how the expansion of new high-yielding varieties,

known as HYVs or simply modern varieties, translated into increased agricul-

tural productivity and reallocated economic activity across sectors. Using the

introduction of the first wave of modern varieties in 1966, I explore the differ-

ential evolution of that adopted regions the new technology versus regions that

did not. I show how the expansion of green revolution technology changed the

evolution of employment in agriculture, manufacturing, and services over three

decades–in ways not always anticipated. Specifically, I show that green revo-

lution technological shocks produced different forms of sectoral reallocation

in the short and long run. By doing so, I reconcile theoretical predictions and

early qualitative observations as to the effects of green revolutions on peasant

labor demand.

This is not a study of technological serendipity, but rather an evaluation of a

grand agricultural intervention (Cullather, 2004,1). The Philippines, a cold war

American ally and recipient of generous Western development aid, became the

epicenter of the rice green revolution. The International Rice Research Institute

(IRRI)–a massive joint effort by the Rockefeller Foundation, Found Foundation,

and the Philippine government–was established in Los Banos in the early 1960s.

Political scientist, Lynn T. White referred to it as the “highest-profile technology

research program in the world" (White, 2009, 6). In 1966 the IRRI released

the first “miracle rice" varieties that defined the green revolution. The mod-

ernizing regime of Philippine President Ferdinand Marcos became the earliest

and most enthusiastic proponent of the IRRI’s innovations, coordinating the

mass adoption of the new technologies after their release. While the Philip-

pines experienced rapid growth in output and productivity in their principal

crop, structural change was slow to come. Agriculture remained the sector of

employment will into the 1980s.

I focuses on the rise and fall of agricultural employment. I show that

patterns of structural change in Philippine municipalities were different in the

short run and the long run. In the short run, the first fifteen years after the

green revolution began, I show that the expansion of new varieties increased

agricultural labor share, along with land-intensive practices and use of farm

capital. However, in the long run, these employment patterns reserved: after

the initial, short-run increase in demand for peasant labor, the agricultural


labor demand significantly declined with a commensurate increase in service

employment. While the classic land-augmenting (labor-biased) nature of green

revolution pulled labor into agriculture in the short run, in the the long run

this labor was displaced by an ever-mechanizing agriculture sector. I point out

empirical patterns that have been described by development scholars, but are

not captured by models of structural change.

The distinct short versus long-run effects of the green revolution in this

study fits early observations of the green revolution. Eminent U.S. policy advisor

and agricultural economist, Wolf Ladejinsky, anticipated the impacts of the

green revolution in the context of this study. Writing in a 1970 Foreign Affairs

article, Ladejinsky, encapsulates the findings of this study:

“The landless farm laborers, though their lot is temporarily im-

proved, are eventually due for a setback. The new type of agri-

culture is labor-intensive, employing more labor due to double-

cropping and other labor-demanding practices it is introducing.

Not surprisingly, therefore, it has been hailed as a solution of the

large problem of unemployment among rural landless. It appears,

however, that even in the most advanced state like Punjab this is

not as promising as anticipated because the technology is both

labor-absorbing and labor-displacing. ... [L]ooking ahead, addi-

tional employment and better wages are not forever, for new farm

practices are bringing in a host of labor-saving devices such as

tractors and threshers and much in between." (Ladejinsky, 1970,

764-765).

The contradictory nature of the green revolution, as first, labor absorbing and

second, labor displacing, were shared by observers from South (Bardhan, 1970;

Cleaver, 1972; Sanyal, 1983) to Southeast Asia (Boyce, 1993; Scott, 1986).3

For the Philippines, specifically, Coxhead and Jayasuriya (1986); Kikuchi and

Hayami (1983) each qualitatively described these patterns following the release

3Prime Minister Charan Singh, writing in 1977 as the Minister for Domestic Affairs voiced

similar concern, “the growing mechanization of agriculture since 1966 has introduced an aber-

ration in our agricutural development greatly limiting its capacity to subserve the objective of

absorbing optimum labor force” (Sanyal, 1983, 39).


of IR8 and HYVs.4 The paradoxical nature of new HYVs led political scientists

James C. Scott to conclude that in Malaysia "the poor have become redundant"

(Scott, 1986, 13).5

While an emergent empirical literature has started to study the effects of

agricultural productivity on structural change, none have accounted for these

nuances observed by early development economists. A series of papers, Foster

and Rosenzweig (2004,0), explore the impact of HYVs in India on the composi-

tion of rural employment. Like this study, they find that agricultural productivity

growth during the green revolution is negatively related to local factory em-

ployment (Foster and Rosenzweig, 2004) but positively related to the presence

of services Foster and Rosenzweig (2007). Bustos, Caprettini, and Ponticelli

(2016) study how new GMO varieties in Brazil increased the demand for labor.

The authors argue that technological change in maize–much like green revo-

lution rice varieties–introduced a second harvest season via land-augmenting

and labor-biased technological change increased agricultural labor demand. On

the other hand, the authors argue that innovations to soy amounted to labor-

augmenting technological change, showing these innovations reallocated labor

towards manufacturing.6 My study argues, like the early observers of the green

revolution, technical change in a single crop can have different effects on the

reallocation of labor across sectors–especially as landlords and small holders

alike respond to higher wages induced by increased labor demand.

4Ladejinsky’s observation is not idiosyncratic. Coxhead and Jayasuriya (1986) discuss the

two phases of the green revolution: “In the first phase of adoption of the new rice technology,

labor use actually increased as a result of greater labor demand for crop care activities ... Since

the mid-1970s, labor use patterns have shown a sharp change. In the irrigated areas where labor

use had increased earlier, new practices reduced labor use very significantly" (Coxhead and

Jayasuriya, 1986, 1058-1059). Similarly, (Kikuchi and Hayami, 1983), “Until 1975 the diffusion

of modem varieties was associated with an increase in labor demand for rice production; although

labor input for land preparation declined due to the concurrent diffusion of hand tractors, this

decline was more than compensated for by the sharp increase in labor use for weeding and for

other crop care needs. After 1975, the total labor input per hectare began to decline with the

introduction of such labor-saving practices as the use of herbicides and threshing machines"

(Kikuchi and Hayami, 1983, 248).5This is a statement about winners and losers from the green revolution, not about the net

welfare effects of the green revolution. Moreover, there is no doubt that the green revolution

often benefitted land holders. As well, there are critics of the view that the green revolution may

have decreased rural wages (See Lal (1976) for discussion).6A closely related paper by Marden (2017) shows how improvements to agricultural produc-

tivity in China, vis-a-vis institutional reforms in the late 1970s and early 1980s, supported the

rise of Chinese manufacturing.

3.2. HISTORICAL CONTEXT AND STYLIZED FACTS 109

Theoretical studies of structural change have emphasized agricultural pro-

ductivity growth as the engine behind structural change. In particular, a domi-

nant literature has highlighted how agricultural productivity growth combined

with non-homothetic preferences, stimulates a demand-driven shift of labor

from agriculture to manufacturing and services (Diao, McMillan, and Rodrik,

2017; Echevarria, 1997; Gollin, Parente, and Rogerson, 2002; Herrendorf,

Rogerson, and Valentinyi, 2014; Matsuyama, 1992).7 However, in other theo-

retical settings, technological change in agriculture may produce quite different

effects.8 None of these papers have considered how technological change, as

exemplified by the green revolution, may be both time labor absorbing and

labor displacing at different points in time.

This paper is organized as follows. Section 2 describes the historical context

of the green revolution and structural change in the Philippines. Section 3

describes the data and digitization effort behind this study. Section 4 describes

my empirical strategy. Section 4 presents the empirical narrative, first showing

the adoption of green revolution technologies; next, showing short-run effects of

the green revolution on structural change and mechanization; and last, showing

the long-run effects on structural change.

3.2 Historical Context and Stylized Facts

In this section, I juxtapose the historical evolution of the Philippine green

revolution with the country’s aggregate patterns of structural change. In doing so,

I drive home two points.First, the green revolution, starting with the introduction

of IR8 in 1966, had a remarkable, sustained effect on aggregate rice productivity.

Second, the Philippines experienced a significant increase in aggregate share of

agricultural employment following 1966, which started a slow decline in the

1980s. This declining agricultural workforce was absorbed by an an expanding

service sector. Remarkably, the manufacturing labor share never changed. In

other words, I juxtapose the giant technological change that occurred in the

Philippines, the epicenter of the green revolution, with peculiar patterns of

7Once again, these theoretical model can trace their intellectual origins to economic history

studies on the industrial revolution (Gilboy, 1932; Mokyr, 1977).8Factors such as the factor bias of technological change (Bustos et al., 2016), trade openness

(Matsuyama, 1992), and factor mobility (Foster and Rosenzweig, 2004,0).


reallocation of labor across sectors.9

A MANHATTAN PROJECT FOR FOOD The green revolution was not a spon-

taneous technological shock. Instead, it was the product of constellation of large

geopolitical forces and large institutions pushing to modernize the developing

world (Cullather, 2013; Gollin et al., 2016; Parayil, 2003; Smith, 2009). An

ambitious agronomic research project, the International Rice Research Institute

(IRRI), was established near the city of Los Banos, Philippines in 1960. The

brainchild of a joint effort by the Ford and Rockefeller Foundations, the IRRI

was created as “a Manhattan Project for food", one that rivaled the ambition

of the Marshall Plan (Cullather, 2004, 233).10 However, the IRRI promised

to modernize third-world agriculture by producing a giant leap in agricultural

productivity through modern genetics.

The first green revolution rice varieties arrived in 1966, after years of

intensive experimentation at IRRI. Known as "miracle rice," IR8-288-3, or

simply IR8, was a nitrogen-responsive cross-breed of Asian, lowland indica

rice, and marked a steady flow of 42 similar hybrid varieties the next 30 years.11

The responsiveness of the HYVs to nitrogen fertilizer is not trivial. Indica rice,

the common rice of the region, was widely understood to be non-responsive

to fertilizers: “improved fertilization [of traditional indica], for instance, will

lead mainly to vegetative growth and lodging rather than significantly increased

yield" (Dalrymple, 1978, 8).

Moreover, unlike traditional varieites, HYVs could be grown over shorter

periods of time and back-to-back, regardless of the season. Specifically, these

cultivars were photo period insensitive, meaning that they were not sensitive to

the length of nighttime and daylight, and thus be grown outside traditional rice

growing seasons. Accordingly, through the 1970s, the Philippine rice sector

9This pattern of structural change is an outlier in Asia. However, it is the similar pattern seen

across many contemporary African economies.10The green revolution wheat varieties had already began in the 1950s. In 1954, Norman

Borlaug invented strains of miracle” dwarf wheat in Mexico. Similarly, the International Maize

and Wheat Improvement Center was established in El Batan in 1966. Over a dozen or so

institutions would be established world wide, including International Institute for Tropical

Agriculture (IITA) in Nigeria, and the International Center for Tropical Agriculture (CIAT)

Colombia.11From 1966 to 1997, Peng, Cassman, Virmani, Sheehy, and Khush (1999) records the release

of 42 indica hybirds for irrigated and lowland rice agriculture.

3.2. HISTORICAL CONTEXT AND STYLIZED FACTS 111

Figure 3.1: Philippines President Ferdinand Marcos and U.S. President Lyndon

B. Johnson at the International Rice Research Institute, 1966

Notes: U.S. and Philippines presidents visit IRRI, October 1966. From right to left: (kneeling)

U.S. President Lyndon B. Johnson; (standing) Philippine President Ferdinand E. Marcos; Dr.

Robert F. Chandler, the founding director of IRRI; (standing), IR8 rice breeder, Peter Jennings;

and (kneeling) IR8 rice breeder, Hank Beachell. Photographed by IRRI photographer, Urbito

Ongleo.


experienced unprecedented gains in rice output (Ishikawa, 1970). General rice

production doubled and yields grew at an average annual rate of 5.3 percent

through the 1970s (Unnevehr, 1986; Unnevehr and Balisacan, 1983).

Figure 3.2 shows the increase in aggregate rice productivity, measured by

quantity of rice output (metric tons) per hectare, from the 1961 to 1990.12

Following the invention of IR8 in 1966 and its dissemination in 1967, waves of

improved varieties coincided with a dramatic increase rice yields.

Notably, the dips in the first half of the 1970s seen in Figure 3.2 are attributed

an unforeseen disasters: tungro virus outbreaks, 1970-1973, and typhoons in

1972 (Atkinson and Kunkel, 1976; Herdt and Capule, 1983).13 Later varieties

were disease/pest resistant and more resilient to extreme weather conditions,

such as IR26 (1973), IR30 (1976), and IR36 (1977) (Peng et al., 1999). In

particular, IR26 was the first cultivar with resistance disease and pests, such as

bacterial blight, blast, brown planthopper, and importantly, green planthopper,

the principal vector of transmission for the tungro virus. Clearly, after 1966

the Philippine Islands experienced a remarkable increase in the productivity of

their major agricultural crop.

Post-war politics meant the widespread adoption of green revolution varieties–

more so than any country during the green revolution rice breakthroughs. Ac-

cording to early agronomists at the International Rice Research Institute, mod-

ern varieties “were adopted more rapidly in the Philippines than in any other

country, which may not be surprising given that IRRI is located there and, as a

consequence, IRRI research may be most relevant in the Philippines" (Herdt

and Capule, 1983, 15). Philippine President Ferdinand Marcos enthusiastically

promoted the first wave of HYVs across the country (Chandler Jr, 1992). IRRI

president Raymond Chandler Jr. noted “one of the chief factors in the rapid

spread of the new varieties in the Philippines was" the Marcos regime and its

enthusiasm for the early varieties (Chandler Jr, 1992, 111). While across Asia,

HYVs accounted for merely 30 percent of rice grown in 1970, in the Philippines

the majority of farms had adopted new varieties in the same period (Dalrymple,

1978; Herdt and Capule, 1983).

12While the traditional structural change literature emphasizes total factor productivity, I

follow the convention of agricultural economics in emphasizing crop yield, also known as yield

per hectare, as the measure of productivity.13The tungro virus is the colloquial name for rice tungro bacilliform virus.


STRUCTURAL CHANGE First and foremost, the green revolution was a

massive increase in agricultural productivity growth propelled by the inven-

tion of scalable modern crop breeding techniques (Evenson and Gollin, 2003;

Gollin et al., 2016). Agricultural productivity has long been considered the

key to structural change and modernization (Nurkse, 1953; Rostow, 1959). Ac-

cordingly, the green revolution has often been characterized as an archetypical

shock to productivity with the potential for fueling structural transformation

(Matsuyama, 1992). However, though the Philippines, as the epicenter of the

green revolution, experienced incredible gains to agricultural productivity, it

never experienced industrial development.

Figure 3.2, bottom panel, shows the pattern of structural change from 1970

to 2000. Each colored line corresponds to the share of total employed in the

agriculture, manufacturing, and services sector. 14 The patterns shown in the

bottom panel of 3.2 fit more with the experiences of African economies than the

Philippine’s Asian neighbors (Diao et al., 2017; Gollin, Jedwab, and Vollrath,

2016).

Three patterns of labor reallocation stand out. First, the agricultural labor

share increases temporarily during the initial years of the green revolution, and

then steadily declines after its 1973 peak. Second, the manufacturing share of

employment is, surprisingly, constant from 1970 to 2000. This pattern stands in

sharp contrast with the “hump shape" evolution of manufacturing labor share

seen along the growth path of nearly all OECD countries (Herrendorf et al.,

2014), as well as many Asian neighbors. Third, the share of service-sector

employment increases steadily through time, surpassing the total agricultural

share by the 1990s.15

While classic models of structural change emphasize the role of agricultural

productivity in reallocating economic activity to modern sectors, the story may

be more nuanced. As Mastuyama (1992) famously noted, a (Hicks-neutral) rise

in agricultural productivity in small open economies can increase the share of

agricultural labor employed in agriculture–the opposite pattern as predicted by

classic development theorists, who often used a closed economy environment.

14Aggregate data for Philippine employment shares comes from Timmer et al. (2015).15The patterns of structural change are similar if one visualizes the share of value added output

by sector. Since this paper will explore within country variation in employment share, I explore

aggregate patterns of structural change using employment share outcomes.

3.3. DATA 115

Moreover, many other factors can impact the reallocation of labor as well.

Forces such as factor mobility, input substitutability, and, in particular, the bias

of agricultural technological change all shape predictions of structural change

(Bustos et al., 2016; Foster and Rosenzweig, 2004,0; Marden, 2017).

This study attempts to unpack the aggregate patterns shown in Figure 3.2 us-

ing within-country variation. Utilizing a newly assembled panel of municipality-

level data, I explore how the rollout of green revolution technologies impacted

the modernization of rural Philippine economies. In doing so, I attempt to

understand how the rise in agricultural productivity during the green revolution

translated into different forms of structural change in the short and long run.

3.3 Data

I combine newly digitized data on Philippine agriculture, along with micro-

data sources, to explore the effect of green revolution productivity increases on

short and long-run labor reallocation. My study focuses on Philippine munici-

palities, or towns, for the periods 1960-2000. In doing so, I digitize agricultural

data from the Republic of Philippines’ Census of Agriculture (CAS). I com-

bined this new machine-readable green revolution data with socioeconomic and

structural change outcomes come from the Census of Population and Housing.

I digitize agricultural outcomes for the 1960, 1970, and 1980 volumes of the

The Republic of Philippines’ Census of Agriculture. Importantly, the Philippine

CAS reports statistical aggregates at the municipality level. For the purpose of

this study, I digitize variables related to 1) basic farm characteristics (e.g. total

farms, total farm area, irrigated farm area; farm labor usage); 2) rice production

(e.g. rice area, effective rice area planted, quantity of rice harvested); 3) farm

industrialization (e.g. use of intensive harvesting techniques; numbers of farms

using harvesters, threshers, and tractors); and 4) institutional outcomes (e.g.

number and area farms under share copping tenancy).

While the CAS is a valuable and underutilized quantitative resource, these

agricultural statistics also have limitations. One shortcoming is that some vari-

ables are not available for the entire panel (1960,1970,1980). Moreover, while

many variables are available for the complete panel, some are not reported in

consistent formats. For example, rice yield calculations are limited to the years


1960 and 1970; as rice production was reported in a different format in the

following census years. Other variables, such as share cropping contract types,

are only reported in the 1960 CAS. Finally, some variables must be harmonized

across census years, such as seasonal employment, which was reported in a

more disaggregated formatted in the 1980 census reports.

Structural change outcomes come from the Philippine Census of Population

and Housing public use files (PUFs). The PUF data extracts were provided

to me by the University of Philippine Population Institute for the years 1970

and 1980.16 For 1990, 1995, and 2000, 10 percent population census sample

extracts come from the University of Minnesota Population Center’s Integrated

Public Use Microdata Series (Minnesota Population Center, 2017).

Structural change outcomes for each municipality are calculated from cen-

sus microdata. I follow the structural-change literature, measuring structural

transformation via the share of the workforce employed in agriculture, manu-

facturing, and service sectors. Unfortunately, data limitations preclude using

pre-treatment variables from the 1960 census, as the 1960 population cen-

sus publication does not report industry or sector employment aggregates by

municipality.

I study the empirical effects of structural change over the short and long run,

and thus construct two different panel data sets. First, a sample of all Philippine

municipalities with outcomes for the years 1970 and 1980 (excluding the

Mindanao region) is used. Second, another set of data uses outcomes for the

years 1970, 1980, 1990, 1995, and 2000, but this sample is limited only to

the Central Luzon region — a principal rice-intensive region on the largest

Philippine island of Luzon.

Due administrative aggregation, the two data sets encompass different

sets of municipalities. First, 932 municipalities are followed for the short-run

panel (1960-1980); second, 754 for the long-run panel (1960-2000). Since the

long-run data uses outcomes from IPUMS, and IPUMS aggregates small munic-

ipalities, this panel contains fewer municipalities due to aggregation. Moreover,

differences in the scope of municipalities are the result of reorganization of

township boundaries during and after Marcos’ Martial Law period.

This study excludes much of the Southern Philippines: specifically, the

16Unfortunately, the author was not what share of the total census these extracts represent.

3.4. EMPIRICS 117

island of Mindanao, which was engaged in separatist struggles and decades-

long Islamic insurgency. Since this study focuses on agrarian municipalities

and labor reallocation, I exclude the metropolis of Manila, or more specifically,

the National Capital Region (NCR).

I calculate a measure of HYV and traditional rice suitability for each Philip-

pine municipality. I utilize cell-level (5 arc-minute grid) data on agro-climactic

crop suitability and other geographic variables from the Food and Agricultural

Organization’s Global Agro-Ecological Zones project (FAO-GAEZ). Aggre-

gates from these cell-level raster data are aggregated to the municipality-level

using means. Importantly, this FAO-GAEZ suitability data are calculated using

the average climatic conditions for the 1961 to 1990, a period that encompasses

the green revolution.

Administrative boundaries are not consistent through time. After World War

II, the Philippines experienced substantial re-organization of municipalities,

including administrative splits, merges, and name changes. New provinces

were also created from larger provinces. These trends were especially apparent

during Ferdinand Marcos’ regime (1965−86), which spans the bulk of this

study. Hence, care is taken to harmonize administrative units to time-consistent

definitions. In doing so, I have mapped municipalities and province to their

original 1960 boundaries. Splits and merges were tracked by an exhaustive

search of the Official Gazette of the Republic of the Philippines.

3.4 Empirics

The goal of my empirical analysis is to explore how the green revolution

impacted structural change in the Philippines–with different short-run and long-

run implications. I study how the introduction of IR8 in 1966, and the march of

HYVs thereafter, increased agricultural yields and reallocated rural economic

activity. I ague that the agricultural development causes by the green revolution

initially increased the demand of agricultural labor but eventually displacing

labor to the service sector.

I tell my empirical story in three sections:

1. Adoption–Before tackling structural change, I explore the dissemination


and rapid adoption of green revolution technologies across the Philip-

pines.

2. Short Run–Second, I study how new rice varieties impacted agricultural

productivity, rural mechanization, and structural change in the short run:

1970-1980.

3. Long Run–Last, I turn to the impact of green revolution technologies

growth on structural change over a long-run horizon, 1970-2000. I argue

that the results can be explained by mechanization that resulted from the

short-run increase in labor demand.

3.4.1 Adoption

First, I establish the rapid dissemination of new technologies and their

determinates. If new green revolution varieties were never adopted, there would

be little to say about their impact on structural change. In the Philippines, unlike

other Asian countries, as well as modern African countries, green revolution

cultivars were rapidly disseminated and adopted, following the first "miracle

rice" variety IR8 in 1966 (Herdt and Capule, 1983; Suri, 2011).

WERE MODERN VARIETIES ADOPTED? The first generation of HYVs

were “adopted more rapidly in the Philippines than in any other country" (Herdt

and Capule, 1983, 14). In 1966, immediately after their advent, the first few

tons of IR8 seeds were disseminated across the fertile Nueva Ecija province of

Central Luzon in 1966. By 1970, according the the Philippine Bureau of Plant

Industry, nearly 50 percent of rice area were comprised of modified varieties. A

mere decade later, the share of modified area rose to 75 percent (ibid; p.16).

The widespread introduction of early generation HYVs is seen even 4

years after their introduction into rice ecosystems in statistics from the 1971

agricultural census. Figure 3.3 plots the area (hectares) of HYV planted in (left)

1970 and (right) 1980 against the area of rice planted in 1960, 6 years before

the advent of IR8. Area is normalized using ln(hectares + 1).17 The top of both

17Since the results are substantively same when using the inverse hyperbolic sine function

with an optimal scaling constant, I show the log-normalized (plus one) variables instead. I use

this normalization throughout the study.

3.4. EMPIRICS 119

panels displays the distribution of the farm area planted to HYV rice for each

year.

Figure 3.3 shows a strong relationship between the share of municipality

area planted with traditional rice varieties in 1960 and the adoption of HYVs in

1970 and 1980. Second, the mass of farms not planting or planting a minuscule

area of seeds shifts drastically in 1980. That is, by 1980 there is a large mass of

locales with a significant portion of area planted with new cultivars. The figures

and preceeding estimates undoubtedly confirm the qualitative narrative that the

Philippines was the fastest adopter of new technology.

I quantify the adoption of green revolution varieties across rice farming

municipalities using the following regression specification,

lnArea HYV Plantedit � α+ β lnRice Areai ,1960 + εit (3.1)

I regress a measure of modern variety adoption lnArea HYV Planted (hectares),

observed for years 1970 and 1980, on the total amount of rice area planted

within the same municipality in 1960: lnRice Area. The coefficient of interest,

β, is an estimate of the average area of HYVs planted per area of 1960 rice

crops planted. This simple bivariate relationship is estimated using a pooled set

of municipalities, indexed by i, for two agricultural census years, indexed by t.

Robust standard errors are clustered on the municipality-level.

Table 3.2, reports the estimates of HYV adoption. Column (1) displays

estimates from Equation 3.4.1, showing that for every one hectare of rice

planted in 1960, .65 hectares of modern varieties were planted over the next two

decades. More generally, column (2) shows that for every hectare of irrigated

farmland in 1960, .34 hectares of new varieties were planted in the same period.

Both estimates are highly significant at the 1 percent level.

The relationships conveyed in Figures 3.3 and estimates from Equation

motivate this study’s measure of modern rice variety adoption in the green

revolution era:

lnHYV Share Plantedit � ln

(Area HYV Plantedit

Rice Areai ,1960

)(3.2)

which corresponds to the area of new varieties planted in 1970 or 1980 as a share

of the total rice area in 1960. The 1960 rice area is used in the denominator,

3.4. EMPIRICS 121

since the measures of rice area planted are not available for the 1980 agricultural

census year.18

WHAT DETERMINED HYV ADOPTION? What determined the adoption

of modern varieties after 1966? Figure 3.4 presents the paper’s preferred mea-

sure of adoption in Equation 3.2, lnHYV Share Planted and its relationship to

geographic variables and pre-treatment covariates from 1960 census of popula-

tion and housing and the 1960 census of agriculture. The plots display results

from simple bivariate regressions, where the main green revolution adoption

measure (Equation 3.2) is regressed on one of the three types of pre-treatment

variables.19 For clarity, each of explanatory variables has been normalized to

be between one and zero. Dots corresponds to the point estimate and line seg-

ments, 95 percent confidence intervals. The plotted left-hand coefficients are

for regressions without any controls, and right-hand side regressions control for

province fixed effects.

Figure 3.4 reveals important determinates of technological adoption across

the Philippines. Importantly, even these basic bivariate relationships eluci-

date qualitative and small-N studies by original agronomy scholarship. Of the

geographic variables, the strongest predictors of HYV adoption are average

temperature, slope index, and the FAO-GAEZ measures of high-input rice suit-

ability (specifically, HYV). All enter as strong positive predictors of modern

variety adoption, even with the inclusion of provincial fixed effects (t �5, 5 and

7, respectively).

The strong agroclimatic predictors of adoption in Figure 3.4 make agro-

nomic sense. In the early days of the green revolution, scientists realized the

success new varieties depended critically on specific environmental conditions

(Barker, Herdt, and Rose, 1985). In particular, early IRRI hybrids were sensitive

to humidity, solar radiation patterns, and required relatively high optimal temper-

atures (Datta, 1981; Veraga, Chu, and Romeo M. Visperas, 1970). Early tropical

varieties were easily damaged by cool temperatures (Maruyama, Yatomi, and

Nakamura, 1990).

18The denominator uses only 1960 rice area, as the total rice area for 1980 has yet to be

digitized.19Robust standard errors clustered on the municipality-level.


Figure 3.4: Municipality Characteristics and Adoption of Green Revolution Tech-

nologies, 1970 −1980

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

No Controls Province Fixed EffectsA

g. Census C

ontrolsG

eographic Controls

Pop. Census C

ontrols

−0.5 0.0 0.5 1.0 −0.5 0.0 0.5 1.0

Farm Size (Avg.) Farm Value

Total Farm Area Tractors

Farm Population Rice Area Rice Yield

Irrigated Area

Highway Length Distance To Highway

Altitude (Avg.) Distance To Urban City

Longitude Area

Temperature (St.Dev.) Rain (Avg.)

Altitude (St.Dev.) Distance To Dam

Rain (St.Dev.) Distance To Port

Distance To Manila/Cebu Rain (Coef. Of Var.) Coconut Suitability

River Length Latitude

Sugar Suitability Low Input Wet Rice Suitability (Avg.) Low Input Dry Rice Suitability (Avg.)

Number Of Rivers High Input Hyv Rice Suitability (Avg.)

Slope (Avg.) Temperature (Avg.)

Hhs With Plumbing Access Open

Hhs With No Ed. Hhs With Electricity

Hhs With Artesian Well Access Hhs With Only Primary Ed.

Total Population Total Households Hhs With Hs Ed.

Hhs With College Ed. Hhs With Pump Access

Estimates (St.Err.)

Varia

bles

restrictions ●● ●No Controls Province Fixed Effects

Figure 3.4: Presents the results for simple bivariate relationship between the adoption (share of

HYV rice grown out of total 1960 rice area) and each variable named on the y-axis. Each point

is the coeffient taken from the simple bivariate regressions; lines represent 95 percent confidence

intervals. The left side presents regression estimates without controls. The right side presents

regression estimates controlling for province-level fixed effects. Robust standard errors are used.

Source: (Bureau of the Census and Statistics, 1962; IIASA/FAO, 2012; National Census and

Statistics Office, 1974,8; National Statistics Office, 1963).

3.4. EMPIRICS 123

Most importantly, green revolution cultivars required intensive, consistent

irrigation (Barker et al., 1985; Farmer, 1979; Otsuka, Gascon, and Asano,

1994).20 Thus, the land gradient, as measured by the FAO-GAZE slope in-

dex, captures an important component of gravity fed irrigation systems used

prominently throughout the Philippines. FAO-GAZE slope index enters as an

important predictor of adoption.

The importance of irrigation in early HYV adoption is further indicated by

the strong positive relationship between the number of households with pump

access, a measure provided by the 1960 population census. Farm households

with access to mechanical pumps likely also use mechanical irrigation tech-

nologies. Clearly, the municipality area under irrigation in the 1960 agricultural

census is the predictor of adoption in the set of pre-treatment covariates. With

this in mind, it is also not surprising the there is a strong negative relationship

between the distance to the nearest dam and adoption.

Reassuringly, the FAO-GAEZ measure for input-intensive modern rice

production — an omnibus measure of agro-climactic HYV suitability — is a

strong predicor of adoption. This is an important relationship to confirm, as

similar FAO-GAEZ measures have been directly used by Bustos et al. (2016)

to study the impact of modern GMOs on economic transformation without

actual agricultural data. The significant positive relationship between the high-

input HYV suitability measure validates the use of this measure even for early

GMOs. Meanwhile, the weak relationship between low input wet and dry rice

agriculture and HYV adoption shows that the high-input measure is not merely

proxying for general rice conditions. In fact, it is capturing the realities of green

revolution activity.

HYV ADOPTION AND FARM ECOSYSTEMS FAO-GAEZ suitability data

are ubiquitous in development research. Figure 3.4 revealed that a) the FAO-

GAEZ measures of high-input HYV rice suitability are not vacuous; indeed

these measures are highly predictive of observed HYV adoption, even in the

early years of the green revolution.

Figure 3.5 presents the relationship between HYV adoption and various

20HYVs for so-called dry rice environments would come later and would be less popular in

the Philippines

3.4. EMPIRICS 125

measures FAO-GAEZ agriculture suitability measures. Panels are arranged

according to their proximity to HYV rice ecosystems — most proximate being

intermediate intensity rice production and least proximate being sugar and

coconut crops. Sugar and coconut are also dominant commodities grown across

the Philippine Islands. Each plot represents a binscatter relationship between the

lnHYV Share Planted measure and average municipality-level crop suitability,

controlling for provincial fixed effects. Each plot contains about 70 “binned"

points–each point corresponding to 50 municipality-year observations. The

panels in Figure 3.5 reveal a key pattern: suitability measures more closely

related to HYV suitability have stronger positive relationships between observed

HYV adoption and suitability measures.

Table 3.3 presents the estimated slopes from Figure 3.5. The fit slopes de-

scend in a gradient as the suitability measures get further from HYV suitability.

Column (1) shows the impact of high input HYV rice suitability is the most

substantial, .148 (1 percent significance). Substantively, this shows that a 1

point increase high input rice suitability is related to a 16 percent increase in the

share of modern varieties adopted. The intermediate input rice measure, column

(2), is slightly smaller: β � .123 with 1 percent significance.

Interestingly, suitability measures that, ex-ante, could be related to HYV

adoption are relatively unrelated to observed adoption. For instance, low input

wet rice and low input dry rice suitability are quite uncorrelated with observed

adoption, with point estimates of .031 and .035 (both insignificant), respectively.

The relationship between HYV adoption and municipalities with high suitability

for general cereal production is even weaker: .008 and insignificant. Last,

adoption is also uncorrelated with the suitability with the dominate Philippine

cash crops, sugar and coconut.

3.4.2 Short-Run Transformation

In this section, I explore how the widespread adoption of green revolution

varieties related to, 1) the rise in agricultural productivity; 2) the industrialization

of agriculture; and most importantly, 3) local patterns of structural change in

the first 4 to 14 years following the invention of high-yielding cultivars. I refer

to this period as the short-run.


I study these relationships with the simple specification, following the

framework used by Bustos et al. (2016):

yit � αi +αt + βHYV Share Plantedit + εit (3.3)

with yit being one of the three types of outcomes recorded in municipality ifor time period t. Municipality fixed effect, αi , control for time invariant town-

level characteristics, and αt are time effects controlling for aggregate shocks.

Importantly, HYV Shareit is my measure of the adoption of green revolution

modern seed varieties: the effective share of rice hectares planted as a share of

total 1960 rice area.

The short-run analysis consists of two periods. Combined with two-way

fixed effects, Equation 3.3 is equivalent to the following first difference specifi-

cation:

Δyi � γ+ βΔHYV Share Plantedi +Δεi (3.4)

The first differences specification in Equation 3.4 admits controls otherwise ab-

sorbed by a fixed effects estimator. Thus, I estimate the preferred specification:

Δyi � γ+ βΔHYV Share Plantedi + X′i ,1960θ+Δεi (3.5)

The preferred specification, 3.5, now includes a vector of municipality-specific,

pre-treatment controls: X1960. This set of geographic variables controls for

core geographic and agro-climactic variables, including (log) area irrigated,

longitude, latitude, average rainfall, average temperature, and the average mu-

nicipality slope.

PRODUCTIVITY AND TECHNOLOGICAL CHANGE If green revolutions

ought to impact structural transformation, they must impact agricultural pro-

ductivity. However, the degree to which green revolutions influence structural

changes wrests on the factor bias of the technological change promoted. This

section shows how HYVs impacted yields, but also impacted both labor ab-

sorbing (land intensive farming practices) and labor displacing farm practices

(mechanization) in the short-run.


Figure 3.6 summarizes changes in rice productivity and the industrialization

of agriculture during the green revolution. With pooled data, (left to right) share

of new varieties adopted in plotted against 1) rice yields; 2) adoption of the

number of farms using succession planting (land intensive practices); and 3)

number of tractors for each municipality (mechanization). Yields are shown

for 1970 only, all other plots are for 1970 and 1980. These raw correlations

reveal a clear relationship between the use of new genetically modified crops

and agricultural productivity, land intensive techniques, and industrialization of

agriculture.

The relationship shown in Figure 3.6 are explored more formally by esti-

mating the first difference specification in Equation 3.5. That is, differences

in (again, log+1) yields, succession cropping farms, and tractor outcomes are

regressed on measures of HYV adoption.

Table 3.4 presents estimates of the aforementioned relationships. Columns

(1) and (2) show estimates for rice yield outcomes between 1960 and 1970,

revealing a strong, significant relationship between average farm yields and

share of modern varieties adopted. In the first decade of the green revolution, a

one percent increase in the share of HYVs adopted increased yields by .3 percent

(1 percent level of significance), after controlling for geo-graphic characteristics.

Table 3.4, columns (3) and (4), confirm that the green revolution sparked the

adoption of land-intensive farming practices (for reasons discussed in Section

3.4.1, back-to-back cropping practices were not physically possible before

1966). The preferred estimates in column (4) indicate that a point increase in

the share of HYV rice leads to a 1.7 percent increase in the number of farms

utilizing these techniques.

Nevertheless, Table 3.4, columns (5) and (6), show that the adoption of

new rice varieties were also associated with the adoption of labor-displacing

technologies. Preferred estimates indicate that a 1 percent increase in the share

of HYVs adopted is related to a 1.6 increase in the number of farms using

tractors.

In the disaggregated data it is clear that green revolution is associated with a

significant change in agricultural productivity. This is key–given productivity is

the necessary ingredient in structural change. However, a more nuanced issued

behind structural change is shown by Table 3.4: what type of technological

3.4. EMPIRICS 129

change was produced by the new revolution rice technologies? The adoption

of modern varieties is associated with the evolution of land intensive practices

(hence, labor-biased) and also labor displacing practices. The following section

explores how green revolution technologies affected agricultural employment —

i.e. explores which dimensions of bias dominated in the short-run.

SHORT-RUN STRUCTURAL CHANGE Did green revolution municipalities

experience structural transformation? If so, what form? I explore to the effect of

HYVs on structural change, measured as the reallocation of labor across broad

sectors: agriculture, manufacturing, and services. In the previous section, I

showed the relationship between the spread of modern varieties on productivity

and technological change. I now explore how these changes translated into the

composition of local economic activity in Philippine municipalities.

I take villages to be small open economies, each of which has a represen-

tative farm with a standard CES technology that combines agricultural labor

and land. A Hicks-neutral technical shift generates a reallocation of labor to

the agricultural sector from the manufacturing (and services).21 Similarly, land-

augmenting technical change reallocates individuals toward the agricultural

sector and is thus labor-biased. On the other hand, labor can be reallocated away

from agriculture if technological change is strongly labor-saving.

I estimate the first differences specification from Equation 3.5 to explore the

relationship between the expansion of modern varieties and changes in employ-

ment shares across municipalities for the years 1970 and 1980. Employment

shares are measured as the share of the total employed population working in

each of the three broad sectors.

The first two columns of Table 3.5 show that areas that adopted more green

revolution rice varieties also increased their share of agricultural employment.

A one percent increase in the share of HYV rice area planted is related to a .02

percent increase (5 percent level of significance) in the share of agricultural

employment in a municipality, according to estimates without controls. Adding

geographic controls increase estimates to .03 (1 percent level of significance).

Accordingly, columns (3) and (4) of Table 3.5 show a corresponding decline

21Specifically Q � A[γ(ALLa )

σ−1σ + (1−γ)(ATTa )

σ−1σ

] σσ−1

, where γ ∈ (0,1) and σ the

elasticity of substitution between factors. I assume that land and labor are complements, σ < 1.


in the intensity of manufacturing employment: a 1 percent rise in the share of

HYV rice planted is associated with a .02 decrease in the share of manufacturing

employment. The estimates are similar with or without geographic controls,

however point estimates are more precise with the inclusion of the covariates.

Interestingly, Table 3.5 shows that the green revolution varieties have no

impact on the share of employment in the service sector.

These short-run patterns of structural change are consistent with general

equilibrium insights by Matsuyama (1992) and Foster and Rosenzweig (2004,

2007). In these contexts, especially increasing an exogenous increase in agri-

cultural productivity leads small open economies to specialize in agricultural

production. In Foster and Rosenzweig (2004, 2007), such an increase in agri-

cultural productivity increases demand for agricultural labor, increasing rural

wages and crowding out rural manufacturing. However, in the Philippine short-

run setting, the increase in agriculture labor demand, and thus peasant incomes,

does not seem to increase the demand for local non-tradables (services).

SHORT-RUN PEASANT LABOR DEMAND The structural change results in

the previous section show that, in the short-run, HYV-adopting municipalities

experienced significant growth (declines) in the share of agricultural (man-

ufacturing) employment. Rather than labor-displacing, the green revolution

encouraged the reallocation of labor to agricultural areas. In the following

section I further explore relationship between green revolution and the rise of

agricultural employment.

I explore how the green revolution related to short-run shifts in peasant labor

demand. Importantly, the Philippine Census of Agriculture allows me to explore

hired seasonal peasant labor to explore the sources of labor reallocation between

1970 and 1980. Theoretically, HYVs were considered land-augmenting (labor

biased) technologies in they promoted back-to-back cropping. If green revo-

lution varieties increased labor demand by virtue of promoting land-intensive

activities, we should see shifts in seasonal labor employment.

The biological nature of HYV innovations hypothetically increase labor

demand across multiple seasons. The principal rice planting season in the Philip-

pines is July - September, with harvesting performed for October - December.

Traditional Asian rice varieties required 160 to 170 days to mature, and were

3.4. EMPIRICS 131

photoperiod sensitive–affected by the length of days and nights during matura-

tion. The new GMO varieties, starting with IR8 and IR36, were photoperiod

insensitive, capable of being planted year-round (under proper conditions). Im-

portantly, the lineage of new green revolution varieties cut the growing season

tremendously. IR8 required around 130 days (about 18 percent less time); IR36,

a mere 110 days (30 percent less).

Table 3.6, panel A shows that the changes in modern varieties coincided

with increases in farm wage labor across the four planting seasons recorded.

Notably, panel A shows the lowest increase labor most outside of the peak

rice period (season 3); the largest effects of HYV adoption on paid peasant are

outside of the primary growing seasons. Panel B adds the succession cropping

covariates, along with the HYV adoption measure. Similarly, the change in

succession farming predicts is positively and significantly related to the increase

in labor demand across all seasons, with the weakest effect for the traditional

season.

There is no a clear relationship between short-run mechanization and (re-

duced) labor demand. Table 3.6, panel C shows a negative, albeit insignificant

, relationship between thresher adoption and the demand for peasant labor in

three out of the four seasons. The relationship is positive and insignificant

for season 4. Threshing tasks were quite labor intensive and, unsurprisingly,

the adoption of threshing technologies has a slightly negative effect on labor

demand. In fact, the mechanization of threshing activity heralded local class

tensions in the Philippines.22 In total, though the green revolution varieties

promoted mechanization, in the short-run these forces had not yet reduced labor

demand

More surprising, Table 3.6, panel D shows a positive relationship between

tractorization and labor demand. However, tractors are often adopted primarily

for land preparation prior to rice planting.23 In the Philippines, tractors often

substituted for draft animals, such as caribou (water buffalo). Moreover, tractor

rental or purchase becomes particular profitable with succession cropping. In

other words, in the short-run, tractorization is endogenous to farming practices

22This is unsurprising, since, historically, the introduction of threshing machines have been met

with resistance–in not only the Philippines, but pre-Industrial Europe (Fegan, 1986; Hobsbawm

and Rude, 1969).23As opposed to miscellaneous tasks, such as transport, etc..


correlated with higher labor demand. While tractors may replace some human

land preparation labor, their adoption is correlated with increased demand for

weeding and harvesting labor. The muted effect of tractorization on hired labor

is consistent with early surveys that found little effect on hired labor for land

preparation Barker, Meyers, Crisostomo, and Duff (1972, 129).24

To summarize, the short-run increase in agricultural labor share is clearly

the result of increased peasant labor demand. This growth in demand seems

driven by demand for preparation labor due to the land-augmenting (labor

biased) nature of HYV innovations. Even though mechanization had begun,

the expansion of double cropping meant that the labor absorbing effects of the

green revolution dominated.

3.4.3 Long-Run Change

I now turn to exploring the impact of green revolution varieties on structural

transformation in the long-run.

To explore the impact of green revolution technologies on long(er)-run

structural change, I estimate the differential evolution of municipalities adopt-

ing more versus less HYVs. To do so, I regress a structural change outcome,

y, for municipality, i, for the years, t, on a measure of green revolution expan-

sion between the years 1966 and 1980. Specifically, I estimate the following

specification,

yit � αi +αt + βt (HYV Share Planted 1966-1980i) +θtXi + εit (3.6)

where HYV Share Planted 1966-1980 is the measure of green revolution ex-

pansion for municipality i. Specification 3.6 also includes time effects, αt , and

municipality fixed effects, αi . The coefficient of interest, βt , is estimated for

every time period, and can be interpreted as the average difference between high

adopting versus low-adopting municipalities, relative to a baseline (omitted)

year, 1970. In other words, the vector of βs capture the average differences for

24One survey of tractor buyers in the Laguna province in the 1960s stated bluntly, "Ninety of

the 150 respondents indicated that they were concerned about the problems of keeping a water

buffalo" Barker et al. (1972, 122).

3.4. EMPIRICS 133

each municipality, relative to 1970 levels, at each point of time: 1980, 1990,

1995, and 2000, giving a sense of the dynamic evolution of green revolution

and non-green revolution municipalities. As in previous specifications, robust

standard errors are clustered at the municipality level.

LONG-RUN STRUCTURAL CHANGE Figure 3.7 shows the patterns of labor

reallocation by plotting estimates from specification (3.6). Solid lines represent

the estimated changes in the labor shares of agriculture (red), manufacturing

(green), and services (blue) between the baseline year, 1970, and 1980, 1990,

1995, 2000, respectively. Table 3.7 shows the point estimates corresponding

Figure 3.7, with and without geographic controls.

The estimated changes in Figure 3.7 correspond to the aggregate patterns of

structural change. First and foremost, the differential dynamic impacts of green

revolution varieties on the agricultural employment share in Figure 3.7 matches

the aggregate pattern shown in Figure 3.2. The long-run estimates capture the

first initial increase in agricultural employment between the 1970 - 1980 period.

This is the same increase shown by the short-run estimates.

Structural change estimates over longer time horizon reveal that the initial

pull of agricultural employment gives way to a significant decline in relative

agriculture employment in high-adopting municipalities. By 2000, the green

revolution has a highly significant and negative impact on agriculture employ-

ment share (relative to 1970). Hence, while the initial labor biased technical

change absorbs agricultural labor, in the long-run other forces reallocate labor

out of agriculture–most likely factors such as capital deepening in agriculture.

Second, in line with theoretical models of structural change in open economies,

it seems that agricultural activity crowds out manufacturing activity in the short

run (as discussed in 3.4.2). However, the negative relationship between green

revolution adoption and manufacturing diminishes through time and is essen-

tially zero by 1990 and 2000.

Third, in the long run, the extent of technological adoption has a large

significant impact on employment in service industry. Importantly, while the

short-run estimates showed no effect of HYV adoption on service employment

(1980), the impact is strong and positive for 1990, 1995, and 2000. It seems

plausible that if agricultural labor was displaced in these later periods, they


Figure 3.7: The Effect of HYV Adoption on Sectoral Labor Reallocation, Relative

to 1970

●

●

●

●

●

●

●●

●●

●●

●

●

●

Agricultural Labor Share

Manufacturing Labor Share

Service Labor Share

1970 1980 1990 2000

−0.05

0.00

0.05

−0.05

0.00

0.05

−0.05

0.00

0.05

Year

Inte

ract

ion

Coe

ffici

ent

Outcome Variable

●●

●

●

Agricultural Labor ShareManufacturing Labor ShareService Labor Share

Figure 3.7: Coefficients, βt , from the regression Equation 3.6, yit � αi +αt+

βt (HYV Share Planted 1966-1980i ) +εit . The plotted coefficients represent the effect of high

versus low HYV adopting municipalities (for 1966-1980), on the growth in sectoral employment

shares for 1980, 1990, 1995, and 2000, relative to 1970 levels. Plots are presented for

agricultural, manufacturing, and service labor share. The Standard errors are clustered at the

municipality level.

3.4. EMPIRICS 135

found their way into low-end service occupations, in a pattern similar to those

explored by Autor and Dorn (2013) for the growth of low-end service employ-

ment in the US. However, the movement of agriculture to services could also

result from increased agricultural wages and the growth in local demand for

non-tradables–an issue I address in the next subsection.

The long-run structural change regressions show that green revolution

municipalities first, experienced an expansion in the share of agricultural em-

ployment, and next, experienced a strong, consistent decline in labor demand.

Thus, while green revolutions are theorized as land-augmenting (labor-biased)

structural change, in the long-run different dynamics are at play.

LABOR SHEDDING AND MECHANIZATION IN THE LONG RUN The

results in Figure 3.7 show that the initial increases in labor demand in the green

revolution were transient. These results were not limited to the Philippines.

Around the world, the green revolution introduced an anomaly to economists:

while labor demand ought to have increased, around the world real wages

were at often stagnant or declining in the periods following the introduction of

new high yielding varieties (Bardhan, 1970; Binswanger, 1986; Coxhead and

Jayasuriya, 1986; Gupta and Shangari, 1979; Lal, 1976). This section attempts

to explain these patterns.

Instead of a land-augmenting, labor-biased technological shock that in-

creased labor demand, other forces were also at work. While in the short-run,

the green revolution increased the demand for agrarian labor, landlords and

farmers often eventually responded to higher relative wages–as well as de-

clining thresher and tractor prices worldwide–with mechanization. For India,

Binswanger (1986) observes a series of patterns that seems to fit the Philippine

context:

“The green revolution led to sharply increased demand for labor,

which caused a big rise in real wages around 1968 ... This in

turn led to increased seasonal and permanent migration, primarily

from Eastern India. But it also led to the use of more tractors

and threshers by Punjab farmers. The combined effect of these

developments was a decline in real wages" (Binswanger, 1986,

33).


In other words, observers of the green revolution in India and the Philippines

alike observed distinct long-run and short-run effects of the green revolution.

The initial increase in labor demand and the decline in agricultural labor is seen

in the long-run structural change regressions in Figure 3.7.

This qualitative story is reflected in aggregate wage data for the Philippines.

The top panel of Figure 3.8 plots changes in the mean real wage of rice workers

for 1975-2000; the grey points represent average annual wages for each Philip-

pine region, and the black plots represent the mean (and 95 percent confidence

intervals) across regions.25 Though the series starts in 1975, a noticeable peak

is observed around 1977, whereby wages start to decline. Real rice farm wages

recover to their mid-1970s level only in the early 1990s.

The bottom panel (B) of Figure 3.8 show the aggregate adoption of tractors

from 1961-2000. Both labor displacing technologies are rapidly adopted from

the 1960s onward, in particular tractors in the mid-1960s and, most prominently,

in the early 1980s. The march of labor displacing technologies are consistent

with the eventual decline of rice farm wages as labor is eventually displaced

from agriculture, as seen in the decline of long-run agricultural employment.

Moreover, the eventual rise in real rice worker wages are also consistent with

the mechanization of rice farming: the wages of substitutable labor (such as har-

vesting and planting activity) fall, as labor is displaced; wages of the remaining

skilled labor increases due to their complementarity with farm capital.

In other words, the simple aggregate patterns in Figure 3.8, are consistent

with the paradoxes of the green revolution and both short-run and long-run

patterns of structural change. To summarize: an increase in real wages reflects

the early increased demand for peasant labor. The eventual fall and recovery of

real wages is consistent with the mechanization of agriculture and the decline

of rural agricultural employment. The rise of rural service employment may

well be the result of displaced, or shed, agricultural labor.

25Unfortunately, pre-1975 data was unavailable from the Bureau of Agriculture series and

have yet to be digitized.


3.5 Conclusion

I study how the green revolution technologies impacted structural change in

their country of origin: the Philippines. Since the birth of the sub-field, develop-

ment economists have argued that rising agricultural productivity was the engine

behind structural transformation: the reallocation of economic activity from

the agricultural sectors to modern manufacturing and service sectors (Johnston

and Mellor, 1961; Nurkse, 1953; Rostow, 1960). However, even after the rapid,

momentous rollout of early green revolution technologies across the islands,

modernization did not follow. In sharp contrast to its Asian contemporaries,

the share of manufacturing labor remained constant, and the agricultural sector

remained the dominant source of employment through the 1980s.

Using newly digitized data on the green revolution, I show that growth in

agricultural productivity produced structural change–but in ways not anticipated

by planners and theoretical models. With a newly constructed panel of Philip-

pine municipalities, I trace how the expansion of new high yielding varieties,

known as HYVs or modern varieties, increased agricultural productivity and

reallocated economic activity across sectors–my measures of structural change.

I focus specifically on how the share of employment in agriculture, manufactur-

ing, and services changed over the next four decades, immediately following

the arrivals of HYVs in 1966.

I show that green revolutions technological shocks had quite different effects

on short-run and long-run structural change, producing unexpected effects on

peasant agricultural labor. I first confirm that after 1966, unlike many Asian

(and current African economies) HYVs were widely adopted across Philippine

townships and were subsequently related to a rapid increase in agricultural

productivity. I then show that in the short run, 1970-1980, the green revolution

translated into labor-absorbing technological change: reallocating labor into

HYV-intensive rice economies. These results are consistent with the increase in

aggregate agricultural employment the decade after the introduction of modern

rice varieties. However, in the long run, 1980-2000, I show how this pattern

is reversed; the green revolution translated into labor-displacing technological

change. In particular, agricultural wage labor was dislocated from agriculture

and pushed into low-skilled service employment. I argue that rising wages

3.5. CONCLUSION 139

and declining prices of capital prompted rice farms to mechanized, and thus

promoted the long-run decline in agricultural employment.

140 REFERENCES

References

Allen, R. C. (2000). Economic structure and agricultural productivity in Europe,

1300-1800. European Review of Economic History 4(1), 1–25.

Atkinson, L. J. and D. E. Kunkel (1976). High-Yielding Varieties of Rice in the

Philippines, Progress of the Seed-Fertilizer Revolution.

Autor, D. H. and D. Dorn (2013). The Growth of Low-Skill Service Jobs and the

Polarization of the US Labor Market. American Economic Review 103(5),

1553–1597.

Bardhan, P. (1970). ’Green Revolution’ and Agricultural Labourers. Economic

and Political Weekly 5(29/31), 1239–1246.




Barker, R., W. H. Meyers, C. M. Crisostomo, and B. Duff (1972). Employ-

ment and Technological Change in Philippine Agriculture. International

Labour Review 106, 111–140.

Binswanger, H. (1986). Agricultural Mechanization: A Comparative Historical

Perspective. The World Bank Research Observer 1(1), 27–56.

Boyce, J. K. (1993). The Philippines: The Political Economy of Growth and

Impoverishment in the Marcos Era. Honolulu, Hawaii: University of

Hawaii Press.

Bureau of Agriculture Statistics (2012). Agricultural Wage Rates of Farm

Workers by Type of Wages, Type of Farm Workers, Geolocation, Sex and

Year.

Bureau of the Census and Statistics (1962). Census of the Philippines : 1960

Housing and Population: Report by Province. Manila: Bureau of the

Census and Statistics.

Bustos, P., B. Caprettini, and J. Ponticelli (2016). Agricultural Productivity and

Structural Transformation: Evidence from Brazil. American Economic

Review 106(6), 1320–1365.

Chandler Jr, R. (1992). An Adventure in Applied Science: A History of the In-

ternational Rice Research Institute. Manila: International Rice Research

Institute.

REFERENCES 141

Clark, G. (1987). Productivity Growth without Technical Change in European

Agriculture before 1850. The Journal of Economic History 47(2), 419–

432.

Clark, G. (1991). Yields Per Acre in English Agriculture, 1250-1860: Evidence

from Labour Inputs. The Economic History Review 44(3), 445–460.

Cleaver, H. M. (1972). The Contradictions of the Green Revolution. The

American Economic Review 62(1/2), 177–186.

Coxhead, I. and S. Jayasuriya (1986). Labor-Shedding with Falling Real Wages

in Philippine Agriculture. Asian Survey 26(10), 1056–1066.





Dalrymple, D. G. (1978). Development and Spread of High-Yielding Varieties

of Wheat and Rice in the Less Developed Nations (6 ed.). Washington,

D.C.: U.S. Department of Agriculture.

Datta, S. K. D. (1981). Principals and Practices of Rice Production. New York:

John Wiley & Sons, Ltd.

Diao, X., M. McMillan, and D. Rodrik (2017). The Recent Growth Boom in

Development Economics: A Structural Change Perspective.

Echevarria, C. (1997). Changes in Sectoral Composition Associated with

Economic Growth. International Economic Review 38(2), 431–452.

Evenson, R. E. and D. Gollin (2003, may). Assessing the Impact of the Green

Revolution, 1960 to 2000. Science 300(5620), 758 LP – 762.

Farmer, B. H. (1979, jul). The ‚Äògreen revolution’ in South Asian ricefields:

Environment and production. The Journal of Development Studies 15(4),

304–319.

Fegan, B. (1986). Tenants’ Non-Violence Resistence to Landowners Clains

in a Central Luzon Village. In J. C. Scott and B. J. T. Kerkvliet (Eds.),

Everyday Forms of Peasant Resistance in South-east Asia (2006 ed.).,

Chapter 5, pp. 87–106. New York, New York: Routledge and Taylor &

Francis Group.

Foster, A. and M. Rosenzweig (2004). Agricultural Productivity Growth, Rural

Economic Diversity, and Economic Reforms: India, 1970–2000.

142 REFERENCES

Economic Development and Cultural Change 52(3), 509–542.

Foster, A. D. and M. R. Rosenzweig (2007). Economic Development and the

Decline of Agricultural Employment. In T. P. S. Economics and J. A.

S. B. T (Eds.), Handbook of Development, Volume Volume 4, Chapter 47,

pp. 3051–3083. Elsevier.

Gilboy, E. W. (1932). "Demand as a Factor in the Industrial Revolution". In A. H.

Cole (Ed.), Facts and Factors in Economic History Articles by former

Students of Edwin Francis Gay, pp. 620–639. Cambridge, Massachusetts.

Gollin, D., C. W. Hansen, and A. M. Wingender (2016). Two Blades of Grass:

The Impact of the Green Revolution.

Gollin, D., R. Jedwab, and D. Vollrath (2016). Urbanization with and without

industrialization. Journal of Economic Growth 21(1), 35–70.

Gollin, D., S. Parente, and R. Rogerson (2002). The role of agriculture in

development. American Economic Review.

Gupta, D. P. and K. K. Shangari (1979). Agricultural Development in Punjab,

1952-53 to 1976-77.

Herdt, R. W. and C. C. Capule (1983). Adoption, Spread, and Production Impact

of Modern Rice Varieties in Asia. Los Banos, Philippines: International


Herrendorf, B., R. Rogerson, and Á. Valentinyi (2014). Growth and Structural

Transformation, Volume 2. Elsevier B.V.

Hobsbawm, E. and G. Rude (1969). Captain Swing. London, England:

Lawrence and Wishart.

IIASA/FAO (2012). Global Agro‚Äêecological Zones (GAEZ v3.0).

Institute for International Rice Research (2017). World Rice Statistics, Version

3.0.

Ishikawa, S. (1970). Agricultural Development Strategies in Asia: Case Studies

of the Philippines and Thailand. Manila: Asian Development Bank.

Johnston, B. F. and J. W. Mellor (1961). The Role of Agriculture in Economic

Development. The American Economic Review 51(4), 566–593.

Kikuchi, M. and Y. Hayami (1983). New Rice Technology, Intrarural Migra-

tion, and Institutional Innovation in the Philippines. Population and

Development Review 9(2), 247–257.

Ladejinsky, W. (1970, jul). Ironies of India’s Green Revolution. Foreign

REFERENCES 143

Affairs 48(4), 758–768.

Lal, D. (1976). Agricultural Growth, Real Wages, and the Rural Poor in India.

Economic and Political Weekly 11(26), A47–A61.

Marden, S. (2017). The agricultural roots of industrial development:‚Äòforward

linkages’ in reform era China.

Maruyama, S., M. Yatomi, and Y. Nakamura (1990). Response of Rice Leaves

to Low Temperature I. Changes in Basic Biochemical Parameters. Plant

Cell Physiology 31(3), 303–309.

Matsuyama, K. (1992, dec). Agricultural productivity, comparative advantage,

and economic growth. Journal of Economic Theory 58(2), 317–334.

Minnesota Population Center (2017). Integrated Public Use Microdata Series,

International: Version 6.5.

Mokyr, J. (1977). Demand vs. Supply in the Industrial Revolution. The Journal

of Economic History 37(4), 981–1008.

National Census and Statistics Office (1974). Census of Agriculture, 1971.

Manila: National Census and Statistics Office.

National Census and Statistics Office (1985). Census of Agriculture, 1980.

National Statistics Office (1963). Census of Philippines, 1960: Agriculture.

Manila: National Statistics Office.



Otsuka, K., F. Gascon, and S. Asano (1994). ‘Second-generation’ MVs and the

evolution of the Green Revolution: the case of Central Luzon, 1966-1990.

Agricultural Economics 10, 283–295.

Parayil, G. (2003). Mapping technological trajectories of the Green Revolution

and the Gene Revolution from modernization to globalization. Research

Policy 32(6), 971–990.

Peng, S., K. G. Cassman, S. S. Virmani, J. Sheehy, and G. S. Khush (1999).

Yield Potential Trends of Tropical Rice since the Release of IR8 and the

Challenge of Increasing Rice Yield Potential. 39, 1552–1559.

Rostow, W. W. (1959). The Stages of Economic Growth. The Economic History

Review 12(1), 1–16.



144 REFERENCES

Sanyal, B. (1983). How Revolutionary Was India’s Green Revolution? Com-

parative Studies of South Asia, Africa and the Middle East III(2), 31–44.

Scott, J. C. (1986). Everyday Forms of Peasant Resistance. The Journal of

Peasant Studies 13(2), 5–35.

Smith, E. (2009, dec). Imaginaries of Development: The Rockefeller Foundation

and Rice Research. Science as Culture 18(4), 461–482.

Suri, T. (2011, jan). Selection and Comparative Advantage in Technology

Adoption. Econometrica 79(1), 159–209.

Timmer, M. P., G. J. de Vries, and K. de Vries (2015). Patterns of structural

Change in Developing Countries. In J. Weiss and M. Tribe (Eds.), Rout-

ledge Handbook of Industry and Development, Chapter 4, pp. 65–83.

New York: Routledge.

Unnevehr, L. J. (1986). Changing Comparative Advantage in Philippine Rice

Production: 1966 to 1982. Food Research Institute Studies 20(1), 43–70.

Unnevehr, L. J. and A. M. Balisacan (1983). Changing Comparative Advantage

in Philippine Rice Production.

Veraga, B. S., Chu, and Romeo M. Visperas (1970). Effect of temperature

on the anthesis of IR8. International Rice Community Newsletter 19(3),

11–17.

White, L. T. (2009). Political Booms: Local Money and Power in Taiwan, East

China, Thailand, and the Philippines. Series on Contemporary China –


REFERENCES 145

Table 3.1: Description Statistics

TOTAL SAMPLE RICE ECONOMIES NON-RICE ECONOMIES

Mean St.Dev. N Mean St.Dev. N Mean St.Dev. N

1960 Agriculture CensusAverage Farm Size 3.29 3.05 2839 3.64 2.15 706 3.17 3.29 2133

Farm Land Value (Nominal Pesos) 1604241.04 1804768.31 2839 2353087.74 2154290.01 706 1356380.86 1598116.27 2133

Irrigated Area 521.89 973.14 2839 1260.55 1602.32 706 277.40 413.69 2133

Farm Area 4712.52 4516.76 2839 8351.59 5246.62 706 3508.02 3495.36 2133

Farm Population 8889.94 6102.58 2839 14066.30 6223.89 706 7176.62 4995.47 2133

Farms Under Full Ownership 596.40 528.88 2839 768.62 695.01 706 539.40 446.70 2133

Farms Under Full Ownership (Area) 2073.55 2753.76 2839 3210.39 3712.35 706 1697.27 2228.78 2133

Farms Under Tenancy 666.06 714.00 2839 1242.53 685.16 706 475.26 613.93 2133

Farm Under Tenancy (Area) 1413.42 1648.68 2839 3091.69 2061.44 706 857.93 985.71 2133

Farms With Tractors 5.93 17.34 2839 12.35 23.89 706 3.81 13.90 2133

Rice Area 1995.65 2123.69 2839 4900.03 2262.37 706 1034.33 770.13 2133

Rice Yield 25.89 10.41 2839 29.18 8.95 706 24.80 10.63 2133

1960 Population CensusHouseholds With Electricity 343.45 1114.01 2177 451.40 1115.18 568 305.35 1111.44 1609

Households Using Open Wells 1196.75 1313.27 2177 1618.14 1614.56 568 1047.99 1153.18 1609

Households Using Artesian Wells 230.53 592.39 2177 216.36 375.23 568 235.54 652.03 1609

Households Using Water Pumps 936.77 1465.54 2177 1878.08 2037.20 568 604.47 1009.75 1609

Households Using Faucets (Plumbing) 514.68 1190.70 2177 534.91 1443.54 568 507.54 1087.94 1609

Households With No Education 8511.72 7465.09 2177 11753.53 6751.18 568 7367.31 7370.09 1609

Households With Primary Education 7260.05 5977.16 2177 10542.96 6079.67 568 6101.14 5491.99 1609

Households With Secondary Education 3118.85 3499.52 2177 4744.26 3935.87 568 2545.05 3138.08 1609

Households With College Education 993.05 2214.25 2177 1474.51 2203.89 568 823.09 2193.47 1609

Total Households 3469.25 2956.22 2177 4878.47 2924.27 568 2971.78 2803.89 1609

Total Population 19886.34 18090.71 2177 28516.10 18032.46 568 16839.92 17106.33 1609

Geographic VariablesArea (Sq.M.) 1.62e+08 2.07e+08 2827 2.23e+08 2.55e+08 706 1.42e+08 1.84e+08 2121

Altitude (Mean, M.) 206.72 265.54 2827 134.22 145.13 706 230.86 290.96 2121

Altitude (St.Dev., M.) 128.69 114.42 2827 102.44 108.30 706 137.43 115.08 2121

Distance to Dam (Deg.Dist.) 0.79 0.31 2827 0.71 0.34 706 0.81 0.29 2121

Distance to Highway (Deg.Dist.) 0.13 0.38 2827 0.10 0.41 706 0.14 0.37 2121

Distance to Manila or Cebu (Deg.Dist.) 1.58 0.95 2827 1.55 0.93 706 1.59 0.95 2121

Distance to City (Deg.Dist.) 0.19 0.23 2827 0.15 0.19 706 0.21 0.24 2121

Distance to Port (Deg.Dist.) 0.53 0.53 2827 0.49 0.45 706 0.54 0.56 2121

Highway Length (Deg.Dist.) 0.64 0.54 2827 0.74 0.45 706 0.60 0.56 2121

Latitude 13.56 2.68 2827 13.97 2.36 706 13.43 2.77 2121

Longitude 122.28 1.63 2827 121.94 1.33 706 122.39 1.71 2121

Slope (Mean, Slope Index) 6638.76 2312.86 2827 7758.73 1951.09 706 6265.96 2303.93 2121

Rain (Mean) 2439.33 517.92 2827 2333.29 434.09 706 2474.63 538.44 2121

Rain (St.Dev.) 108.82 104.86 2827 96.93 93.61 706 112.77 108.08 2121

Rain (Coeff. Var.) 62.14 25.60 2827 65.02 20.54 706 61.18 27.02 2121

Rivers, Total Length (Deg.Dist.) 0.20 0.27 2827 0.29 0.34 706 0.16 0.23 2121

Rivers, Numbers 3.37 3.79 2827 4.86 5.05 706 2.88 3.11 2121

Temperature (Mean) 26.20 1.46 2827 26.55 0.79 706 26.08 1.61 2121

Temperature (St.Dev.) 6.74 6.06 2827 5.31 5.58 706 7.21 6.14 2121

146 REFERENCES

Table 3.2: Adoption of Early Green Revolution Technologies, 1970 −1980

Dependent Variable Effective Area HYV Planted

(1) (2)

Ln Rice Farm Area, 1960 0.6468∗∗∗(0.0276)

Ln Irrigated Farm Area, 1960 0.3449∗∗∗(0.0177)

r2 0.105 0.071

N 2839 2839

Note: Table 3.2 displays estimates from a regression on pooled, cross-sectional data: regressing

the (ln) total area of HYV area planted, 1970-1980 on two measures of 1960 rice output. Column

(1) shows regressions on (ln) total area of rice harvested (1960). Column (2) shows regressions

on (ln) total irrigated rice area. Robust standard errors are clustered on the municipality-level.

Asterisks indicate statistical significance at the 1% ∗∗∗, 5% ∗∗, and 10% ∗ levels.

REFERENCES 147

Table 3.3: Adoption of HYVs and FAO-GAEZ Suitability Measures

Dependent Variable Ln HYV Share Adopted

(1) (2) (3) (4) (5) (6) (7)

High HYV Suitability 0.1475∗∗∗(0.0189)

Inter HYV Suitability 0.1235∗∗∗(0.0206)

Low Wet Rice Suitability 0.0317

(0.0232)

Low Dry Rice Suitability 0.0351

(0.0264)

General Cereal Suitability 0.0080

(0.0257)

Sugar Suitability 0.0093

(0.0280)

Coconut Suitability -0.0083

(0.0247)

r2 0.205 0.211 0.188 0.188 0.198 0.198 0.198

N 2817 2289 2817 2817 2289 2289 2289

Note: Table 3.3 reports results from regressions using pooled cross-sectional data, regressing the

(ln) total area of HYV area planted, 1970-1980, on FAO-GAEZ normalized suitability measures.

Each column reports estimates of the bivariate relationships between HYV adoption and (1)

HYV suitability; (2) intermediate input HYV rice suitability; (3) low input wet rice suitability;

(4) low dry rice suitability; (5) general cereal crop suitability; (6) sugar suitability; and (7)

coconut suitability. Robust standard errors are clustered on the municipality-level. Asterisks

indicate statistical significance at the 1% ∗∗∗, 5% ∗∗, and 10% ∗ levels.

148 REFERENCES

Tabl

e3.

4:P

rod

uct

ivit

y,In

ten

sifi

cati

on

,M

ech

aniz

atio

n,

and

HY

Vs,

19

70−1

98

0

Dep

enden

tV

aria

ble

Δln

Yie

lds

Δln

Succ

essi

on

Cro

ppin

gΔ

lnT

ract

ors

(1)

(2)

(3)

(4)

(5)

(6)

ΔL

nH

YV

Adopti

on

0.4

191∗∗∗

0.2

848∗∗∗

1.0

208∗∗∗

1.6

561∗∗∗

2.3

206∗∗∗

1.5

764∗∗∗

(0.0

554)

(0.0

634)

(0.2

606)

(0.2

928)

(0.1

335)

(0.1

743)

Contr

ols

No

Yes

No

Yes

No

Yes

r20.0

43

0.0

67

0.0

18

0.2

25

0.1

05

0.1

41

N921

917

932

928

1874

1866

No

te:

Tab

le3

.4re

po

rts

esti

mat

esu

sin

gS

pec

ifica

tio

n(5

).C

olu

mn

s(1

)-(2

)re

po

rtth

ees

tim

ated

rela

tio

nsh

ipb

etw

een

chan

ge

in(l

n)

rice

yie

lds

bet

wee

n1

96

0

and

19

70

and

the

(ln

)H

YV

ado

pti

on

du

rin

gth

esa

me

per

iod

.C

olu

mn

s(3

)-(4

).

Rep

ort

chan

ges

in(l

n)

nu

mb

ero

ffa

rms

eng

agin

gin

succ

essi

on

cro

pp

ing

fro

m

1960-1

970

and

1970-1

980

and

(ln)

HY

Vad

opti

on

over

the

sam

eper

iods.

Colu

mn

(5)-

(6)

report

sa

sim

ilar

rela

tionsh

ip,

but

for

chan

ge

(ln)

num

ber

of

trac

tors

.

Co

lum

ns

(1).

Ro

bu

stst

and

ard

erro

rsar

ecl

ust

ered

on

the

mu

nic

ipal

ity

-lev

el.

Co

lum

ns

(1),

(3),

(5)

rep

ort

esti

mat

esw

ith

ou

tco

ntr

ols

,w

hil

eco

lum

ns

(2),

(4),

and

(6)

incl

ude

contr

ols

.C

ontr

ols

incl

ude

(log)

area

irri

gat

ed,

longit

ude,

lati

tude,

aver

age

rain

fall

,av

erag

ete

mper

ature

,an

dth

eav

erag

em

unic

ipal

ity

slo

pe.

Ro

bu

stst

and

ard

erro

rsar

ecl

ust

ered

on

the

mu

nic

ipal

ity

-lev

el.

Ast

eris

ks

ind

icat

est

atis

tica

lsi

gn

ifica

nce

atth

e1

%∗∗∗

,5

%∗∗ ,

and

10

%∗ l

evel

s.

REFERENCES 149

Tabl

e3.

5:S

ho

rtR

un

Em

plo

ym

ent

Sh

ares

and

Gre

enR

evo

luti

on

Tec

hn

olo

gie

s

Dep

enden

tV

aria

ble

ΔE

mplo

ym

ent

Shar

e

Agri

cult

ure

Man

ufa

cturi

ng

Ser

vic

es

(1)

(2)

(3)

(4)

(5)

(6)

ΔL

nH

YV

Adopti

on

0.0

204∗∗

0.0

269∗∗∗

-0.0

238∗∗∗

-0.0

219∗∗

-0.0

065

-0.0

038

(0.0

082)

(0.0

097)

(0.0

070)

(0.0

087)

(0.0

093)

(0.0

114)

Gra

phic

Contr

ols

No

Yes

No

Yes

No

Yes

r20.0

06

0.0

33

0.0

10

0.0

43

0.0

01

0.0

12

N932

928

932

928

932

928

No

te:

Tab

le3

.5re

po

rts

esti

mat

esu

sin

gS

pec

ifica

tio

n(5

),w

her

est

ruct

ura

lch

ang

eo

utc

om

esar

ere

gre

ssed

on

HY

Vad

op

tio

nm

easu

res.

Sp

ecifi

call

y,co

lum

ns

(1)-

(2)

reg

ress

edch

ang

esin

(ln

)sh

are

of

agri

cult

ura

lem

plo

ym

ent

on

chan

ges

in(l

n)

shar

eof

HY

Vad

opti

on;

colu

mns

(3)-

(4)

report

sim

ilar

spec

ifica

tions

for

(ln)

shar

eof

man

ufa

cturi

ng

labor;

and

(5)-

(6),

for

shar

eof

serv

ice

labor,

Colu

mns

(1),

(3),

(5)

report

esti

mat

esw

ithout

contr

ols

,w

hil

eco

lum

ns

(2),

(4),

and

(6)

incl

ud

eco

ntr

ols

.C

on

tro

lsin

clu

de

(lo

g)

area

irri

gat

ed,

lon

git

ud

e,la

titu

de,

aver

age

rain

fall

,av

erag

ete

mp

erat

ure

,an

dth

eav

erag

em

un

icip

alit

ysl

op

e.R

obu

st

stan

dar

der

rors

are

clu

ster

edo

nth

em

un

icip

alit

y-l

evel

.A

ster

isk

sin

dic

ate

stat

isti

cal

sig

nifi

can

ceat

the

1%∗∗∗

,5

%∗∗ ,

and

10

%∗ l

evel

s.

150 REFERENCES

Table 3.6: Seasonal Peasant Employment and Green Revolution Technologies

Dependent Variable: Δ ln Peasant Labor Hired in

Season 1 Season 2 Season 3 Season 4

Panel A: Effect of HYVs Only (1) (2) (3) (4)

Δ Ln HYV Adoption 1.2250∗∗∗ 0.8302∗∗∗ 0.5172∗∗∗ 0.7760∗∗∗(0.2126) (0.2478) (0.1838) (0.2034)

Panel B: Effect of Succession Cropping & HYVs (1) (2) (3) (4)

Δ Ln Succession Farms 0.0498∗∗ 0.0886∗∗∗ 0.0421∗ 0.0529∗∗(0.0238) (0.0264) (0.0235) (0.0269)

Δ Ln HYV Adoption 1.1426∗∗∗ 0.6834∗∗∗ 0.4475∗∗ 0.6885∗∗∗(0.2188) (0.2525) (0.1861) (0.2070)

Panel C: Effect of Threshers & HYVs (1) (2) (3) (4)

Δ Ln Threshers -0.0252 -0.0241 -0.0316 0.0333

(0.0196) (0.0230) (0.0211) (0.0218)

Δ Ln HYV Adoption 1.3372∗∗∗ 0.9373∗∗∗ 0.6581∗∗∗ 0.6280∗∗∗(0.2343) (0.2756) (0.2132) (0.2237)

Panel D: Effect of Tractors & HYVs (1) (2) (3) (4)

Δ Ln Tractors 0.0855∗∗∗ 0.0780∗∗∗ 0.0810∗∗∗ 0.0837∗∗∗(0.0248) (0.0276) (0.0280) (0.0305)

Δ Ln HYV Adoption 1.1030∗∗∗ 0.7189∗∗∗ 0.4017∗∗ 0.6566∗∗∗(0.2106) (0.2500) (0.1870) (0.2030)

Controls Yes Yes Yes Yes

N 928 928 928 928

Notes: Table 3.6 reports regressions for the change in Ln Peasant Labor demand across different growing seasons on changes in farming

practices. Columns (1) shows estimates of peasant employment for the January to March season; column (2), April - June; column (3), peak

season employment for the July - September season; and column (4) is for October - December. Panel A reports regressions of changes in

seasonal employment on the (ln) HYV adoption measure. Panel B reports the same specification as A, but including a measure for the change

in (ln) number of farms practicing succession cropping. Panel C similarly includes changes in the (ln) number of municipality threshers, and

Panel D includes the (ln) number of tractors. Controls include (log) area irrigated, longitude, latitude, average rainfall, average temperature,

and the average municipality slope. Robust standard errors are clustered on the municipality-level. Asterisks indicate statistical significance at

the 1% ∗∗∗ , 5% ∗∗ , and 10% ∗ levels.

REFERENCES 151

Table 3.7: Long Run Employment Shares and Green Revolution Technologies

Dependent Variable: Log Share of Agriculture Employment Manufacturing Employment Service Employment

(1) (2) (3) (4) (5) (6)

ln HYV Adoption (66-80) × 1980 0.0178 0.0266 -0.0333∗∗∗ -0.0355∗∗ -0.0057 -0.0036

(0.0137) (0.0165) (0.0119) (0.0142) (0.0144) (0.0172)

× 1990 0.0044 0.0114 -0.0276∗∗ -0.0284∗∗ 0.0248∗ 0.0168

(0.0143) (0.0164) (0.0115) (0.0137) (0.0144) (0.0164)

× 1995 -0.0320∗ -0.0253 -0.0115 -0.0125 0.0423∗∗∗ 0.0397∗∗(0.0166) (0.0194) (0.0120) (0.0144) (0.0156) (0.0184)

× 2000 -0.0452∗∗∗ -0.0361∗ -0.0066 -0.0056 0.0493∗∗∗ 0.0408∗∗(0.0167) (0.0187) (0.0118) (0.0140) (0.0162) (0.0184)

Municipality Fixed Effects Yes Yes Yes Yes Yes Yes

Year Effects Yes Yes Yes Yes Yes Yes

Geographic Controls No Yes No Yes No Yes

Clusters 743 743 743 743 743 743

r2 0.921 0.923 0.795 0.800 0.887 0.889

N 3671 3671 3671 3671 3671 3671

Note: Table 3.7 reports estimates from specification 3.6. Specifically, the table reports estimates

of the interaction of (ln) HYV adoption (between 1966-1980) and year indicators, 1980, 1990,

1995, and 2000. These estimated interactions convey the average growth in sectoral employment

for the interaction year relative to a 1970 levels. Columns (1) and (2) report outcomes for (ln)

agricultural employment share; columns (3)-(4), (ln) manufacturing; and (ln) services. All

estimates controls for municipality and year fixed effects. Columns (2), (4), and (6) include

geographic controls. Controls include (log) area irrigated, longitude, latitude, average rainfall,

average temperature, and the average municipality slope. Robust standard errors are clustered on

the municipality-level. Asterisks indicate statistical significance at the 1% ∗∗∗, 5% ∗∗, and 10% ∗levels.

152 REFERENCES

4. The Historical State, Local

Collective Action, and Economic

Development in Vietnam *

4.1 Introduction

The past century has witnessed a large-scale divergence in economic

prosperity within the developing world. In particular, initially poor economies

in Northeast Asia - such as Japan, Taiwan, and South Korea - have developed

much more rapidly on average than economies in Southeast Asia - such as the

Philippines and Cambodia.

One central difference between these regions is the nature of the historical

state, but it is challenging to deduce what role this played in the divergence

since many factors differ. Progress can be made by focusing on a single country

- Vietnam - that lies at the intersection of Northeast and Southeast Asia. This

study uses a regression discontinuity design to compare nearby Vietnamese

villages that belonged to different historical states, employing rich historical

data to elucidate channels of persistence. We hypothesize that the historical state

crowded in local collective action, and that these norms persisted, influencing

civic engagement, public goods provision, and economic development long

after the historical state had disappeared.

*This chapter was coauthored with Melissa Dell and Pablo Querubin. We thank Minh Trinh,

Nhung Le, Minh Tuan Nguyen, Thao Ngo, and Huyen Cao for providing excellent research

assistance. We are also grateful to seminar participants at Berkeley, the Canadian Institute

for Advanced Research, Central European University, Columbia University, CUNY, Duke

Economics, Duke Political Science, Harvard, IIES, LACEA, LSE, MIT, Munich, NYU, Oxford,

Queen Mary, UC-Santa Barbara, Universidad de Piura, University of Illinois Urbana-Champaign,

Warwick, World Bank and Yale for their helpful comments and suggestions. Support used to

fund this project was received from the Weatherhead Center for International Affairs (Harvard),

the Milton Fund (Harvard), and the Canadian Institute for Advanced Research.

153

154 The Historical State in Vietnam

Northeast Asia historically had central states with well-developed tax

systems, bureaucracies, and legal codes. Importantly, the village was the fun-

damental administrative unit. The central state set quotas for tax and military

conscript contributions at the village level and often designed legal codes with

an eye towards preventing local strongmen from becoming sufficiently pow-

erful to challenge the national government. These pre-modern states did not,

however, have the capability to micro-manage local administration, and villages

had considerable autonomy in how to implement policies. Villagers had to work

together to provide local public goods and meet the village-level tax and mili-

tary quotas. In contrast, Southeast Asian states followed a more decentralized

patron-client model. Power relations were personalized, with peasants paying

tribute and receiving protection from landowning patrons, who in turn had their

own network of relations with higher level patrons. The village was not a central

unit of administrative organization.

The northern Vietnamese state of Dai Viet was governed by China during

the first millennium CE, and it maintained many features of the Chinese state

following independence. During the 14th through 19th centuries, Dai Viet

gradually expanded southward, establishing village governance norms in what

is now central Vietnam (see Figure 4.1). Most of this expansion came via

conquering the territory of Champa, a relatively weak state that historically

occupied central Vietnam. Local affairs in Dai Viet villages were coordinated by

village councils elected by popular male suffrage.1 In contrast, the southernmost

part of modern Vietnam was historically a peripheral tributary of the Cambodian

state of Khmer. Khmer had a weak control of its periphery, and patron-client

relationships, not villages, were the central feature of local administration.

While some Vietnamese settlers had entered the region earlier, the Khmer areas

of modern Vietnam were not administratively organized along Vietnamese

lines until 1833, leaving little time for Vietnamese institutions to take root

before French colonial occupation in 1859. Colonial rule reinforced the pre-

existing differences, with the French exploiting village structures where they

were well-established and relying on more direct means of control where they

1Evidence on the impacts of elections in pre-modern states is limited, but Martinez-Bravo,

Padró i Miquel, Qian, and Yao (2012) show that in recent years the introduction of local elections

in China significantly increased public goods provision.


were not.

Table 4.1 summarizes key characteristics of the Khmer and Vietnamese

states. Section 4.2 discusses the many similarities between Dai Viet and other

Northeast Asian states, and between Khmer and other Southeast Asian states.

While care must be taken with external validity, these similarities suggest that

the Vietnamese context is informative about differences between Northeast and

Southeast Asia more generally.

This study examines the boundary between Khmer and Dai Viet, which

was established in 1698 and is denoted by a thick line in Figure 4.1. Villages on

one side belonged to Dai Viet for over 150 years prior to French colonization,

whereas villages on the other were organized along Vietnamese administrative

lines only a few decades prior to the arrival of the French. Section 4.2 docu-

ments that this boundary was meaningful and was the result of idiosyncratic

circumstances that prevented Dai Viet from conquering further Khmer territory

until they resolved an internal civil war. Geographic characteristics are similar

on either side, suggesting that villages just to the Khmer side are a reasonable

counterfactual for those just to the Dai Viet side. We also examine the other

boundaries of Dai Viet’s expansion but do not expect major long-run effects,

since villages along them were all governed by Dai Viet for centuries prior to

French colonization.

Using a regression discontinuity design to compare nearby villages on

either side of the Dai Viet-Khmer boundary, we find that household consumption

in Dai Viet villages is around one third higher today. Results are highly robust

to the selection of bandwidth and RD functional form. Ho Chi Minh City, the

administrative center of Dai Viet’s 1698 expansion, is in our study area, and the

estimates change little when it - and the entire surrounding region - is dropped.

They also change little when all provincial capitals are dropped, or when we

limit to villages along segments of the boundary not formed or only formed

by rivers. Moreover, the estimates are similar when we extend the sample to

all of South Vietnam, rather than focusing on the boundary. We document that

economic impacts also obtain historically, using data from the French colonial

period collected between 1878 and 1926, household income data from the

1970s, and a variety of other historical economic indicators.

After considering contemporary living standards, we examine channels of


persistence. While as always a number of mechanisms could be relevant, the

historical and empirical evidence suggests that local collective action - through

civil society and local government - is an important channel, and one that

is interesting in its own right. This would often be unobserved, particularly

historically, but Vietnam offers unusually rich historical data that allow us

to examine it. Detailed information on civil society and local government are

available for nearly all 18,000 South Vietnamese hamlets for the period between

1969 and 1973, and public opinion surveys were also collected for a random

sample of hamlets.

Citizens in Dai Viet villages historically had to work together to meet their

village’s obligations to the central state, provide public goods, and elect their

leaders, and these patterns persisted long after the dissolution of the Dai Viet

central state. Citizens in villages with a historically strong state were nearly

twice as likely between 1969 and 1973 to participate in local civic organizations.

They were also more likely to organize and participate in self-development

projects and local self-defense forces, to attend local government meetings,

and to have civic organizations that redistributed to needy households. Today,

though data on civil society are quite limited, available information shows

that households in Dai Viet villages are more likely to donate to charitable

organizations. These results - and the other results on mechanisms - survive

using latent class analysis to address multiple hypothesis testing and are robust

to the specification tests outlined above, including dropping Ho Chi Minh

City. They are also robust to controlling for village size. Barriers to migration

between villages in Vietnam are high, which may help explain why local social

capital has been so persistent, as argued in Besley (1995).

A 1967 constitutional reform granted villages expansive budgetary powers

and public goods provision responsibilities, with citizens electing village heads

and councils. This makes 1969 to 1973 a particularly illuminating time to

examine the historical state’s long-run impacts on local governance. Dai Viet

villages were more likely to collect taxes and the village head was more likely to

actually reside in the village. They were also more likely to have all the positions

filled on the village committee, which provides public goods. Importantly, Dai

Viet villages provided better access to basic health care, education, and law

enforcement. Citizens in Dai Viet villages reported that the local government


was more responsive to their needs and that local officials were more successful,

and they had better knowledge of the village administrative structure. These

results suggest that participatory governance reforms work best in places with a

history of participatory village governance.

There are no effects on public goods provided by the provincial govern-

ment and Dai Viet citizens had more negative views of the national government,

indicating that effects are unlikely to be driven by higher levels of government

or more positive attitudes towards government in general. More recently, we

continue to observe effects on access to secondary schooling, and in Dai Viet

areas, individuals have almost a year of additional schooling.

The literature finds that ethnic heterogeneity is an important determinant

of collective action, but it does not appear to be a direct cause of the effects we

find.2 While ethnic composition may have differed historically, today there is

almost no ethnic heterogeneity within villages. Nearly everyone identifies as

Vietnamese, which was also true 50 years ago.

The study also examines plausible alternative mechanisms. Extensive

evidence indicates that the effects are unlikely to be driven by differential

impacts of the Vietnam War, with a variety of measures suggesting that conflict

was similar across the boundary. Effects likewise do not appear to be driven

by recent land inequality. Dai Viet households are less likely to be agricultural

today, but within agriculture there is not a difference in average farm size.

Moreover, while 97% of French-owned land was located in Khmer areas at the

close of the colonial period, there were almost no French estates near the Dai

Viet boundary.

However, a lower share of land is formally titled in Dai Viet villages

today. This contrasts with conventional wisdom - based mostly on evidence

from the West - which associates capable states with the protection of formal

property rights and promotion of impersonal markets. Historically Dai Viet

implemented certain formal institutions - a legal code, an impersonal tax system,

a census registry, local elections, etc - whereas in Khmer areas personalistic

ties dominated. However, Dai Viet also emphasized redistribution of property

within the village, in order to prevent the emergence of local strongmen who

2See Alesina, Baqir, and Easterly (1999); Alesina and La Ferrara (2005); Bazzi and Gudgeon

(2016); Easterly and Levine (1997); Montalvo and Reynal-Querol (2005a,0)


could challenge the state. Accepted traditional norms of de facto property

management by civil society networks and local governments plausibly play a

more central role in ensuring property rights in Dai Viet villages than formal

titles introduced relatively recently by the central government in Hanoi. This

suggests a crowding out of formal institutions that is similar to the processes

documented in Greif (1994) and Greif and Tabellini (2010).

The study contributes to our understanding of the relationship between

civil society and the state. It is unclear from the existing literature whether

we would expect the state and civil society to be complements or substitutes.

On the one hand, scholars such as Gouldner (1980) and Fukuyama (1995)

argue that in the presence of a weak state, civil society emerges to substitute

the state by providing protection and social insurance, whereas a powerful

state can repress or co-opt any organizations that threaten it. Acemoglu, Reed,

and Robinson (2014) show that powerful ruling families in Sierra Leone are

able to co-opt civil society organizations. In contrast, Skocpol (1995) argues

that strong states can directly promote civil initiatives through legal protection

and public services, and the state’s legitimacy also relies on citizen’s active

participation and trust in institutions.3 Padro i Miquel, Qian, Xu, and Yao (2015)

document that Chinese villages with temples - a measure of social capital -

experienced larger increases in public goods following the introduction of

local elections.4 Sociologist Peter Evans (1995) hypothesizes that in Japan and

South Korea, a capable state and an active civil society are complements that

provided the engine for rapid economic growth. Our results suggest that when

the state and civil society do compliment each other, long-run growth is more

likely. Moreover, complementarities between the state and civil society are

consistent with the broader hypothesis that complementarities between culture

and institutions play a central role in generating persistence (see Alesina and

Giuliano (2015) for a review).

The study also speaks to an important social science debate about the

nature of village development. In his classic The Moral Economy of the Peasant,

James Scott (1977) argues that the village is the key institution of pre-capitalist

3See Lehning (1998), Hoover (2000) and Woolcock and Narayan (2000) for a review of the

theory on the relationship between states and social capital.4For more information on challenges for village governance in China see Qian (2014).


society, characterized by an adherence to social arrangements that insure vil-

lagers against subsistence crises. Similarly, Hayami (1980, p. 27) argues that

the East Asian village acts as “a community which mobilizes collective actions

to supply essential public goods.” In contrast, other scholars view prisoners

dilemmas, free rider problems, and other barriers to collective action - such as

zero sum mentalities - as prohibitive, and hence argue that when development

happens in rural areas, it is in spite of village social arrangements.5 Notably

Samuel ? argues that organizing to supply public goods is precisely what vil-

lagers find very difficult, due to the limited abilities of peasants to generate

village-wide insurance or welfare arrangements.

Interestingly, Scott conducted his fieldwork in central Vietnam, an area

where villages had been ruled by Dai Viet for many centuries, whereas Popkin

focused his work on Cochinchina, the French province in southern Vietnam

where the former Khmer areas are located. Given our results, it is not surprising

that they reached different conclusions. Some social scientists have examined

qualitatively why village collective action differs, focusing largely on environ-

mental determinism (Wade, 1994). This is essentially an efficient institutions

view (Demsetz, 1967) - collective action will emerge where environmental

factors lead the benefits to outweigh the costs. Our results provide quantitative

evidence for another important explanation, the village’s historical relationship

with the central state.

Finally, the study contributes more broadly to the literatures on the persis-

tence of social norms and institutions. The persistent impacts of the historical

state, despite the upheavals that came with colonialism, the Vietnam War, and

communism, are consistent with Roland (2012), who argues that culture in

transition countries is more influenced by participation in historical empires

for over 100 years than by the communist experience.6 Our results are also

consistent with work by Michalopoulos and Papaioannou (2013) and Gennaioli

and Rainer (2012) documenting that the organization of pre-colonial states

affects long-run prosperity in Africa. By focusing on a single country with rich

historical data, we are able to delve into mechanisms. More generally, this study

5See for example Foster (1965).6Our results are also plausibly consistent with Alesina and Fuchs-Schündeln (2007), who do

find that communism affected social values but argue that these effects will be short lived.


relates to a large literature that highlights the relevance of historical states and

historical institutions for long-run development and emphasizes the importance

of persistent cultural traits.7 In particular, the results support the strand of the

cultural economics literature that highlights social capital as a highly persistent

determinant of economic divergence (Guiso, Sapienza, and Zingales, 2016;

Putnam, Leonardi, and Nanetti, 1994). This literature has focused extensively

on Italy, whereas our results suggest that a social capital mechanism is also plau-

sibly at play in explaining the relative prosperity of Northeast versus Southeast

Asia, another central puzzle of modern economic growth.8

In the next section, we provide an overview of the historical context. Sec-

tion 4.3 discusses identification and section 4.4 tests whether the historical state

impacts contemporary living standards. Section 4.5 examines the mechanisms

through which the impacts of the historical state persist. Finally, Section 4.6

offers concluding remarks.

4.2 Historical Background

4.2.1 Historical Overview

For most of the first millennium, the northern part of modern Vietnam was

subject to Chinese overlordship. After gaining independence, the Vietnamese

state of Dai Viet - whose original borders are shown by the northernmost

polygon in Figure 4.1 - adopted the general political form of the Chinese state,

over time modifying it to Vietnamese needs. The Vietnamese state maintained

a competitive bureaucratic tradition, with an exam system used to select village

bureaucrats.9 In 1461 the system was reformed so that village councils were

7See Acemoglu, Cantoni, Johnson, and Robinson (2011); Acemoglu, Garcia-Jimeno, and

Robinson (2015); Acemoglu, Johnson, and Robinson (2001,0); Alesina, Giuliano, and Nunn

(2013); Becker, Boeckh, Hainz, and Woessmann (2016); Bukowski (2015); Dell (2010); Fer-

nández and Fogli (2009); Giuliano (2007); Grosjean (2011,1); Guiso, Sapienza, and Zingales

(2008); Lowes, Nunn, Robinson, and Weigel (2015); Luttmer and Singhal (2011); Nunn (2008);

Nunn and Wantchekon (2011); Oto-Peralías and Romero-Ávila (2014); Spolaore and Wacziarg

(2013); Tabellini (2008,1); Voigtländer and Voth (2012).8Of course, depending on what sorts of activities civil society engages in, it could also have

led to very negative experiences, as in Satyanath, Voigtlaender, and Voth (2013).9Porter, 1993, p. 4-5; Lieberman, 2003, p. 381-384; Woodside, 1971, p. 156-157; Do, 2003,

p. 53

4.2. HISTORICAL BACKGROUND 161

elected by villagers through popular male suffrage, with national bureaucrats

still selected through an exam system.10 These policies made Vietnamese local

governance unusually participatory, by global standards and relative to the

original Chinese model.11

Detailed legal codes institutionalized the relationship between the central

state, which served as the impetus and enforcer for most policies, and local

functionaries, who were responsible for implementation.12 The central state

imposed tax and military recruitment quotas on the village, leaving the village

authority to allocate tax burdens within their jurisdiction.13 The village main-

tained multiple population and property lists for the central state, and cadastral

records allowed for periodic land redistribution, as well as the collection of

property taxes beginning in the 1690s.14 The state also entrusted the village with

the supervision of public works.15 Historical consensus holds that Dai Viet state

capacity was “exceptionally penetrating by Southeast Asian standards.”16 Com-

paring Vietnam to Cambodia, Laos, and Thailand, Victor Lieberman (1993, p.

484) argues: “Chinese bureaucratic norms...tended to encourage in that country

[Vietnam] a more impersonal, territorially uniform, and locally interventionist

system than was found in Indianized polities to the west.”

Over hundreds of years, Dai Viet expanded southward (Figure 4.1).

Through its conquests it sought to make conquered territories integral to the Viet-

namese state. While conquered areas were initially settled as military colonies,

they were ultimately converted into Vietnamese administrative villages, whose

citizens had the same rights and obligations as areas that had been part of

Dai Viet for much longer.17 The Vietnamese first conquered the fragmented,

patron-client state Champa, which ruled central Vietnam through a system of

loose personalistic alliances.18 The Cham had been fully absorbed by the late

17th century (Figure 4.1), bringing the Vietnamese into conflict with the larger

10Meyer and Nguyen, 2005, p. 10311Cotter, 1968, p. 1612Haines, 1984, p. 309; Yu, 2001, p. 165; Lieberman, 2003, p. 382; Porter, 1993, p. 4-513Lieberman, 2003, p. 393; Zottoli, 2011, p. 10; Woodside, 1971; Porter, 1993, p. 5-614Yu, 2001, p. 165; Li, 1998, p. 49-56; Pastor, 199715Mus, 1949, p. 26616Lieberman, 2003, p. 38217Nguyen, 1985, p. 8-918Lieberman, 2003, p. 393


and more militarily powerful Khmer (Cambodian) empire to the south.

Dai Viet left behind a rich paper trail that historians have used to develop

a nuanced understanding of local and national political economy. In contrast,

the absence of a record-keeping state in the Khmer periphery has resulted in

very little quantifiable knowledge about life on the Khmer frontier prior to

Vietnamese invasion.19 Nevertheless the general features of Khmer society are

reasonably well-understood. The Khmer lacked a centralized bureaucracy, and

the state’s control over the periphery was weak.20 Southeast Asian historian

Shawn McHale (2013) argues that the Khmer periphery in Vietnam was the

lowland equivalent of highland Zomia in James Scott’s The Art of Not Being

Governed: an area with limited state capacity where peasants could escape the

exactitudes of the state.

In Khmer, political appointments and land distribution were personalistic,

and taxation was controlled by a temple-based system.21 Land-owning elites

solidified their claims to land by building a temple. They used the temple to

collect tribute from peasants and in turn passed a share up to higher level elites,

who legitimized their claims to land.22 Royal patronage, not administrative

specialization, were driving features of government service.23 Moreover, while

Dai Viet had a law code with nearly 1000 articles - 15 percent of them aimed at

protecting the existence of independent farmers - the Khmer legal code focused

instead on the preservation of patron-client relations.24

Table 4.1 summarizes key differences between the Khmer and Vietnamese

states.

Dai Viet and Khmer are representative more generally of Northeast and

Southeast Asian civilizations.25 The literature commonly divides Asian soci-

eties into two groups - the Indic states of Southeast Asia and the Sinic states of

Northeast Asia. Dai Viet, Korea (Choson), and Japan adopted a Chinese-style ad-

ministrative bureaucracy, including the exam system.26 They had Chinese-style

19Hall, 1968, p. 121-123,12620?, p. 231-234; Ebihara, 1984, p. 282; Osborne, 1966, p. 421Osborne, 1969; Sahai, 1970, p. 139-148; Chandler, 198322Lieberman, 1993, p. 227; Hall, 2011, p. 162; ?23Mabbett and Chandler, 1995, p. 166-167; Ebihara, 1984, p. 285; ?, p. 1024Ebihara, 1984, p. 285-286; Woodside, 1984, p. 318-319; Haines, 1984, p. 31025Hall, 1973; Cotterell, 201426Woodside, 1971; Woodside, 2006; Liu, 2007


legal norms, a high degree of centralization, and the village was a fundamental

administrative unit.27 All three had a tributary relationship with China at some

point, with political ties precipitating the adoption of Chinese statecraft.28

In contrast, a large literature on state formation in Southeast Asia classifies

Laos, Siam (Thailand), Bagan (Myanmar), and Khmer, as well as states such

as Srivijaya and Majapahit in island Southeast Asia, as Indianized “mandala”

states.29 Since at least the second century, most states across mainland and

island Southeast Asia were impacted by Hindu-Buddhist statecraft and elite

culture imported from India.

Bureaucracies, to the extent that they did exist, were never profession-

alized, even in the more centralized of the Indic polities and periods; central

states had weak control over the periphery; and the village was not typically a

fundamental administrative unit.30 The lines between private and state affairs

were blurred, and regime stability depended largely on monarchical personality

politics, as opposed to the codified rules of succession seen in China.31 South-

east Asian states also shared common Buddhist-Hindu legal origins, though

Islamic tradition eventually influenced Indonesia and Malaysia.32 Legal codes

tended to emphasize the preservation of patron client relations, which some

scholars have argued is due to the influence of the Indian caste system.33

Mere decades after the organization of the Khmer areas as Vietnamese

provinces in 1833, the French began colonizing Vietnam. Our study region

belonged entirely to the directly administered province of Cochinchina, es-

tablished in 1862. The French method of extracting surplus varied with the

pre-existing institutions.34 Where existing village structures were strong and

deeply rooted, they could be leveraged to meet extractive aims.35 In contrast,

where the village was weak and already lacked legitimacy, village leaders lost

27Jansen, 2000; Barnes, 2007; Liu, 2007; Lewis, 2009; Lewis, 2011; Kang and Cha, 2010;

Whitmore, 1979; Palais, 199628Kang and Cha, 2010; Kang, 201029Cœ dès, 1966; Mabbett, 1977a; Kulke, 1986; Tambiah, 1977; Wolters, 1999; Tambiah, 201330Lieberman, 1993; Lieberman, 200331Lieberman, 1987; Chutintaranond, 199032Hooker, 1978b; Hooker, 1978a; Harding, 2001; Acharya, 201333Mabbett, 1977b34Anh, 2003, p. 117; Booth, 200735Nguyen, 1985, p. 160


further legitimacy in attempting to collect taxes for the French. The French

relied on externally appointed officials to facilitate tax collection, and French

landowners took control of many estates that had previously been held by the

Khmer landed gentry.36 We digitized data on all French landownership in Viet-

nam at the close of the colonial period, and 97.5% of French lands in Vietnam

were on the Khmer side of the boundary.37 Nearly all of these lands are further

south than our study region, and thus are unlikely to explain our results, but the

overall patterns support the assertion that the French worked through existing

societal structures.

French colonial strategy halted the Vietnamization of former Khmer ter-

ritories that would have otherwise plausibly taken place. If the historical state

had differed, colonial policy would have as well, as it exploited and preserved

pre-existing norms. Colonial policy is hence an outcome that is important for

understanding persistence, not just in Vietnam but in a variety of Asian countries

that also experienced foreign interference.

Following World War II, the Vietnamese engaged in a successful anti-

colonial struggle against the French. The Geneva Accords of 1954 demarcated

Vietnam at the 17th parallel into two zones - communist North Vietnam and

pro-western South Vietnam - until elections to be held in 1956 would select a

unified Vietnamese government. These elections never occurred, and ongoing

conflict gradually escalated into the Vietnam War. Our study region is well

within South Vietnam, with the 17th parallel falling near the boundary of the

northernmost region in Figure 1. Importantly for our study, in 1967 there was

a major constitutional reform in South Vietnam that decentralized political

power, granting villages new budgetary powers, control over local councils, and

the ability to elect village councils and shape local development projects. Our

results from the South Vietnamese era thus shed light on the impacts of the

historical state in a context with a high degree of decentralization.

In 1975, Vietnam was reunited under a communist government, which

attempted unsuccessfully to collectivize land in the south and implement a

command economy. Liberalization began in the 1990s, and presently Vietnam

36Anh, 2003, p. 119; Osborne, 1969, p. 151; Popkin, 1979, p. 432; Wolf, 1969, p. 17737These data were compiled from French records by the Stanford Research Institute (Bredo,

1968).


is one of the more decentralized countries in Southeast Asia. Like the dynamics

of fifteenth century Dai Viet, local-state relations are critical and the Com-

munist party-led administration has continually attempted to penetrate to the

village-level. An administrative hierarchy defines a fiscal relationship between

the village and the central government, with provincial and district bodies in

between. Fiscal administration is conducted at the provincial level, whereas

village governments continue to play a role in administering a variety of ser-

vices.38 Officials are selected by communist party bodies.39 However, locally

selected village heads continue to exist in a more informal capacity, carrying

out important de facto functions in local politics along with village-level party

officials, although the central government does not formally recognize them.40

4.2.2 The Dai Viet - Khmer Boundary

The 1698 boundary between Dai Viet and Khmer is the southernmost

one in Figure 4.1, shown with a thick black line. Areas just to the east of

this boundary were part of Dai Viet for around 150 years prior to French

colonization, whereas areas just to the west were organized under Vietnamese

administrative lines in 1833, just decades prior to the commencement of French

colonization. Dai Viet exercised a strong control over its periphery, and the

Vietnamese state believed “firmly in well-defined borders as an alternative of

wayward conquering.”41 Systematic data do not exist for this region prior to

Vietnamese conquest. Nevertheless, the historical evidence suggests that the

location of the boundary is the result of a highly contingent set of historical

circumstances that with small perturbations would have produced different

boundaries, as opposed to reflecting underlying economic potential.

Upon completion of the conquest of Champa, southern expansion brought

Dai Viet into conflict with the Khmer kingdom. In 1623, the Vietnamese pro-

cured the rights from Khmer to establish a customs house at Prey Nokor (today

Ho Chi Minh Port), which was at the time a small Khmer fishing village. Prey

Nokor historically played a marginal role in Southeast Asian trade and “would

38See Kerkvliet and Marr (2004).39Marr, 2004, p. 4840Kerkvliet and Marr, 2004, p. 4-741Osborne, 1969, p. 13


only become important much later when it had been developed as an administra-

tive center [of Dai Viet].”42 Other natural ports such as Ha Tien, Ninh Kieu, and

Binh Long were located on the Khmer side of the boundary and initially played

a larger role in trade. Throughout the 17th century, Vietnamese settlers fleeing

civil conflict in Vietnam moved into the surrounding region.43 The Vietnamese

annexed much of the eastern Mekong as Gia Dinh Province in 1698, in part

using rivers to demarcate the territory, and the Khmer crown was unable to stop

this since they were engulfed in a war with Thailand. Dai Viet then set about

consolidating the area into administrative districts.44

Dai Viet would have likely annexed further Khmer territory in the imme-

diate aftermath of 1698, but this process was halted due to a contingent set of

circumstances at home.45 Vietnam had witnessed a series of bloody civil wars

between the Nguyen family in the south and the Trinh family in the north. In

1672, a truce was declared, and the country was split in two. Conquering a more

substantial chunk of Khmer territory further south would have required a full-

scale offensive by the Nguyen army against Khmer and Thailand, which also

aspired to conquer Cambodia.46 This would have left the Nguyen vulnerable

to an attack in the north from the Trinh. Civil conflict likewise constrained the

Khmer state, which had been in decline since the 15th century. The Khmer

crown oscillated between pro-Siamese and pro-Vietnamese royal factions in

a series of bloody conflicts.47 Vietnamese had settled in the Khmer areas dur-

ing the 18th century, albeit with a risk of ethnic cleansing, but they did not

have the political rights and obligations of those residing in Dai Viet.48 This

political equilibrium persisted until the latter quarter of the 18th century, when

large-scale conflict in Vietnam broke out. The Nguyen, who governed the south,

united all of Vietnam under their rule in 1802, allowing the annexation of the

remainder of modern Vietnam to proceed.

42Vickery, 1996, p. 415; Parthesius, 201043Cœ dès, 1966; Taylor, 2013, p. 303-31044Briggs, 1947, p. 35845Dieu, 1999, p. 1746Dieu, 1999, p. 1747Wook, 2004, p. 29348Taylor, 2013, p. 325-336; Engelbert, 1994, p. 170-175

4.3. ESTIMATION FRAMEWORK 167

4.3 Estimation Framework

This study’s research design exploits the discontinuous change in exposure

to the historical state across the Khmer-Dai Viet boundary, comparing house-

holds in areas incorporated into Dai Viet in 1698 to households in areas that

remained under Khmer. The boundary forms a multi-dimensional discontinuity

in longitude-latitude space, and regressions take the form:

outv � α+γDai Vietv + f (geographic locationv)+

n∑i�1

se giv +βdist_hcmv +εv

(4.1)

where outv is the outcome variable of interest in village v, and Dai Vietv

is an indicator equal to 1 if village v was on the Dai Viet side of the 1698 bound-

ary and equal to zero otherwise. f (geographic locationv) is the RD polynomial,

which controls for smooth functions of geographic location.

The se giv split the boundary into 25 km segments and equal 1 if village v

is closest to segment i and zero otherwise. The boundary segment fixed effects

ensure that the specification is comparing villages across the same segment of

the boundary, and the appendix shows that results are highly robust to the choice

of segment length. Finally, dist_hcmv is the distance of village v from Ho Chi

Minh City and is included in all regressions to explicitly control for proximity

to the region’s largest urban area. For regressions with equivalent household

consumption on the left-hand side, we also include a vector of demographic

variables giving the number of infants, children, and adults in the household.

The baseline specification limits the sample to villages within 25 kilo-

meters of the threshold. Following Gelman and Imbens (2014), we use a local

linear RD polynomial for the baseline and document robustness to a wide

variety of different bandwidths and RD polynomials.

The key identifying assumption is that all relevant factors besides treat-

ment vary smoothly at the Dai Viet-Khmer boundary. That is, letting c1 and

c0 denote potential outcomes under treatment and control, x denote longitude,

and y denote latitude, identification requires that E[c1 |x , y] and E[c0 |x , y] are

continuous at the discontinuity threshold. This assumption is needed for obser-

vations located just across the Khmer side of the boundary to be an appropriate


counterfactual for observations located just across the Dai Viet side.

To assess the plausibility of this assumption, Table 4.2 examines a variety

of geographic characteristics, using gridded geographic data and regressions of

the form described in equation (4.1). The unit of analysis is a 10 km x 10 km

grid cell.49 To be conservative, we treat grid cells as independent observations,

as the use of spatially correlated standard errors tends to slightly increase

their magnitude. Ideally we would be able to look at social and economic

characteristics before Dai Viet settled our study region, during the period when

the entire area was loosely controlled by Khmer. However, because the state

was weak, no systematic data were collected. Suitability for rice - the dominant

crop - was plausibly the most relevant characteristic given the agrarian nature

of the society at that time.

Columns (1) and (2) of Table 4.2 examine elevation and slope, respectively.

The point estimates on Dai Viet are small relative to the mean and statistically

insignificant. Column (3) shows that temperature is likewise balanced. Column

(4) does find a modest difference in precipitation that is marginally significant

at the 10% level, but the coefficient is quite small relative to the mean. Column

(5) documents that suitability for rice - the region’s principal crop - is similar on

either side of the boundary. Column (6) examines flow accumulation, a measure

constructed by the USGS Hydrosheds project that calculates how many cells

are uphill from the cell under question. The higher the number, the more water

we would expect to flow through the cell. There is not a statistically significant

difference. Finally, column (7) examines the kilometers of river flowing through

each cell, which is also balanced.

An additional assumption is no selective sorting across the treatment

threshold. This would be violated if the historical state provoked substantial out-

migration of relatively productive individuals from Khmer to Dai Viet, leading

to a larger indirect effect. The historical state would still exert long-run impacts,

but the interpretation would be different. As Dai Viet expanded southward, it

initially set up military colonies in newly acquired areas with settlers from the

north. This process happened throughout southern Vietnam prior to the arrival

of the French. However, there is little evidence of selective migration from

Khmer to established Dai Viet villages. The historical literature argues that

49Results are similar when other sized cells or villages are used as the unit of analysis.

4.4. LONG RUN EFFECTS ON ECONOMIC PROSPERITY 169

negative attitudes towards outsiders created substantial barriers to moving into

established villages:

“An outsider who was allowed to live in a village had fewer

rights to village possessions than did insiders. His descendants, fur-

thermore, might not receive full citizenship–and with it, the right

to own property and be notables–for several generations. Such

marked distinctions made it exceedingly difficult, if not impossible,

for a man to move into a village and take over another man’s land.

Even well into the period of French rule, a person from another vil-

lage who tried to farm was likely to have his crops destroyed...The

emphasis on village citizenship, therefore, encouraged local own-

ership" (Popkin, 1979, p. 89).

Moreover, the Pacification Attitudes and Analysis Survey, conducted in

the early 1970s, asked individuals if they would hypothetically be willing to

move to a different village or province if they received an offer for a higher

paying job. Only 21% and 12% of respondents answered yes, respectively.

Finally, we use the 2009 census to compare current place of residence to

place of residence in 2004 and find low levels of migration between historically

Khmer and Dai Viet areas. 2.5% of households in areas historically under

Dai Viet reported having lived in historically Khmer areas in 2004. 1% of

households in historically Khmer areas reported having lived in historically Dai

Viet areas in 2004. While migration is unlikely to be a primary driver of results,

we will examine its potential effect on the estimates in Section 4.4.

4.4 Long Run Effects on Economic Prosperity

This section examines the impacts of the historical state on economic

prosperity across the past century and a half. It begins by considering effects on

contemporary household consumption and then explores a variety of historical

measures of economic activity.

The biennial Vietnam Household Living Standards Surveys (VHLSS)

were collected between 2002 and 2012 by the General Statistics Office of


Vietnam with technical assistance from the World Bank.50 The set of sampled

villages remains mostly constant across 2002-2008, and then changes substan-

tially in 2010. In each year, a core survey is administered to a large number

of households, and an additional module on expenditures is administered to a

subsample of households. In order to create a panel, there is a 50% rotation of

households from one survey round to the next. To avoid repeated observations

for the same household, we drop all households in 2004 that were also surveyed

in 2002, all households in 2006 that were also surveyed in 2004 and so forth.

Results are quantitatively similar if all observations are retained (Appendix

Table A.1). To construct a measure of consumption that reflects productive ca-

pacity, we subtract transfers received from total consumption, though estimates

are similar when transfers are included (Table A.2).51

Table 4.3 reports estimates from equation (4.1), using the log of equivalent

household consumption as the dependent variable. Following Deaton (1997),

we assume that children aged 0 to 4 are equal to 0.4 adults and children aged 5

to 14 are equal to 0.5 adults. All regressions control for survey year fixed effects

and the number of household members aged 0-4, 5-14, and 15 and older.52

Standard errors are clustered at the village level, and none of the significance

levels in Table 4.3 change if errors are clustered at a higher administrative level

or adjusted for spatial dependence.

Overall, the point estimates suggest that household consumption is around

a third higher in Dai Viet villages. Column (1) uses a local linear polynomial

in latitude and longitude, whereas column (2) uses instead a local linear poly-

nomial in distance to the boundary, and column (3) includes both. Results

are similar across these specifications. In a regression discontinuity there are

many options for how to specify the RD polynomial and bandwidth. We fol-

low Gelman and Imbens (2014) in specifying the baseline as a local linear

polynomial but are not aware of a widely accepted optimal bandwidth for a

multi-dimensional RD, employing a variety of outcomes. Fortunately the choice

of bandwidth and RD polynomial makes little difference. Each panel in Figure

4.2 plots point estimates of γ using equation (4.1) and different bandwidth

50The survey was collected during the 1990s but only for a very small sample of villages.51We classify transfers as remittances and gifts received by the household as well as all income

from social welfare and charity organizations.52Household demographics are balanced.


values between 10-100 kilometers, with the bandwidth under consideration

denoted on the x-axis. Thin lines show 95% confidence intervals while the

slightly thicker lines show 90% confidence intervals. The panels in different

rows employ different functional forms for the RD polynomial: linear latitude-

longitude polynomial (row 1), linear distance to the boundary polynomial (row

2), both linear latitude-longitude and linear distance to the boundary polyno-

mials (row 3), and analogous specifications using quadratic functional forms

(rows 4 through 6). The estimates in the first column include the full border

and show that impacts are remarkably robust to alternative bandwidth and RD

polynomial choices, though by construction estimates for smaller bandwidths

tend to be noisier, particularly for quadratic polynomials. The second column

shows that estimates change little when limiting the sample to portions of the

boundary not formed by rivers.

The results can be seen graphically in Figure 4.3. Each sub-figure shows

a scatter plot for one of the paper’s main outcomes. These plots are the three-

dimensional analogues to standard two-dimensional RD plots, with each vil-

lage’s longitude on the x-axis, its latitude on the y-axis, and the outcome shown

using an evenly-spaced monochromatic color scale. The background shows

predicted values, for a finely spaced grid of longitude-latitude coordinates, from

a regression for the outcome under consideration using equation (4.1). In the

typical RD, the predicted value plot is a two-dimensional curve, whereas here it

is a three-dimensional surface, with the third dimension indicated by the color

gradient. Lighter shades indicate higher values. The data are not binned, the way

they often are in a two-dimensional RD, so will tend to show greater variation.

Panel (a) for household consumption illustrates the predicted jump across the

boundary and darker dots tend to overlay darker-shaded areas, indicating a good

fit.

The cluster of points on the Dai Viet side of the boundary is Ho Chi Minh

City, and one concern is that it drives the effects. Its placement is not by chance -

it was the administrative center of Dai Viet’s 1698 expansion - but if it drove the

results the interpretation would be different. While Ho Chi Minh City is near

the boundary, it does not directly border the boundary, and hence its influence

on the RD estimates is limited. Column (4) shows that results barely change

upon dropping all urban districts comprising Ho Chi Minh City, and column (5)


documents that results are also unchanged when all of Ho Chi Minh Province

- which includes urban and rural areas - is excluded. Column (6) shows that

results are also robust to dropping all provincial capitals, which largely removes

urban areas.

Villages in Dai Viet do tend to be slightly closer to Ho Chi Minh City, but

this does not change discontinuously at the boundary. Hence, the RD controls

for it. One additional concern, though, is that travel costs could be discontinuous

along boundary segments formed by rivers - today bridges are widespread but

this may not have always been true. On the other hand, river segments might be

preferred to the extent that they constitute exogenous barriers that were used to

separate different historical polities. Column (7) limits the sample to villages

closest to boundary segments that do not coincide with rivers, and column

(8) does the same for segments that are formed by rivers. The point estimates

are of similar magnitude, suggesting that effects are unlikely to be driven by

discontinuous travel costs or by unobservables that change discontinuously

only along non-river segments. This robustness extends across bandwidths and

RD polynomials (Figure 4.2). Another alternative hypothesis is that results

could be driven by higher levels of government. Provinces change across the

study period, so we aggregate these changes to create provinces with consistent

boundaries across time. Comparing villages within these if anything makes the

estimates larger (column 9).

An additional question to consider is whether differential rates of mi-

gration today may be responsible for living standards differences across the

boundary. Given that in-migration to provinces historically under Dai Viet is

about 2.5%, we omit the 2.5% of the Dai Viet sample with the highest consump-

tion. To be conservative we similarly omit the 1% of the Khmer sample with

the lowest consumption, as in-migration to Khmer areas is 1%. The estimate

in Column (10) based on the trimmed sample remains similar, indicating mi-

gration today is not large enough to drive the differences. We have no way to

measure migration - or how selective it was - historically. However, as discussed

in Section 4.3, the fact that migrants and their descendants faced substantial

barriers in gaining full village citizenship suggests that selective migration is

unlikely to fully drive our results.

A final concern is that the boundary may be an unusual place. We address


this by examining two alternative samples. The first considers only places 25-

100 km from the boundary, omitting the boundary region itself. The second

compares all of South Vietnam that belonged to Dai Viet historically to all of

South Vietnam that belonged to Khmer.53 While these estimates are no longer

causally identified, they remain very similar to the baseline estimates, demon-

strating that the effects near the boundary are not a fluke. Results (available

upon request) are also robust to dropping other places that may be unusual, such

as coastal villages.

The appendix documents additional robustness. The baseline specification

uses 25 km boundary segment fixed effects. Appendix Tables A.3 through

A.6 show robustness to instead including 100, 75, 50, or 10 km segment fixed

effects, respectively. Tables A.7 and A.8 document that results are broadly

similar before and after the redefinition of the VHLSS sample in 2010.

Finally, Appendix Table A.9 reports several placebo tests. First, the rivers

coinciding with the Dai Viet boundary also flow through areas that are not along

the boundary (Appendix Figure 4.1), providing an additional opportunity to

examine whether estimates simply capture the effect of being on different sides

of a river. Column (1) estimates the baseline regression on the sample of districts

bordering other portions of the rivers that partially form the boundary, assigning

as treated whichever side of the river segment is richer. The difference across the

rivers is statistically insignificant. Column (2) performs a placebo comparing

across the provincial boundaries in the study area, in order to see whether

income differentials of the magnitude found along the Dai Viet boundary are

typical. This requires assigning some places as treated and others as untreated,

and in order to stack the test in favor of finding a difference we assign the richer

side of each provincial boundary segment as treated. The specification does not

reveal a statistically significant discontinuity. Finally, Column (3) considers a

placebo test using other historical boundaries of Dai Viet’s southward expansion.

To increase power, we pool all observations in proximity to the 1306, 1407,

1471, 1611, 1651 and 1693 boundaries, and the treatment indicator equals 1

if the district is located on the side of the boundary conquered earlier. Since

53We focus on South Vietnam to increase comparability, since the North had a very different

history under Communist North Vietnam between independence and reunification. We do not

include an RD polynomial since the sample contains places very far from the boundary, but

estimates are larger when it is included.


all of these places were organized under the village government system for

hundreds of years, we would expect there to be little or no long run effect of

being brought into this system modestly earlier. This contrasts to the Khmer-Dai

Viet boundary, where villages on the Khmer side were organized along Dai

Viet administrative lines for only a short time before French colonization. The

estimate is indeed small and statistically insignificant.

The historical state has a robust impact on current living standards. We

turn next to an examination of economic variables across the past century and

a half, in order to examine whether similar effects obtain historically. Data

from the pre-colonial period are not systematically available, in particular for

the Khmer side of the boundary. When the French arrived, they did collect

some systematic data, but disaggregated data are rare. The only source of extant

village level information is maps held by the Bibliotheque Nationale de France,

for 1878, 1901, 1910, and 1926. Each map shows different types of infrastruc-

ture - specifically roads, railroads, telegraph lines, and military posts - though

not all types of infrastructure are shown in all maps. Since our entire study

region is within the same colonial administrative unit, Cochinchina, we would

not expect these outcomes to differ if pre-existing conditions were not different.

The colonial state and private companies plausibly invested in transport and

communications infrastructure in areas with the greatest economic surplus. We

georeferenced these maps and intersected them with village boundaries.

Table 4.4, columns (1) and (2) consider density (in km per village area)

of telegraph lines in 1878 and 1901, respectively. Telegraph lines were more

prominent on the Dai Viet side of the boundary, and the coefficients are large

relative to the sample means. In contrast, while only 3% of villages in the sample

contain a military post, these were less prominent in 1878 in Dai Viet areas,

plausibly because these villages could be more easily governed through the pre-

existing village structure (column 3). The 1878 map also shows lines denoting

a rail or road (of any type), and there is not a statistically significant difference

across the boundary (column 4). However, by 1910, the maps reveal that railroad

density was higher in Dai Viet villages (column 6), and the coefficient on motor

roads (which may be paved or unpaved) is positive but noisy (column 5). Finally,

the 1926 map shows a strong positive effect on paved roads (column 7). The

railroad effect, in contrast, is no longer statistically significant, and the density


of railroads by this time had fallen nearly in half relative to 1910 (column 8).

Appendix Figure 4.2 shows that these estimates are broadly robust to the choice

of bandwidth. We plot coefficients only for the linear RD baseline specification,

to avoid reporting tens of thousands of coefficients over the course of the study,

but estimates are also similar when other RD polynomials are used.

Next, we turn to data from the South Vietnamese period. Income data

are available for a sample of hamlets through the Pacification Attitudes and

Analysis Survey (PAAS, U.S. National Archives RG 330 and 472). PAAS was

a joint U.S.-South Vietnamese effort, in which responses were compiled by

Vietnamese enumerators. It was launched in March of 1970 and was conducted

monthly until December of 1972, though unfortunately not all months have

been preserved.54 Each month, surveys were conducted in 6 randomly selected

hamlets per province. 15 respondents were randomly selected per hamlet, with

stratification on demographic characteristics. The survey focused on citizens’

attitudes and opinions, but also asked about household income in the past year.

Households identify which income bin describes their situation, and we assign

their income as the midpoint of the bin. The data are not of the same quality

as modern expenditure surveys but are nevertheless a rare example of income

measurement in a developing country before the advent of living standards

surveys.

We also obtain a variety of economic indicators from the Hamlet Evalua-

tion System (HES, RG 472), collected jointly by the United States and South

Vietnam between 1969 and 1973. HES contains information on economic, so-

cial, political, and security conditions in all South Vietnamese hamlets, with

data collected on a quarterly basis.55 The information was compiled by US

and Vietnamese advisers, in conjunction with local officials, and the questions

are at the hamlet or village level. The data provide unusually rich local level

information covering a broad set of variables.

The HES questions have categorical responses. We code questions with

multiple responses into binary indicators that preserve variation (see the data

appendix for more details). For example, a coding of a question about non-rice

54Tapes containing information for May, 1970 through February, 1971 and for August and

September of 1971 were not preserved.55Most questions are quarterly, but a few of the security questions were collected monthly.


food availability into no or limited availability versus ample availability pre-

serves significantly more variation than alternative codings, since few hamlets

completely lack non-rice staples.56 These indicators are then averaged across

the full period of data availability. We also report estimates from a Bayesian la-

tent class analysis, to address potential concerns about the coding of categorical

responses and multiple hypothesis testing. Based on the observed categorical

responses to all economic questions, latent class analysis estimates the posterior

probability that each hamlet belongs to one of two latent groups associated with

“high” and “low” economic prosperity.57

Estimates for economic outcomes during the South Vietnamese period

are reported in Table 4.5. Column (1) examines log household income between

1970 and 1972. Income on the Dai Viet side of the boundary is around 16

percentage points higher, and the estimate is statistically significant at the 1%

level. Column (2) documents that hamlets historically under Dai Viet are 16

percentage points more likely to be in the high prosperity latent class (s.e.=

0.055), and the effect is significant at the 1% level. See also Figure 4.3, panel

b). The results for the individual outcomes that contribute to the LCA show a

similar pattern. In places with a strong state historically, non-rice foodstuffs are

28 percentage points more likely to be amply available (column 3, s.e.= 0.06),

and manufactured goods are 20 percentage points more likely to be available

(column 4, s.e.= 0.07). Surplus goods are also more likely to be produced,

households are less likely to require assistance to subsist, and households

are more likely to have access to a vehicle (columns 5-7). However, there is

no difference in whether land is left fallow due to poor security (column 8).

The next section will show that security did not differ substantially across the

Dai Viet boundary, alleviating the concern that these effects could be largely

driven by the war. Finally, column (9) shows that there is no difference in

quarterly population growth, suggesting that differential migration during this

period is unlikely to contribute substantially to the effects. Appendix Figure 4.3

56An alternative would be to estimate a multinomial logit, but this does not converge well

since there is often little variation in some of the categories.57We include questions that are available for the full sample period. Results are similar if we

include questions that were only asked during part of the sample period. When we compute the

LCA, we include all observations, to avoid needing to recompute the LCA when changing the

bandwidth. However, if we just include hamlets within 25 kilometers of the boundary in the LCA

computations, results are very similar.

4.5. MECHANISMS 177

documents that results are broadly robust to the choice of bandwidth.

One concern is that the results could be partially driven by hamlet size.

While we do not control for this in the baseline, since it is endogenous, Table

A.10 documents that results are similar when this control is included. Results

are also similar when we drop Ho Chi Minh City, which at the time was its own

province, or when we drop all provincial capitals (Tables A.11 and A.12).

Finally, we digitized district level data on land ownership and rice cul-

tivation during 1975-1985, the period after Vietnamese reunification, from

provincial yearbooks and declassified Vietnamese Communist Party documents.

The main drawback of these data is that there are relatively few districts, and

thus we lack statistical power. In order to have enough observations for regres-

sion analysis, we extend the bandwidth to 100 kilometers. Estimates in Table

A.13 suggest that districts in historically Dai Viet areas had a higher share of

state-owned land, a lower share of private land, and may have had less land

cultivated with paddy rice.58 Conditional on land being in paddy, it was more

likely to be irrigated and mechanized. Most effects are marginally significant at

the 10% level, except the effect on land in paddy, which falls short of statistical

significance but is large in magnitude. While these effects are noisy, they are

highly consistent with the persistent impacts of the historical state on economic

conditions, documented from the mid-19th century through the present.

4.5 Mechanisms

This section explores mechanisms through which the historical state has

influenced long-run development. Our hypothesis is that Dai Viet’s long history

of village government translated into greater local collective action that has

persisted through a series of upheavals following the dissolution of Dai Viet,

including colonial conquest, the Vietnam War, and an effort to implement a

communist command economy. Long after the Dai Viet central state ceased

to exist - and even in the face of subsequent state policies that have at times

aimed to discourage local collective action - these norms have persisted and

have plausibly had important impacts on local public goods provision and

58In addition to state and private land, the third category of land is collectively farmed land. In

1979, when these data were compiled, there was no collectively farmed land in our study region.


economic development. We do not claim that local collective action is the

only mechanism linking the historical state to long-run development, but the

historical and empirical evidence make it difficult to tell a story where it does

not play an important role.

Before examining the data in detail, it is useful to consider one important

reason local collective action may vary across the boundary - ethnic hetero-

geneity.59 In the past ethnicity may well have differed, but more recently the

vast majority of people throughout the region self-identify as Vietnamese. In

VHLSS, 97% of respondents are Vietnamese and almost none identify as Khmer

or Cham. The only minority group of any quantitative significance is the Chi-

nese, who are concentrated in large urban areas. Results change little when

these areas are dropped. Within villages, there is almost no ethnic diversity.

Moreover, during the 1960s and 1970s, Vietnamese is identified as the primary

ethnicity in over 98% of hamlets. While it is possible that individuals could

self-identify as Vietnamese but still practice the customs of other ethnic groups,

we do not find differences across the boundary in patrilocal marriage patterns,

an important difference between Northeast and Southeast Asian ethnicities.

Engagement with civil society and local government - two important com-

ponents of local collective action - is often unobserved, particularly historically,

but between 1969 and 1973 the U.S. and South Vietnamese governments com-

piled unique local level data on civil society, village government, and public

opinion. The Hamlet Evaluation System (1969-1973) contains monthly and

quarterly data on economics, civil society, local government, and security for

nearly all of South Vietnam’s 18,000 hamlets, and the PAAS Public Opin-

ion Survey (1970-1972) provides public opinion data for a random sample of

hamlets. These data are described in more detail in the previous section.

Table 4.6 uses the HES data to examine non-communist civil society (the

Viet Cong insurgency will be considered subsequently). To address potential

concerns about multiple hypothesis testing, we start by constructing a summary

measure using Bayesian latent class analysis and the individual civil society

questions, as described in the previous section. The dependent variable in

column (1) is the posterior probability that the hamlet is in the high civil society

59See Alesina et al. (1999); Alesina and La Ferrara (2005); Bazzi and Gudgeon (2016);

Easterly and Levine (1997); Montalvo and Reynal-Querol (2005a,0).

4.5. MECHANISMS 179

group. Hamlets historically under a strong state are 18 percentage points more

likely to be classified in the high group (s.e.= 0.035), relative to a sample mean

posterior probability of 0.76. This effect is shown in Figure 4.3, panel c).

Columns (2) through (12) consider individual outcomes. In hamlets his-

torically under Dai Viet, households are 26 percentage points more likely to

participate in civil organizations, relative to a sample mean of 0.37 and are

21 percentage points more likely to participate in local economic trainings

(columns 2-3). Both effects are significant at the one percent level. Households

in Dai Viet villages are 8 percentage points more likely to participate in the

People’s Self Defense Forces - a local self defense organization - and the effect

is statistically significant at the one percent level (column 4). Dai Viet villages

are also substantially more likely to have self-development projects underway

(column 5). We do not find that the village council is more likely to discuss

citizens’ grievances with them - the mean of this variable is over 90% - nor are

Dai Viet villages more likely to have organized youth activities (columns 6-7).

Dai Viet households are 10 percentage points more likely to attend meetings

held by the village government, relative to a mean of 0.37 (column 8). In the

final columns, we consider outcomes that may be selected - participation in

Rural Development (RD) Cadre activities and civil organization assistance to

needy households. Not all villages have RD Cadre - South Vietnamese devel-

opment aid workers - or households that require assistance to subsist, but only

villages that have RD cadre can participate in their activities and only villages

that answer that they have households requiring assistance to subsist can then

specify that they have civic organizations to provide this assistance. Otherwise,

the response to these is coded as zero.60 We do not find a difference in whether

the RD cadre are present in the hamlet, and in Dai Viet villages, households are

more likely to participate in RD cadre-initiated development activities (columns

9-10). In contrast, Dai Viet villages are substantially less likely to have house-

holds that require assistance to subsist (column 11). Nevertheless, they are 17

percentage points more likely to have organizations that provide assistance

to such households, relative to a sample mean of 0.24 (column 12).61 These

60These outcomes are not included in the LCA since they may be selected, but LCA impacts

would be even stronger if they were included.61The response to this question is 0 by construction if villages don’t have such households.


results are broadly robust to the selection of bandwidth (Appendix Figure 4.4),

to controlling for population, to dropping Ho Chi Minh City, and to dropping

all provincial capitals (Tables A.14 to A.16).

Next, we turn to local governance and public goods provision, with out-

comes drawn from HES (1969-1973). During this period, a constitutional reform

had decentralized many governance and public goods provision roles to the lo-

cal level, making it a time when local government was particularly relevant. As

discussed in the introduction, it is not clear from the existing literature whether

we would expect impacts on local governance to go in the same direction as

those on civil society.

The results indicate that local governance and civil society in Vietnam

are complements. They also suggest that participatory decentralization reforms

work best in areas with a history of participatory governance. Table 4.7, Column

(1) considers the posterior probability that the hamlet belongs to the class

associated with good local government administration. Dai Viet areas are more

likely to be classified in the high local governance class, and the effect is

statistically significant at the 5% level.

A classic measure of state capacity is tax collection, and column (2) shows

that local governments in Dai Viet villages are six percentage points more likely

to systematically collect taxes, relative to a sample mean of 0.84. Column (3)

documents that Dai Viet villages are also more likely to have all the positions

on their village committee - which organizes public goods provision - filled.

The village chief is more likely to be regularly present in Dai Viet villages,

though there is not a statistically significant effect on the presence of the hamlet

chief (columns 4 and 5). Moreover, police are 17 percentage points more likely

to be regularly present, relative to a sample mean of 0.56 (column 6). Next we

consider outcomes that might be selected - the village head’s control over the

RD cadre and the provision of government assistance. There is not a difference

in the presence of RD cadre (Table 4.6, column 9), but village heads in Dai

Viet villages are more likely to effectively control the RD teams (column 7).

Households in Dai Viet villages are less likely to require assistance to subsist

(Table 4.6, column 11), but despite this governments in Dai Viet villages are 14

percentage points more likely to provide assistance, relative to a sample mean

of 0.3 (column 8).

4.5. MECHANISMS 181

Examining policies under the control of the provincial government can

serve as a useful placebo. In principal, the historical state could affect provincial

governments or the distribution of provincial resources, but if the mechanism

is about organizing within the village we would expect these effects to be less

pronounced. HES asked three questions about provincial government: whether

projects have failed due to a lack of provincial technical assistance (column 9),

whether provincial technical personnel - such as agricultural extension workers

- regularly visit the village (column 10), and whether provincial land affairs

officials visit the village to assist with land reform (column 11). All coefficients

are small relative to the mean and statistically insignificant.

Local public goods provision also shows a discontinuity at the Dai Viet

boundary. Data are available for health care, education, and law enforcement,

the principal public goods in this context. Table 4.8, Column (1) reports the LCA

for health care provision, documenting that Dai Viet villages are substantially

more likely to be in the high health care provision latent class (see also Figure

4.3, panel d)). Government-provided medical services are 20 percentage points

more likely to be available, relative to a sample mean of 0.39 and mobile health

workers are more likely to regularly visit all hamlets in the village (columns

2-3). There is more likely to be a health clinic in the village, but we do not find

an impact on the presence of a maternity clinic (columns 4-5).

We also consider education. The LCA is not statistically significant,

though it is positive (column 6). This is partially driven by the fact that there is

not a difference in whether school attendance is restricted by poor security (col-

umn 11). Dai Viet villages are 6 percentage points more likely to have access

to a primary school, and the primary school completion rate is 9 percentage

points higher, relative to a sample mean of 0.61 (columns 7 and 8). There is not

a statistically significant impact on whether there is a secondary school in the

village - though the coefficient is large and positive - but the secondary school

attendance rate is 3 percentage points higher, relative to a sample mean of 0.18

(columns 9 and 10). Finally, in Dai Viet villages, authorities are 22 percentage

points more likely to enforce the law day and night (column 12). The local

administration and public goods results are robust to alternative bandwidths,

to controlling for population, to dropping Ho Chi Minh City, and to dropping

provincial capitals (Figures 4.5 to 4.6 and Tables A.17 to A.22).


Public opinion data, examined in Table 4.9, can corroborate the above re-

sults with an independent data source and also provide additional nuance. Since

different questions are asked in different months, the number of observations

can vary substantially.

Column (1) documents that respondents in Dai Viet villages are 9.8 per-

centage points more likely to report that their local government is responsive

to the needs of its citizens, relative to a sample mean of 0.37. They are also 20

percentage points more likely to report that their local government is successful,

relative to a sample mean of 0.52, and they have more knowledge of their vil-

lage administrative structure (columns 2-3). This is consistent with the previous

result that they are more likely to attend government meetings. Respondents in

Dai Viet areas are also 35 percentage points more likely to feel that the Land to

Tiller (LTT) program - South Vietnam’s land reform - was administered fairly

in their village (column 4).

These views do not extend to the national government (columns 5-6).

Respondents in Dai Viet villages are 11 percentage points more likely to respond

that the national government performs poorly, an effect that is significant at the

one percent level. They are also more likely to feel that the national government

has done a poor job of managing the economy. These results suggest that effects

cannot be explained by Dai Viet areas having more positive attitudes towards

government in general.

A final set of questions considers civic engagement. These have a smaller

sample size and should be interpreted cautiously. Respondents are asked who

has the primary responsibility to make community life better, the people or the

government, and respondents in Dai Viet villages are 27 percentage points more

likely to feel that this is the responsibility of the people (column 7). This may

indicate a less severe free-rider problem, in which citizens recognize their own

responsibility to improve village life. Dai Viet respondents are also more likely

to be active in an interest group and are more likely to report that the people of

the village decide which self-development projects will be implemented, rather

than government heads (columns 8-9). The public opinion results are broadly

robust to alternative bandwidths (Figure 4.7) and to controlling for population,

dropping Ho Chi Minh City, and dropping all provincial capitals (Tables A.23

through A.25).

4.5. MECHANISMS 183

Current data about civil society, local government, and public opinion

are sparsely available for Vietnam, a communist country where officially civic

engagement occurs through the Party and local government is managed by

communist officials. As discussed in Section 4.2, informal institutions reflecting

village structures remain active, but the state has been hesitant to acknowledge

or collect information on these arrangements. Unlike many household surveys,

VHLSS does not have a social capital module. Nevertheless, in the available

data, we continue to see legacies of the historical state. The closest question

to civic engagement systematically available in VHLSS asks about household

expenditures on donations to charitable organizations, and we code an indicator

for whether the household has donated to a charitable group. Table 4.10, column

1 documents that households in Dai Viet villages are 12 percentage points more

likely to make charitable contributions (see also Figure 4.3, panel e).

Columns (2) through (6) consider human capital. Column (2) uses district-

level information from provincial yearbooks (1999-2004) on the share of com-

munes in each district with a secondary school, showing a greater prevalence

in Dai Viet areas. In order to have enough districts for regression analysis, we

need to extend the bandwidth to 100 km, so this estimate should be interpreted

with caution. We do not examine primary schools or health posts because access

today is nearly universal. Columns (3) through (6) use individual-level data

from VHLSS on years of schooling. Column (3) reports the average effect for

all individuals over 25, whereas columns (4) through (6) consider different co-

horts separately. We focus on adult cohorts as they are likely to have completed

schooling. The estimates are positive and statistically significant, documenting

that individuals in areas historically under a strong state have an additional 0.9

years of schooling. This is shown graphically in Figure 4.3, panel f). While the

absolute effect is roughly similar across cohorts, the effect is proportionally

larger for older individuals, since the older cohort has only half the schooling

of the younger cohort. The estimates are consistent with the historical results

on access to schooling and suggest some convergence over time. The direct

impacts of education today are large enough to explain about a third of the

economic differences, using typical returns to education. Figure 4.8 documents

that the estimates are broadly robust to the choice of bandwidth. Tables A.26

through A.28 show that results remain similar when Ho Chi Minh City and


Province are dropped and when provincial capitals are dropped.62

This study documents that the historical state exerts long-run impacts on

collective action. There could be many mechanisms linking the historical state

to economic outcomes, but we argue that it is difficult to tell a compelling story

where this does not play a role. Figure 4.11 shows that there is a strong relation-

ship in the raw data between the civil society LCA and the economic LCA, as

well as between the civil society LCA and the health care and education LCAs.

Providing local public goods that promote development inherently requires

working together. Villages that are richer might be able to afford to invest more

in local collective action, creating a virtuous feedback loop that is sustained in

the long-run.

We next consider some other potential mechanisms, starting with the

Vietnam War. The war is unlikely to be an omitted variable. Our study region

is in the same military corps region, with no reason to expect military strategy

to change discontinuously at the boundary, and we find economic impacts of

the historical state prior to the war. Moreover, Dell and Querubin (2016) do not

find long-run effects of bombing - one of the most destructive features of the

war - using the same data sources as this study, and Miguel and Roland (2011)

likewise find no long-run impacts of bombing.

If the historical state impacted insurgency, the war could be a contributing

mechanism, but Table 4.11 finds little evidence for major differences in conflict

across the Dai Viet boundary, using a variety of detailed data drawn from

the U.S. National Archives. Column (1) considers the security LCA, which

combines the security questions available in HES. The coefficient is small in

magnitude and statistically insignificant. Columns (2) through (7) examine

representative individual outcomes that enter the LCA. Columns (2) and (3)

do not find impacts on whether there are Viet Cong (VC) forces or a VC

base nearby. Dai Viet villages are 6.5 percentage points more likely to have

a VC village guerrilla squad, which consists entirely of locals who are part-

time fighters (column 4). At the same time, they are less likely to have a VC

main squad, which consists of regular forces from elsewhere (columns 5). VC

62The impact on the share of households that contribute to charity is no longer statistically

significant when all of Ho Chi Minh Province is excluded, but the coefficient remains large and

positive.

4.5. MECHANISMS 185

supporters in Dai Viet villages appear better able to organize into a guerrilla

squad, but the VC compensated this by sending more regular forces to former

Khmer areas. Columns 6 and 7 do not find impacts on presence of the VC

Infrastructure, which organized VC political activities, or on VC taxation.

HES likewise contains information on whether friendly (U.S. and South

Vietnamese) forces operated nearby in the past month and whether friendly

air or artillery strikes hit near populated areas (columns 8 and 9). While the

coefficients are negative, neither is close to being statistically significant. We

can likewise examine security using administrative data from the U.S. and

South Vietnamese armed forces that track ground troop activity (“Situation

Report Army”, RG 218). There is not an impact on U.S. initiated attacks near

the hamlet (column 10), but South Vietnamese initiated attacks are lower in

Dai Viet villages (column 11). This is likely because South Vietnamese ground

troops pursued VC main force squads. Finally, we consider data on South

Vietnamese regional defense forces from the the “Territorial Forces Evaluation

System” (RG 472) and the “Territorial Forces Activity Reporting System” (RG

330) and again do not find a discontinuity (column 12).

These results are broadly robust to widening the bandwidth, with security

on the Dai Viet side tending to be better relative to the Khmer side as the

bandwidth is extended towards 100 kilometers (Appendix Figure 4.9). Results

are robust to controlling for population, dropping Ho Chi Minh City, and

dropping all provincial capitals (Tables A.29 to A.31).

Effects likewise do not appear to be driven by recent land inequality.

Recall from Section 4.2 that there were almost no French estates near the

1698 boundary - nearly every village is a zero in these data - but we also

examine agricultural and land outcomes more recently. The dependent variable

in column (1) of Table 4.12 is an indicator equal to 1 if the household is

engaged in agricultural production, taken from VHLSS. The estimates show

that Dai Viet households are less likely to work in agriculture, consistent with

the economic effects discussed earlier. Column (2) examines agricultural land

size, in hectares, for agricultural households. Though some caution is warranted

since this outcome is selected, the coefficient is small relative to the mean

and statistically insignificant. This indicates that differences in average farm

size are unlikely to drive the observed economic differences. In Column (3),


the dependent variable is an indicator equal to 1 if the individual works in a

manufacturing industry, again from VHLSS. We restrict analysis to prime-age

men, in order to avoid conflating effects with selection into the labor force. The

point estimate is small and statistically insignificant, suggesting that households

in Dai Viet areas are more likely to move out of agriculture into owner-operated

businesses and services. We also examined manufacturing in detail using the

2012 Enterprise Census and did not find major differences in the distribution of

employment across manufacturing sectors (results available upon request).

However, the VHLSS commune questionnaire does reveal that a lower

share of land is formally titled in Dai Viet villages. Columns (4) through (6)

estimate equation (4.1) using the fraction of area of different types of land

in each commune with a land use certificate as the dependent variable. The

estimates show a lower prevalence of land-use certificates in areas historically

under a strong state, for annual, perennial, and residential land. If property rights

are de facto secure for villagers due to strong communal enforcement, they may

demand fewer formal titles, or there may be social pressure to participate in

community norms rather than formal titling.

On a related note, column (7) examines whether the use of formal financial

services is more or less widespread in historically Dai Viet areas. Results are

again consistent with less active impersonal markets. Households in Dai Viet

villages are 10 percentage points less likely to make interest expenses on formal

financial instruments, despite being wealthier. This contrasts to Italy, where

Guiso, Sapienza, and Zingales (2004) find that in high-social-capital areas,

households are more likely to use checks, invest less in cash and more in stock,

have higher access to institutional credit, and make less use of informal credit.

Market institutions arrived in Vietnam recently, and may still be less effective

than non-market institutions in places where social capital is high, and hence

non-market arrangements work relatively well. Finally column (8) considers

informal sector employment, limiting to prime age males. In contrast to the

land titling and financial sector results, informal sector employment is lower in

Dai Viet villages. This is largely driven by the fact that there is less agriculture,

which makes up the bulk of the informal sector. These results are broadly robust

to the choice of bandwidth (Figure 4.10), with the impact on land size becoming

negative and significant when wider bandwidths are used. The results are also

4.6. CONCLUSION 187

robust to excluding Ho Chi Minh City and Province and provincial capitals

(Tables A.32 to A.34).

Vietnam has become increasingly globalized in recent years, and an ad-

ditional hypothesis is that Dai Viet villages are richer in part because they

have been better at attracting foreign investment. For example, a review by

Nielsen, Asmussen, and Weatherall (2017) suggests that places with higher

human capital are often better able to attract FDI. However, using data from the

2011 Enterprise Census, Table A.35 shows that foreign sector employment is ac-

tually lower in Dai Viet villages.63 This is consistent with historically tight-knit

villages being more closed towards outsiders, although other explanations could

also be at play. In any case, a greater prevalence of FDI in Dai Viet villages is

unlikely to drive the results.

4.6 Conclusion

Using a regression discontinuity design across the Dai Viet-Khmer bound-

ary, this study documents that areas historically under a strong state have higher

living standards today and better economic outcomes over the past 150 years.

Rich historical data reveal that in villages with a strong historical state, citizens

have been better able to organize for public goods and redistribution through

civil society and local government.

The strong historical state plausibly crowded in village-level collective

action, and these norms persisted long after the original state disappeared.

While care must be taken with external validity, this study provides support for

the theory that the existence of a strong historical state in East Asia - which

encouraged local collective action - played a central role in the 20th century

divergence between this region and much of the developing world.

This suggests that a collective action mechanism is also plausibly at play

in explaining the relative prosperity of Northeast versus Southeast Asia, another

central puzzle of modern economic growth.

63The Enterprise Census includes formal firms and identifies the location of the headquarters.

188 REFERENCES

References

Acemoglu, D., D. Cantoni, S. Johnson, and J. A. Robinson (2011). The conse-

quences of radical reform: The french revolution. American Economic

Review 101, 3286–3307.

Acemoglu, D., C. Garcia-Jimeno, and J. A. Robinson (2015). State capacity and

economic development: A network approach. The American Economic

Review 105(8), 2364–2409.

Acemoglu, D., S. Johnson, and J. A. Robinson (2001). The colonial origins

of comparative development: An empirical investigation. The American

Economic Review 91(5), 1369–1401.

Acemoglu, D., S. Johnson, and J. A. Robinson (2002). Reversal of fortune:

Geography and institutions in the making of the modern world income

distribution. The Quarterly Journal of Economics 117(4), 1231–1294.

Acemoglu, D., T. Reed, and J. A. Robinson (2014). Chiefs: Economic develop-

ment and elite control of civil society in sierra leone. Journal of Political

Economy 122(2), 319–368.

Acharya, A. (2013). Civilizations in Embrace: The Spread of Ideas and the

Transformation of Power: India and Southeast Asia in the Classical Age,

Volume 2. Singapore: Institute of Southeast Asian Studies.

Alesina, A., R. Baqir, and W. Easterly (1999). Public goods and ethnic divisions.

The Quarterly Journal of Economics 114(4), 1243–1284.

Alesina, A. and N. Fuchs-Schündeln (2007, September). Goodbye lenin (or

not?): The effect of communism on people. American Economic Re-

view 97(4), 1507–1528.

Alesina, A. and P. Giuliano (2015, December). Culture and institutions. Journal

of Economic Literature 53(4), 898–944.

Alesina, A., P. Giuliano, and N. Nunn (2013). On the origins of gender roles:

Women and the plough. The Quarterly Journal of Economics 128(2),

469–530.

Alesina, A. and E. La Ferrara (2005). Ethnic diversity and economic perfor-

mance. Journal of Economic Literature 43(3), 762–800.

Anh, N. T. (2003). Village Versus State: The Evolution of State-Local Relations

in Vietnam Until 1945. South East Asian Studies 41(1), 101–123.

REFERENCES 189

Barnes, G. L. (2007). State Formation in Japan: Emergence of a 4th-century

Ruling Elite. London, England and New York, New York: Routledge,

Taylor & Francis.

Bazzi, S. and M. Gudgeon (2016). Local government proliferation, diversity,

and conflict. Working paper, Boston University.

Becker, S. O., K. Boeckh, C. Hainz, and L. Woessmann (2016). The empire is

dead, long live the empire! long-run persistence of trust and corruption

in the bureaucracy. The Economic Journal 126(590), 40–74.

Besley, T. (1995). Nonmarket institutions for credit and risk sharing in low-

income countries. The Journal of Economic Perspectives 9(3), 115–127.

Booth, A. (2007). Nightwatchman, Extractive or Developmentail States? Some

Evidence from Late Colonial South-East Asia. Economic History Re-

view 60(2), 241–266.

Bredo, W. (1968). Land Reform in Vietnam. Stanford Research Institute.

Briggs, L. P. (1947). A Sketch of Cambodian History. The Far Eastern

Quarterly 6(4), 345.

Bukowski, P. (2015). What Determines the Long-Run Persistence of the Em-

pires? The Effect of the Partition of Poland on Education. Technical

report.

Chandler, D. P. (1983). A History of Cambodia David Chandler (2000 ed.).

Boulder, Colorado: Westview Press.

Chutintaranond, S. (1990). "Mandala", "Segmentary State" and Politics of

Centralisation in Medieval Ayudhya. Journal of the Siam Society 78(1),

89–100.

Cœ dès, G. (1966). The Making of South East Asia (1983 ed.). Berkeley and

Los Angeles, California and London, England: University of California

Press.

Cotter, M. G. (1968). Towards a Social History of the Vietnamese Southward

Movement. Journal of Southeast Asian History 9(1), 12–24.

Cotterell, A. (2014). A History of Southeast Asia. West Sussex, U.K.: Wiley

Blackwell.

Deaton, A. (1997). The Analysis of Household Surveys: A Microeconometric

Approach to Development Policy. World Bank.

Dell, M. (2010). The persistent effects of peru’s mining mita. Economet-

190 REFERENCES

rica 78(6), 1863–1903.

Dell, M. and P. Querubin (2016, July). Nation building through foreign inter-

vention: Evidence from discontinuities in military strategies. Working

Paper 22395, National Bureau of Economic Research.

Demsetz, H. (1967). Toward a theory of property rights. American Economic

Review 57(2), 347–359.

Dieu, N. T. (1999). The Mekong River and the Struggle for Indochina. Praeger.

Do, T. (2003). Vietnamese Supernaturalism. Views from the Southern Region.

Taylor & Francis.

Dùc, T. H. and N. Tao (1972). Gia Dinh Thành Thông Chí. Nha Ván hóa, Phù

QuóÃÇc-vu-khanh Dac-trách Van-hóa.

Easterly, W. and R. Levine (1997). Africa’s growth tragedy: Policies and ethnic

divisions. The Quarterly Journal of Economics 112(4), 1203–1250.

Ebihara, M. (1984). Social Organization in Sixteenth and Seventeenth Century

Cambodia. Journal of Southeast Asian Studies 15(2), 280–295.

Engelbert, T. (1994). The Khmer in Southern Vietnam-Cambodians or Viet-

namese? Nationalism and Ethnicity in Southeast Asia 1, 155–196.

Evans, P. B. (1995). Embedded Autonomy: States and Industrial Transformation.

Princeton University Press.

Fernández, R. and A. Fogli (2009, January). Culture: An empirical investigation

of beliefs, work, and fertility. American Economic Journal: Macroeco-

nomics 1(1), 146–177.

Foster, G. M. (1965). Peasant society and the image of limited good. American

Anthropologist 67(2), 293–315.

Fukuyama, F. (1995). Trust: The Social Virtues and the Creation of Prosperity.

Free Press.

Gelman, A. and G. Imbens (2014, August). Why high-order polynomials should

not be used in regression discontinuity designs. Working Paper 20405,

National Bureau of Economic Research.

Gennaioli, N. and I. Rainer (2012). The Modern in Africa Centralization Impact

of Precolonial. Journal of Economic Growth 12(3), 185–234.

Giuliano, P. (2007). Living arrangements in western europe: Does cultural

origin matter? Journal of the European Economic Association 5(5),

927–952.

REFERENCES 191

Gouldner, A. W. (1980). The Two Marxisms: Contradictions and Anomalies in

the Development of Theory. Critical Social Studies. Macmillan.

Greif, A. (1994). Cultural beliefs and the organization of society: A histori-

cal and theoretical reflection on collectivist and individualist societies.

Journal of Political Economy 102(5), 912–950.

Greif, A. and G. Tabellini (2010, May). Cultural and institutional bifurcation:

China and europe compared. American Economic Review 100(2), 135–

140.

Grosjean, P. (2011, May). The weight of history on european cultural integration:

A gravity approach. American Economic Review 101(3), 504–508.

Grosjean, P. (2014). A history of violence: The culture of honor and homicide

in the us south. Journal of the European Economic Association 12(5),

1285–1316.

Guiso, L., P. Sapienza, and L. Zingales (2004, June). The role of social capital

in financial development. American Economic Review 94(3), 526–556.

Guiso, L., P. Sapienza, and L. Zingales (2008). Alfred marshall lecture: Social

capital as good culture. Journal of the European Economic Associa-

tion 6(2-3), 295–320.

Guiso, L., P. Sapienza, and L. Zingales (2016). Long-term persistence. Journal

of the European Economic Association 14(6), 1401–1436.

Haines, D. (1984). Reflections of Kinship and Society Under Vietnam’s Le

Dynasty. Journal of Southeast Asian Studies 15(2), 307–314.

Hall, D. (1968). A History of South-East Asia. Macmillan and Company,

Limited and St. Martin’s Press.

Hall, D. (1973). The Integrity of Southeast Asian History. Journal of Southeast

Asian Studies 4(2), 159–168.

Hall, K. R. (2011). A History of Early Southeast Asia: Maritime Trade and

Social Development, 100-500. Rowman & Littlefield Publishers.

Harding, A. (2001). Comparative Law and Legal Transplantation in South East

Asia: Making Sense of the ‘Nomic Din’. In D. Nelken and J. Feest (Eds.),

Adapting legal cultures. Hart Publishing.

Hayami, Y. (1980). Economic approach to village community and institutions.

Journal of Rural Development 3, 27–49.

Hooker, M. B. (1978a). A Concise Legal History of South-East Asia. Clarendon

192 REFERENCES

Press and Oxford University Press.

Hooker, M. B. (1978b). The Indian-Derived Law Texts of Southeast Asia. The

Journal of Asian Studies 37(02), 201–219.

Hoover, D. R. (2000). The sources of social capital reconsidered: Voluntary

associations, advocacy and the state. In D. McConkey and P. Lawler

(Eds.), Social Structures, Social Capital, and Personal Freedom. Praeger.

Jansen, M. B. (2000). China in the Tokugawa World. Harvard University Press.

Kang, D. C. (2010). Hierarchy and Legitimacy in International Systems: The

Tribute System in Early Modern East Asia. Security Studies 19(4), 591–

622.

Kang, D. C. and V. D. Cha (2010). East Asia Before the West. Five Centuries of

Trade and Tribute. Columbia University Press.

Kerkvliet, B. J. and D. G. Marr (2004). Beyond Hanoi: Local Government in

Vietnam. Institute of Southeast Asian Studies / Nordic Institute of Asian

Studies.

Kulke, H. (1986). Early and the Imperial Kingdoms in Southeast Asian History.

In D. G. Marr and A. C. Milner (Eds.), Southeast Asia in the 9th to 14th

Century. Institute of Southeast Asian Studies and Australian National

University, Research School of Pacific Studies.

Lehning, P. B. (1998). Towards a multicultural civil society: The role of social

capital and democratic citizenship. Government and Opposition 33(2),

221–242.

Lewis, M. E. (2009). China’s Cosmopolitan Empire. Belknap Press and Harvard

University Press.

Lewis, M. E. (2011). China Between Empires: The Northern and Southern

Dynasties. Harvard University Press.

Li, T. (1998). Nguyen Cochinchina: Southern Vietnam in the Seventeenth and

Eighteenth Centuries. SEAP Publications.

Lieberman, V. (1987). Reinterpreting Burmese History. Comparative Studies

in Society and History 29(01), 162.

Lieberman, V. (1993). Local Integration and Eurasian Analogies: Structuring

Southeast Asian History, c. 1350-c. 1830. Modern Asian Studies 27(03),

475–572.


REFERENCES 193

800-1830, Volume I: Integration on the Mainland. Cambridge, U.K and


Liu, H. (2007). Influence of China’s Imperial Examinations on Japan, Korea

and Vietnam. Frontiers of History in China 2(4), 493–512.

Lowes, S., N. Nunn, J. A. Robinson, and J. Weigel (2015, December). The

evolution of culture and institutions: Evidence from the kuba kingdom.

Working Paper 21798, National Bureau of Economic Research.

Luttmer, E. F. P. and M. Singhal (2011, February). Culture, context, and the taste

for redistribution. American Economic Journal: Economic Policy 3(1),

157–79.

Mabbett, I. W. (1977a). The ‘Indianization’ of Southeast Asia : Reflections on

the Historical Sources. Journal of Southeast Asian Studies 8(2), 143–161.

Mabbett, I. W. (1977b). Varnas in angkor and the indian caste system. The

Journal of Asian Studies 36(03), 429–442.

Mabbett, I. W. and D. P. Chandler (1995). The Khmers. Oxford, U.K. and

Cambridge, Massachusetts: Blackwell Publishers.

Marr, D. G. (2004). A Brief History of Local Government in Vietnam. In B. J. T.

Kerkvliet and D. G. Marr (Eds.), Beyond Hanoi: Local Government in

Vietnam. Institute of Southeast Asian Studies / Nordic Institute of Asian

Studies.

Martinez-Bravo, M., G. Padró i Miquel, N. Qian, and Y. Yao (2012). Elections

in china. Working Paper 18101, National Bureau of Economic Research.

McHale, S. (2013). Ethnicity, Violence, and Khmer-Vietnamese Relations: The

Significance of the Lower Mekong Delta, 1757‚Äì1954. The Journal of

Asian Studies 72(2), 367–390.

Meyer, K. E. and H. V. Nguyen (2005). Foreign investment strategies and

sub-national institutions in emerging markets: Evidence from vietnam.

Journal of Management Studies 42(1), 63–93.

Michalopoulos, S. and E. Papaioannou (2013). Pre-Colonial Ethnic Institutions

and Contemporary African Development. Econometrica 81(1), 113–152.

Miguel, E. and G. Roland (2011). The long-run impact of bombing vietnam.

Journal of Development Economics 96(1), 1–15.

Montalvo, J. and M. Reynal-Querol (2005a). Ethnic diversity and economic

development. Journal of Development Economics 76(2), 293–323.

194 REFERENCES

Montalvo, J. and M. Reynal-Querol (2005b). Ethnic polarization, potential

conflict, and civil wars. American Economic Review 95(3), 796–816.

Mus, P. (1949). The Role of the Village in Vietnamese Politics. Pacific

Affairs 22(3), 265–272.

Nguyen, A. T. (1985). The Vietnamese Monarchy under French Colonial Rule

1884-1945. Modern Asian Studies 19(1), 147–162.

Nielsen, B. B., C. G. Asmussen, and C. D. Weatherall (2017). The location

choice of foreign direct investments: Empirical evidence and method-

ological challenges. Journal of World Business 52(1), 62 – 82.

Nunn, N. (2008). The long-term effects of africa’s slave trades. The Quarterly

Journal of Economics 123(1), 139–176.

Nunn, N. and L. Wantchekon (2011, December). The slave trade and the origins

of mistrust in africa. American Economic Review 101(7), 3221–3252.

Osborne, M. (1966). History and Kingship in Contemporary Cambodia. Journal

of Southeast Asian History 7(1), 1–14.

Osborne, M. (1969). The French presence in Cochinchina and Cambodia: Rule

and Response (1859-1905). Ithaca, New York and London, U.K.: Cornell

University Press.

Oto-Peralías, D. and D. Romero-Ávila (2014). The Distribution of Legal

Traditions Around the World: A Contribution to the Legal-Origins Theory.

Journal of Law and Economics 57(3), 561–628.

Padro i Miquel, G., N. Qian, Y. Xu, and Y. Yao (2015, April). Making democracy

work: Culture, social capital and elections in china. Working Paper 21058,


Palais, J. B. (1996). Confucian Statecraft and Korean Institutions: Yu Hyong-

won and the Late Choson Dynasty. Seattle, Washington: University of

Washington Press.

Parthesius, R. (2010). Dutch Ships in Tropical Waters. The Development of the

Dutch East India Company (VOC) Shipping Network in Asia 1595-1660.

Amsterdam studies in the Dutch golden age. Amsterdam University Press.

Pastor, J. (1997). Land and Economy in Ancient Palestine. Journal of Southeast

Asian Studies 1(02), 306.

Popkin, S. L. (1979). The Rational Peasant: The Political Economy of Rural

Society in Vietnam.

REFERENCES 195

Porter, G. (1993). Vietnam: The Politics of Bureaucratic Socialism. Cornell

University Press.

Putnam, R. D., R. Leonardi, and R. Y. Nanetti (1994). Making democracy work:

Civic traditions in modern Italy. Princeton University Press.

Qian, N. (2014). Village governance in china. In The Oxford Companion to the

Economics of China. Oxford University Press, USA.

Roland, G. (2012). The long-run weight of communism or the weight of

long-run history? In Economies in Transition, pp. 153–171. Springer.

Sahai, S. (1970). Les Institutions Politiques et l’Organisation Administrative du

Cambodge Ancien (VIe-XIIIe siècles), Volume 1. Paris: École française

d’Extréme-Orient Publications.

Satyanath, S., N. Voigtlaender, and H.-J. Voth (2013, July). Bowling for fascism:

Social capital and the rise of the nazi party. Working Paper 19201,


Scott, J. C. (1977). The Moral Economy of the Peasant: Rebellion and Subsis-

tence in Southeast Asia. Yale University Press.

Skocpol, T. (1995). Protecting Soldiers and Mothers. Harvard University Press.

Spolaore, E. and R. Wacziarg (2013, June). How deep are the roots of economic

development? Journal of Economic Literature 51(2), 325–369.

Su Quan Trieu Nguyen, Q. and T. D. Pham (1992). Dai Nam Nhat Thong Chi.

Nha Xuat Ban Thuan Hoa.

Tabellini, G. (2008). Presidential address: Institutions and culture. Journal of

the European Economic Association 6(2-3), 255–294.

Tabellini, G. (2010). Culture and institutions: Economic development in the

regions of europe. Journal of the European Economic Association 8(4),

677–716.

Tambiah, S. J. (1977). The Galactic Polity: The Structure of Traditional

Kingdoms in Southeast Asia. Annals of the New York Academy of Sci-

ences 293(1 Anthropology), 69–97.

Tambiah, S. J. (2013). The Galactic Polity in Southeast Asia. HAU: Journal of

Ethnographic Theory 3(3), 503–534.

Taylor, K. (2013). A History of the Vietnamese. Cambridge University Press.

Vickery, M. (1996). Mak Phœun : Histoire du Cambodge de la fin du XVIe

au début du XVIIIe siècle. Bulletin de l’Ecole française d’Extrême-

196 REFERENCES

Orient 83, 405–415.

Voigtländer, N. and H.-J. Voth (2012). Persecution perpetuated: The medieval

origins of anti-semitic violence in nazi germany. The Quarterly Journal

of Economics 127(3), 1339–1392.

Wade, R. H. (1994). Village Republics: Economic Conditions for Collective

Action in South India. International Center for Self-Governance. San

Francisco: Institute for Contemporary Studies.

Whitmore, J. K. (1979). Merit Subjects: China, Korea, Vietnam. Ming Stud-

ies 1979(1), 42–51.

Wolf, E. R. (1969). Peasant Wars of the Twentieth Century. University of

Oklahoma Press.

Wolters, O. (1999). History, Culture and Region in Southeast Asian Perspectives.

Southeast Asia Program Publications, Cornell University.

Woodside, A. B. (1971). Vietnam and the Chinese model a Comparative

Study of Vietnamese and Chinese Government in the First Half of the

Nineteenth Century (1988 ed.). Cambridge, Massachusetts and London,

England: Council on East Asian Studies, Harvard University and Harvard

University Press.

Woodside, A. B. (1984). Medieval Vietnam and Cambodia: A Comparative

Comment. Journal of Southeast Asian Studies 15(02), 315–319.

Woodside, A. B. (2006). Lost Modernities : China , Vietnam , Korea and the

Hazards of World. Harvard University Press.

Wook, C. B. (2004). Southern Vietnam Under the Reign of Minh Mang (1820-

1841): Central Policies and Local Response. SEAP Publications.

Woolcock, M. and D. Narayan (2000). Social capital: Implications for develop-

ment theory, research, and policy. World Bank Research Observer 15(2),

225–49.

Yu, I. (2001, June). The Changing Nature of the Red River Delta Villages During

the Lê Period (1428-1788). Journal of Southeast Asian Studies 32(2),

151–172.

Zottoli, B. A. (2011). Reconceptualizing Southern Vietnamese History Form the

15th to 18th Centuries: Competition Along the Coasts Form Guangdong

to Cambodia. Ph.D Dissertation, The University of Michigan.

REFERENCES 197

Figure 4.1: Dai Viet Historical Boundaries

0 200 400100 Kilometers

U1:7,000,000

Historical BoundariesConquest Date

1069Capture of BoChinh, Dia Ly,Ma Linh

1306Huyen Tranmarriage

1407Ming boundary

1471Annexation ofQuang Nam

1611 Conquest ofPhu Yen

1651Defeat of PoNraup

1693Fall of Champa

1698Establishmentof Gia Dinh

1833Org. underMinh-Mang

Sources: Dùc and Tao (1972); Su Quan Trieu Nguyen and Pham (1992).

198 REFERENCES

Figure 4.2: Robustness of Household Consumption EstimatesFull Border No Rivers

● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

● ● ● ● ●●

●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

●●

● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

● ● ● ● ●●

● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●●

●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

● ● ● ●●

●●

● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●●

●●

● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

● ● ● ● ●●

● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●●

●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

0.0

0.2

0.4

0.6

0.0

0.2

0.4

0.6

0.0

0.2

0.4

0.6

0.0

0.2

0.4

0.6

0.0

0.2

0.4

0.6

0.0

0.2

0.4

0.6

Lin. Lat−LonLin. D

ist. to Bnd.

Lin. Dist. to B

nd., Lat−LonQ

uad. Lat−LonQ

uad. Dist. to B

nd.Q

uad. Dist. to B

nd., Lat−Lon

10 20 30 40 50 60 70 80 90 100 10 20 30 40 50 60 70 80 90 100bw

Equ

ival

ent C

onsu

mpt

ion

Notes: Each sub-figure plots the point estimates of γ (vertical axis) from equation (4.1) for

different bandwidth values between 10-100 kilometers in 1 km increments (horizontal axis).

Thin lines stemming from the point estimates show 95% confidence intervals while the slightly

thicker lines show 90% confidence intervals. The panels in different rows correspond to different

polynomial functions for geographic location. The estimates in the first column are based on the

full border while those in the second column exclude households closest to boundary segments

that coincide with a river.

200 REFERENCES

Table 4.1: Comparing Dai Viet and Khmer Kingdoms in Precolonial Vietnam

Dai Viet Khmer

Colonial outpost of China (111 BCE-

939 CE)

Indic patron-client statea

Maintained bureaucratic Chinese gov-

ernment system since independencebAccelerated decline after invasion by

Siam (1430); weak control of periph-

eryc

Centralized state; impersonal central-

ized bureaucracy under dynastic court;

uniform territorial administrationd

Decentralized state; personalistic rule

through court; semi-independent provin-

cial rulee

Institutionalized role of village chiefs &

village councils (elected since 1461)fPersonalistic political appointments &

land distributiong

Bureaucratic control of local taxation,

military recruitmenthTemple-based public finance systemi

aLieberman, 2003bWoodside, 1971cCoedes, 1966; Tarling, 1999dLieberman, 2003eWoodside, 1971; Tarling, 1999fYu, 2001gOsborne, 1969; Chandler, 1983hWoodside, 1971; Yu, 2001iTarling, 1999; Hall, 2010

REFERENCES 201

Table 4.2: Balance Checks

Dependent variable is:

Elev. Slope Temp. Precip. Rice Suit. Flow Accum. Km Rivers

(1) (2) (3) (4) (5) (6) (7)

Dai Viet -1.562 0.356 0.052 2.084 0.371 -1.758 -1.195

(10.027) (0.596) (0.053) (1.206) (0.611) (1.483) (2.666)

Obs 120 120 120 120 119 120 120

Clusters 120 120 120 120 119 120 120

Mean 51.39 2.59 26.65 168.34 27.28 0.66 17.01

The unit of analysis is the grid cell. All regressions include a linear RD polynomial in latitude

and longitude, a control for distance to Ho Chi Minh City, and boundary segment FE. Robust

standard errors are reported in parentheses.

202 REFERENCES

Tabl

e4.

3:C

on

tem

po

rary

Ho

use

ho

ldC

on

sum

pti

on

Dep

end

ent

var

iab

leis

log

ho

use

ho

ldex

pen

dit

ure

.S

pec

ifica

tio

nis

:

Dis

t.L

at-L

on

No

No

No

Co

nsi

st.

Tri

m2

5to

Lat

-Lo

nB

nd

.&

Dis

t.U

rban

HC

MP

rov.

No

On

lyP

rov.

Fo

r1

00

All

Po

lyn

om

ial

HC

MC

Pro

v.C

ap.

Riv

erB

nd

.F

EM

igr.

Km

SV

N

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

(9)

(10

)(1

1)

(12

)

Dai

Vie

t0

.33

10

.25

90

.28

00

.34

40

.32

20

.32

90

.31

10

.34

00

.40

50

.28

10

.32

90

.35

1

(0.0

54

)(0

.05

9)

(0.0

59

)(0

.05

5)

(0.0

76

)(0

.05

9)

(0.0

65

)(0

.08

4)

(0.0

63

)(0

.05

0)

(0.0

85

)(0

.02

6)

Ob

s4

,31

94

,31

94

,31

93

,48

32

,56

52

,86

63

,59

77

22

4,3

19

4,2

40

6,7

89

25

,61

7

Clu

ster

s4

50

45

04

50

36

22

58

31

23

74

76

45

04

50

67

02

58

1

The

unit

of

anal

ysi

sis

the

house

hold

.C

olu

mns

(1)

and

(3)

thro

ugh

(11)

incl

ude

ali

nea

rpoly

nom

ial

inla

titu

de

and

longit

ude,

and

colu

mns

(2)

and

(3)

incl

ud

ea

lin

ear

po

lyn

om

ial

ind

ista

nce

toth

eb

ou

nd

ary.

All

colu

mn

sin

clu

de

aco

ntr

ol

for

dis

tan

ceto

Ho

Ch

iM

inh

Cit

y,d

emo

-

gra

ph

icco

ntr

ols

for

the

nu

mb

ero

fin

fan

ts,

chil

dre

n,

and

adu

lts

inth

eh

ou

seh

old

,an

dy

ear

fixed

effe

cts.

Co

lum

ns

(1)

thro

ug

h(1

1)

incl

ud

e

bo

un

dar

yse

gm

ent

fixed

effe

cts,

and

colu

mn

(9)

incl

ud

esco

nsi

sten

tp

rov

ince

fixed

effe

cts.

Ro

bu

stst

and

ard

erro

rs,cl

ust

ered

atth

ev

illa

ge

level

,ar

ere

po

rted

inp

aren

thes

es.

Tabl

e4.

4:E

con

om

icO

utc

om

es:

Co

lon

ial

Per

iod

Dep

end

ent

var

iab

leis

:

Tel

egra

ph

Mil

itar

yR

ail/

Ro

adM

oto

rR

oad

Rai

lP

aved

Rai

l

Den

sity

Po

stD

ensi

tyD

ensi

tyD

ensi

tyR

oad

Den

s.D

ensi

ty

18

78

19

01

18

78

18

78

19

10

19

10

19

26

19

26

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

Dai

Vie

t0

.01

30

.06

1-0

.05

5-0

.01

10

.05

80

.03

60

.12

10

.00

7

(0.0

06

)(0

.03

2)

(0.0

25

)(0

.03

1)

(0.0

37

)(0

.02

1)

(0.0

43

)(0

.01

6)

Ob

s6

91

69

16

91

69

16

91

69

16

91

69

1

Clu

ster

s6

91

69

16

91

69

16

91

69

16

91

69

1

Mea

n0

.01

0.1

70

.03

0.1

90

.42

0.0

90

.43

0.0

6

The

unit

of

anal

ysi

sis

the

vil

lage.

All

regre

ssio

ns

incl

ude

ali

nea

rR

Dpoly

nom

ial

inla

titu

de

and

longit

ude,

a

con

tro

lfo

rd

ista

nce

toH

oC

hi

Min

hC

ity,

and

bo

un

dar

yse

gm

ent

fixed

effe

cts.

Ro

bu

stst

and

ard

erro

rsar

e

rep

ort

edin

par

enth

eses

.

204 REFERENCES

Tabl

e4.

5:E

con

om

icO

utc

om

es:

So

uth

Vie

tnam

ese

Per

iod

Dep

end

ent

var

iab

leis

:

Lo

gN

on

-ric

eM

anu

f.S

urp

lus

Ho

use

ho

lds

%H

HL

and

Fam

ily

Eco

nF

oo

dG

oo

ds

Go

od

sR

equ

ire

Acc

ess

Un

farm

edP

op

Inco

me

LC

AA

vai

l.A

vai

l.P

rod

uce

dA

ssis

t.V

ehic

.B

adS

ec.

Gro

wth

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

(9)

Dai

Vie

t0

.15

60

.16

30

.27

90

.19

60

.18

4-0

.07

50

.14

30

.00

9-0

.00

4

(0.0

41

)(0

.05

5)

(0.0

60

)(0

.06

5)

(0.0

52

)(0

.03

8)

(0.0

20

)(0

.04

7)

(0.0

06

)

Ob

s5

,92

62

,28

53

88

38

83

88

2,3

30

2,3

32

33

02

,27

6

Clu

ster

s1

72

39

23

88

38

83

88

39

73

96

33

03

96

Mea

n9

.72

0.8

20

.71

0.6

30

.44

0.6

10

.34

0.2

60

.01

Th

eu

nit

of

anal

ysi

sis

the

ho

use

ho

ld,h

amle

t,o

rv

illa

ge.

All

reg

ress

ion

sin

clu

de

ali

nea

rR

Dp

oly

no

mia

lin

lati

tud

e

and

longit

ude,

aco

ntr

ol

for

dis

tance

toH

oC

hi

Min

hC

ity,

and

boundar

yse

gm

ent

fixed

effe

cts.

Robu

stst

andar

der

rors

,

clu

ster

edat

the

vil

lag

ele

vel

,ar

ere

po

rted

inp

aren

thes

es.

REFERENCES 205

Tabl

e4.

6:C

ivil

So

ciet

y

Dep

end

ent

var

iab

leis

:

Civ

il%

Ho

use

ho

lds

%H

HS

elf-

Dev

.C

ou

nci

lO

rg.

%H

HR

D%

HH

Ho

use

ho

lds

Civ

.S

oc.

So

ciet

yP

arti

cpat

ein

Act

ive

Pro

ject

Dis

cuss

esY

ou

thA

tten

dC

adre

inP

art.

Req

uir

eP

rov

ides

LC

AC

ivic

Org

Eco

n.

Tra

in.

inP

SD

FU

nd

erw

ayG

riev

.A

ctiv

.G

ov

t.M

tgs.

Ham

let

RD

Cad

reA

ssis

tan

ce

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

(9)

(10

)(1

1)

(12

)

Dai

Vie

t0

.17

50

.26

20

.21

10

.07

70

.09

9-0

.01

1-0

.03

60

.10

40

.00

90

.17

4-0

.07

50

.16

5

(0.0

35

)(0

.02

7)

(0.0

27

)(0

.02

8)

(0.0

24

)(0

.02

0)

(0.0

33

)(0

.02

8)

(0.0

30

)(0

.03

5)

(0.0

38

)(0

.04

2)

Ob

s2

,28

52

,32

52

,34

82

,33

03

88

38

43

88

2,3

31

2,3

37

2,3

14

2,3

30

2,2

06

Clu

ster

s3

92

39

73

99

39

73

88

38

43

88

39

73

97

39

63

97

38

8

Mea

n0

.76

0.3

70

.22

0.6

20

.83

0.9

30

.78

0.3

70

.76

0.5

20

.61

0.2

4

Th

eu

nit

of

anal

ysi

sis

the

ham

let

or

vil

lag

e.A

llre

gre

ssio

ns

incl

ud

ea

lin

ear

RD

po

lyn

om

ial

inla

titu

de

and

lon

git

ud

e,a

con

tro

lfo

rd

ista

nce

toH

oC

hi

Min

hC

ity,

and

bo

un

dar

y

seg

men

tfi

xed

effe

cts.

Ro

bu

stst

and

ard

erro

rs,

clu

ster

edat

the

vil

lag

ele

vel

,ar

ere

po

rted

inp

aren

thes

es.

206 REFERENCES

Tabl

e4.

7:L

oca

lA

dm

inis

trat

ion

Dep

end

ent

var

iab

leis

:

Lo

cal

Gov

t.V

ilg

.V

ilg

.H

amle

tP

oli

ceC

hie

fG

ov

t.L

ack

Tec

h.

Pro

v.L

and

Ad

min

.S

yst

.C

om

m.

Ch

ief

Ch

ief

Reg

ula

rly

Co

ntr

ols

Pro

vid

esP

rov.

Tec

h.

Per

s.A

ffai

rs

LC

AT

axes

Fil

led

Pre

sen

tR

DC

adre

Ass

ist.

Ass

ist.

Vis

itV

isit

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

(9)

(10

)(1

1)

Dai

Vie

t0

.03

90

.06

00

.06

20

.05

70

.03

60

.16

60

.08

10

.14

10

.01

20

.00

30

.00

1

(0.0

17

)(0

.03

5)

(0.0

28

)(0

.03

3)

(0.0

24

)(0

.04

8)

(0.0

20

)(0

.04

8)

(0.0

30

)(0

.04

8)

(0.0

55

)

Ob

s2

,28

53

88

38

83

88

2,3

17

2,3

39

38

22

,22

13

87

38

63

08

Clu

ster

s3

92

38

83

88

38

83

96

39

73

82

39

03

87

38

63

08

Mea

n0

.98

0.8

40

.87

0.9

30

.92

0.5

60

.88

0.3

00

.18

0.5

30

.72

Th

eu

nit

of

anal

ysi

sis

the

ham

let

or

vil

lage.

All

reg

ress

ions

incl

ud

ea

lin

ear

RD

poly

nom

ial

inla

titu

de

and

lon

git

ud

e,a

con

tro

lfo

rd

ista

nce

toH

o

Ch

iM

inh

Cit

y,an

db

ou

nd

ary

seg

men

tfi

xed

effe

cts.

Ro

bu

stst

and

ard

erro

rs,

clu

ster

edat

the

vil

lag

ele

vel

,ar

ere

po

rted

inp

aren

thes

es.

REFERENCES 207

Tabl

e4.

8:P

ub

lic

Go

od

s

Dep

end

ent

var

iab

leis

:

Hea

lth

Gov

t.H

ealt

hH

ealt

hM

at.

Pri

mar

yS

eco

nd

ary

Att

end

.L

aw

Car

eM

ed.

Ser

v.W

krs

.V

isit

Cli

nic

Ed

uc.

Sch

oo

lS

cho

ol

Res

tr.

En

forc

ed

LC

AA

vai

l.R

eg.

inV

illa

ge

LC

AA

cces

s.C

om

ple

tio

nIn

Vil

g.

Att

end

.b

yS

ec.

Day

/Nig

ht

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

(9)

(10

)(1

1)

(12

)

Dai

Vie

t0

.14

00

.20

40

.33

90

.14

00

.02

20

.05

20

.05

70

.08

90

.08

50

.03

0-0

.01

80

.21

5

(0.0

43

)(0

.03

7)

(0.0

42

)(0

.05

0)

(0.0

67

)(0

.04

4)

(0.0

22

)(0

.03

1)

(0.0

59

)(0

.01

3)

(0.0

13

)(0

.04

6)

Ob

s2

,28

52

,33

92

,33

63

88

38

82

,28

52

,33

63

88

38

83

88

2,3

33

2,3

33

Clu

ster

s3

92

39

73

97

38

83

88

39

23

96

38

83

88

38

83

96

39

7

Mea

n0

.86

0.3

90

.47

0.7

90

.61

0.8

20

.90

0.6

10

.35

0.1

80

.02

0.7

9

Th

eu

nit

of

anal

ysi

sis

the

ham

let

or

vil

lag

e.A

llre

gre

ssio

ns

incl

ud

ea

lin

ear

RD

po

lyn

om

ial

inla

titu

de

and

lon

git

ud

e,a

con

tro

lfo

rd

ista

nce

toH

oC

hi

Min

h

Cit

y,an

db

ou

nd

ary

seg

men

tfi

xed

effe

cts.

Ro

bu

stst

and

ard

erro

rs,

clu

ster

edat

the

vil

lag

ele

vel

,ar

ere

po

rted

inp

aren

thes

es.

208 REFERENCES

Tabl

e4.

9:P

ub

lic

Op

inio

n

Dep

end

ent

var

iab

leis

:

Gv

t.L

oca

lK

now

sV

ilg

LT

TN

atl.

Gv

t.P

eop

leA

ctiv

ein

Peo

ple

Res

po

nsi

ve

Offi

cial

sA

dm

in.

Fai

rly

Per

form

sM

an.

Eco

n.

Res

po

ns.

Inte

rest

Dec

ide

Cit

izen

sS

ucc

essf

ul

Str

uct

.W

ell

Ad

min

ist.

Po

orl

yP

oo

rly

Co

mm

.L

ife

Gro

up

SD

P

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

(9)

Dai

Vie

t0

.09

80

.20

20

.11

90

.35

00

.11

00

.08

70

.27

10

.39

50

.22

2

(0.0

44

)(0

.07

5)

(0.0

44

)(0

.07

5)

(0.0

40

)(0

.04

0)

(0.0

75

)(0

.14

6)

(0.0

52

)

Ob

s2

,77

93

,48

71

,45

79

99

2,8

11

5,7

78

87

92

43

35

3

Clu

ster

s1

90

18

38

91

01

18

22

15

10

63

55

3

Mea

n0

.37

0.5

20

.22

0.5

70

.19

0.3

10

.18

0.1

80

.23

The

unit

of

anal

ysi

sis

the

indiv

idual

.A

llre

gre

ssio

ns

incl

ude

ali

nea

rR

Dpoly

nom

ial

inla

titu

de

and

longit

ude,

aco

ntr

ol

for

dis

tance

to

Ho

Chi

Min

hC

ity,

and

boundar

yse

gm

ent

fixed

effe

cts.

Robust

stan

dar

der

rors

,cl

ust

ered

atth

evil

lage

level

,ar

ere

port

edin

par

enth

eses

.

REFERENCES 209

Table 4.10: Current Outcomes


Contributed Share Years Schooling

to Charity Communes Cohort

Fund Lower Sec. > 25 25-40 40-60 >60

(1) (2) (3) (4) (5) (6)

Dai Viet 0.122 0.292 0.950 0.899 0.989 0.982

(0.035) (0.069) (0.195) (0.192) (0.237) (0.232)

Obs 5,889 124 42,189 17,985 17,273 6,931

Clusters 450 124 453 452 453 442

Mean 0.70 0.78 7.45 8.41 7.67 4.38

The unit of analysis is the household, district, or individual. All columns include a

linear RD polynomial in latitude and longitude, a control for distance to Ho Chi

Minh City, and boundary segment fixed effects. Columns (1) and (3) through (6)

include year fixed effects. Robust standard errors, clustered at the village level, are

reported in parentheses.

210 REFERENCES

Tabl

e4.

11:A

dd

itio

nal

Mec

han

ism

s-

Th

eV

ietn

amW

ar

Dep

end

ent

var

iab

leis

:

VC

VC

Vil

g.

VC

VC

Fri

end

lyA

ir/A

rt.

U..

S.

SV

NT

erri

t.

Sec

uri

tyF

orc

esB

ase

Gu

err.

Mai

nIn

fra

VC

Fo

rces

Str

ke

Init

iate

dF

orc

es

LC

AP

rese

nt

Nea

rby

Sq

uad

Sq

uad

Act

ivit

yT

axat

ion

Nea

rby

Nea

rby

Att

ack

Pre

sen

t

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

(9)

(10

)(1

1)

(12

)

Dai

Vie

t0

.00

3-0

.04

9-0

.03

50

.06

5-0

.10

20

.02

3-0

.01

4-0

.02

4-0

.04

0-0

.00

0-0

.10

9-0

.01

5

(0.0

43

)(0

.03

5)

(0.0

52

)(0

.03

8)

(0.0

39

)(0

.03

2)

(0.0

16

)(0

.03

7)

(0.0

31

)(0

.00

1)

(0.0

24

)(0

.02

6)

Ob

s2

,28

52

,33

53

90

39

03

90

2,3

39

38

93

89

38

82

,35

82

,35

82

,34

8

Clu

ster

s3

92

39

83

90

39

03

90

39

83

89

38

93

88

40

04

00

39

9

Mea

n0

.80

0.1

50

.49

0.2

00

.23

0.0

90

.07

0.4

90

.13

0.0

00

.71

0.2

4

The

unit

of

anal

ysi

sis

the

ham

let

or

vil

lage.

All

regre

ssio

ns

incl

ude

ali

nea

rR

Dpoly

nom

ial

inla

titu

de

and

longit

ude,

aco

ntr

ol

for

dis

tance

to

Ho

Ch

iM

inh

Cit

y,an

db

ou

nd

ary

seg

men

tfi

xed

effe

cts.

Ro

bu

stst

and

ard

erro

rs,

clu

ster

edat

the

vil

lag

ele

vel

,ar

ere

po

rted

inp

aren

thes

es.

REFERENCES 211

Table 4.12: Additional Mechanisms - Land and MarketsDependent variable is:

Agric. Main Share H.H. Employed

Agric. Land Job in Annual Perennial Residential Interest Informal

H.H. Size Industry Land Certified Expenses Sector

(1) (2) (3) (4) (5) (6) (7) (8)

Dai Viet -0.190 -0.020 -0.013 -0.118 -0.125 -0.218 -0.090 -0.069

(0.034) (0.103) (0.022) (0.041) (0.049) (0.064) (0.032) (0.021)

Obs 16,419 4,518 20,357 176 173 170 4,553 20,343

Clusters 453 285 453 131 129 128 251 453

Mean 0.24 0.87 0.25 0.93 0.92 0.94 0.25 0.62

The unit of analysis is the household, individual, or commune. All columns include a linear RD polynomial in

latitude and longitude, a control for distance to Ho Chi Minh City, year fixed effects, and boundary segment fixed

effects. Robust standard errors, clustered at the village level, are reported in parentheses.

212 REFERENCES

A Appendix

Figure 4.1: Placebo: River as Boundary

Rivers

Placebo boundary

Treatment Boundary

Placebo districts

A. APPENDIX 213

Figure 4.2: Alternative Bandwidths: Economic Outcomes During the Colonial

Period

● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

● ●●

●●

●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

Telegraph Density 1878 Telegraph Density 1901

Military Post 1878 Rail or Road Density 1878

Motor Road Density 1910 Rail Density 1910

Paved Road Density 1926 Rail Density 1926

−0.1

0.0

0.1

0.2

−0.1

0.0

0.1

0.2

−0.1

0.0

0.1

0.2

−0.1

0.0

0.1

0.2

10 20 30 40 50 60 70 80 90 100 10 20 30 40 50 60 70 80 90 100bw

Out

com

e

Notes: Each sub-figure plots the point estimates and confidence intervals of γ (vertical axis)

from equation (1) for different bandwidth values between 10-100 kilometers in 1 km increments

(horizontal axis).

214 REFERENCES

Figure 4.3: Alternative Bandwidths: Economic Outcomes (1969-1973)

● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

Log Family Income Econ. LCA

Non−rice Food Avail. Manuf. Goods Avail.

Surplus Goods Produced Households Require Assist.

% HH Access Vehic. Land Unfarmed Bad Sec.

Population Growth

−0.2

0.0

0.2

0.4

−0.2

0.0

0.2

0.4

−0.2

0.0

0.2

0.4

−0.2

0.0

0.2

0.4

−0.2

0.0

0.2

0.4

10 20 30 40 50 60 70 80 90 100bw

Out

com

e



(horizontal axis).

A. APPENDIX 215

Figure 4.4: Alternative Bandwidths: Civil Society

● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

Civic Society LCA % Households Participate in Civic Org.

% Households Participate in Econ. Train. % Households Active in PSDF

Self. Dev. Project Underway Council Discusses Grievances

Organized Youth Activities % Households Attend Govt. Meetings

RD Cadre in Hamlet % Households Part. in RD Cadre Activites

Civic Society Provides Assistance

−0.1

0.0

0.1

0.2

0.3

−0.1

0.0

0.1

0.2

0.3

−0.1

0.0

0.1

0.2

0.3

−0.1

0.0

0.1

0.2

0.3

−0.1

0.0

0.1

0.2

0.3

−0.1

0.0

0.1

0.2

0.3

10 20 30 40 50 60 70 80 90 100bw

Out

com

e



(horizontal axis).

216 REFERENCES

Figure 4.5: Alternative Bandwidths: Local Administration

● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

Local Admin. LCA Govt. Syst. Taxes

Vilg. Comm. Filled Vilg. Chief Present

Hamlet Chief Present Police Regularly Present

Chief Controls RD Cadre Govt. Provides Welfare Assist.

Lack Prov. Tech. Assist. Tech. Pers. Visit

Prov. Land Affairs Visit

−0.1

0.0

0.1

0.2

0.3

0.4

−0.1

0.0

0.1

0.2

0.3

0.4

−0.1

0.0

0.1

0.2

0.3

0.4

−0.1

0.0

0.1

0.2

0.3

0.4

−0.1

0.0

0.1

0.2

0.3

0.4

−0.1

0.0

0.1

0.2

0.3

0.4

10 20 30 40 50 60 70 80 90 100bw

Out

com

e



(horizontal axis).

A. APPENDIX 217

Figure 4.6: Alternative Bandwidths: Public Goods

● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

Health Care LCA Govt. Med. Serv. Available

Health Workers Visit Regularly Health Clinc in Village

Maternity Clinic in Village Education LCA

Primary School Accessible Primary School Completion

Secondary School in Village Secondary School Attendance

School Attend. Restricted by Security Law Enforced Day and Night

−0.2

0.0

0.2

0.4

−0.2

0.0

0.2

0.4

−0.2

0.0

0.2

0.4

−0.2

0.0

0.2

0.4

−0.2

0.0

0.2

0.4

−0.2

0.0

0.2

0.4

10 20 30 40 50 60 70 80 90 100 10 20 30 40 50 60 70 80 90 100bw

Out

com

e



(horizontal axis).

218 REFERENCES

Figure 4.7: Alternative Bandwidths: Public Opinion

● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

● ● ●

●

●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

Gvt. Responsive Citizens Local Officials Successful

Know Vilg. Admin. Struct. LTT Fairly Administ.

Natl. Gvt. Performs Poorly Natl. Gvt. Man. Econ. Poorly

People Respons. Comm. Life Active in Interest Group

People Decide SDP

0.0

0.5

1.0

0.0

0.5

1.0

0.0

0.5

1.0

0.0

0.5

1.0

0.0

0.5

1.0

10 20 30 40 50 60 70 80 90 100bw

Out

com

e



(horizontal axis).

A. APPENDIX 219

Figure 4.8: Alternative Bandwidths: Modern Outcomes

● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●●

●●

●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●●

●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●●

●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●●

●●

●●

●●

●●

● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

Contributed to Charity Fund Years of Schooling (25 or Older)

Years of Schooling (25−40) Years of Schooling (40−60)

Years of Schooling (60 or Older)

0.0

0.5

1.0

1.5

0.0

0.5

1.0

1.5

0.0

0.5

1.0

1.5

10 20 30 40 50 60 70 80 90 100bw

Out

com

e



(horizontal axis).

220 REFERENCES

Figure 4.9: Alternative Bandwidths: Security

● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

Security LCA VC Forces Present

VC Base Nearby VC Village Guerrilla Squad

VC Main Force Squad VC Infrastructure Activity

VC Taxation Friendly Forces Nearby

Friendly Air/Art. Strike Nearby U.S. Initiated Attack

SVN Initiated Attack Territorial Forces Present

−0.2

−0.1

0.0

0.1

0.2

−0.2

−0.1

0.0

0.1

0.2

−0.2

−0.1

0.0

0.1

0.2

−0.2

−0.1

0.0

0.1

0.2

−0.2

−0.1

0.0

0.1

0.2

−0.2

−0.1

0.0

0.1

0.2

10 20 30 40 50 60 70 80 90 100 10 20 30 40 50 60 70 80 90 100bw

Out

com

e



(horizontal axis).

A. APPENDIX 221

Figure 4.10: Alternative Bandwidths: Other Mechanisms

● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

●●

●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

Agricultural Households Agricultural Land Size

Main Job in Industry Share Annual Land Titled

Share Perennial Land Titled Share Residential Land Titled

Interest Expenditure Employed in Informal Sector

−0.4

−0.2

0.0

−0.4

−0.2

0.0

−0.4

−0.2

0.0

−0.4

−0.2

0.0

10 20 30 40 50 60 70 80 90 100 10 20 30 40 50 60 70 80 90 100bw

Out

com

e



(horizontal axis).

Figure 4.11: Correlation Plots

.4.6

.81

econ

omic

s

0 .2 .4 .6 .8 1civil society

(a) Economic Index and CivilSociety Index

.5.6

.7.8

.91

loca

l adm

inis

trat

ion


(b) Local Admin Index andCivil Society Index

.2.4

.6.8

1ed

ucat

ion


(c) Education Index and CivilSociety Index

.2.4

.6.8

1he

alth


(d) Health Care Index andCivil Society Index

Notes: Each point is an outcome averaged within a bin. The regression line is fit on the raw data.

A. APPENDIX 223

A.1 Appendix Tables

224 REFERENCES

Tabl

eA

.1:C

on

tem

po

rary

Ho

use

ho

ldC

on

sum

pti

on

Incl

ud

ing

Pan

elH

ou

seh

old

s

Dep

end

ent

var

iab

leis

log

ho

use

ho

ldex

pen

dit

ure

.S

pec

ifica

tio

nis

:

Dis

t.L

at-L

on

No

No

No

Co

nsi

st.

Tri

m2

5to

Lat

-Lo

nB

nd

.&

Dis

t.U

rban

HC

MP

rov.

No

On

lyP

rov.

Fo

r1

00

All

Po

lyn

om

ial

HC

MC

Pro

v.C

ap.

Riv

erB

nd

.F

EM

igr.

Km

SV

N

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

(9)

(10

)(1

1)

(12

)

Dai

Vie

t0

.29

90

.23

00

.24

90

.30

20

.28

80

.29

10

.29

40

.28

10

.38

90

.25

20

.30

90

.34

8

(0.0

50

)(0

.05

5)

(0.0

55

)(0

.05

0)

(0.0

71

)(0

.05

3)

(0.0

60

)(0

.07

0)

(0.0

60

)(0

.04

7)

(0.0

80

)(0

.02

3)

Ob

s5

,53

95

,53

95

,53

94

,46

23

,29

63

,69

04

,61

29

27

5,5

39

5,4

46

8,7

34

32

,84

8

Clu

ster

s4

55

45

54

55

36

72

63

31

53

79

76

45

54

55

68

92

68

6

The

unit

of

anal

ysi

sis

the

house

hold

.C

olu

mns

(1)

and

(3)

thro

ugh

(11)

incl

ude

ali

nea

rpoly

nom

ial

inla

titu

de

and

longit

ude,

and

colu

mns

(2)

and

(3)

incl

ud

ea

lin

ear

po

lyn

om

ial

ind

ista

nce

toth

eb

ou

nd

ary.

All

colu

mn

sin

clu

de

aco

ntr

ol

for

dis

tan

ceto

Ho

Ch

iM

inh

Cit

y,d

emo

gra

ph

ic

contr

ols

for

the

num

ber

of

infa

nts

,ch

ildre

n,an

dad

ult

sin

the

house

hold

,an

dyea

rfi

xed

effe

cts.

Colu

mns

(1)

thro

ugh

(11)

incl

ude

boundar

y

segm

ent

fixed

effe

cts,

and

colu

mn

(9)

incl

udes

consi

sten

tpro

vin

cefi

xed

effe

cts.

Robu

stst

andar

der

rors

,cl

ust

ered

atth

evil

lage

level

,ar

ere

port

ed

inp

aren

thes

es.

A. APPENDIX 225

Tabl

eA

.2:C

on

tem

po

rary

Ho

use

ho

ldC

on

sum

pti

on

Incl

ud

ing

Tra

nsf

ers

Dep

end

ent

var

iab

leis

log

ho

use

ho

ldex

pen

dit

ure

.S

pec

ifica

tio

nis

:

Dis

t.L

at-L

on

No

No

No

Co

nsi

st.

Tri

m2

5to

Lat

-Lo

nB

nd

.&

Dis

t.U

rban

HC

MP

rov.

No

On

lyP

rov.

Fo

r1

00

All

Po

lyn

om

ial

HC

MC

Pro

v.C

ap.

Riv

erB

nd

.F

EM

igr.

Km

SV

N

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

(9)

(10

)(1

1)

(12

)

Dai

Vie

t0

.31

50

.25

70

.27

20

.28

30

.32

10

.30

00

.28

40

.35

90

.35

60

.30

30

.39

00

.36

7

(0.0

44

)(0

.04

8)

(0.0

48

)(0

.04

9)

(0.0

45

)(0

.07

0)

(0.0

52

)(0

.06

2)

(0.0

53

)(0

.04

3)

(0.0

79

)(0

.02

4)

Ob

s4

,45

24

,45

24

,45

22

,95

63

,58

82

,63

53

,71

07

42

4,4

52

4,2

94

6,9

55

26

,24

7

Clu

ster

s4

50

45

04

50

31

23

62

25

83

74

76

45

04

50

67

02

58

1

The

unit

of

anal

ysi

sis

the

house

hold

.C

olu

mns

(1)

and

(3)

thro

ugh

(11)

incl

ude

ali

nea

rpoly

nom

ial

inla

titu

de

and

longit

ude,

and

colu

mns

(2)

and

(3)

incl

ud

ea

lin

ear

po

lyn

om

ial

ind

ista

nce

toth

eb

ou

nd

ary.

All

colu

mn

sin

clu

de

aco

ntr

ol

for

dis

tan

ceto

Ho

Ch

iM

inh

Cit

y,d

emo

gra

ph

ic

contr

ols

for

the

num

ber

of

infa

nts

,ch

ildre

n,an

dad

ult

sin

the

house

hold

,an

dyea

rfi

xed

effe

cts.

Colu

mns

(1)

thro

ugh

(11)

incl

ude

boundar

y

segm

ent

fixed

effe

cts,

and

colu

mn

(9)

incl

udes

consi

sten

tpro

vin

cefi

xed

effe

cts.

Robu

stst

andar

der

rors

,cl

ust

ered

atth

evil

lage

level

,ar

ere

port

ed

inp

aren

thes

es.

226 REFERENCES

Tabl

eA

.3:1

00

Kil

om

eter

Bo

un

dar

yS

egm

ent

Fix

edE

ffec

ts

Dep

end

ent

var

iab

leis

log

ho

use

ho

ldex

pen

dit

ure

.S

pec

ifica

tio

nis

:

Dis

t.L

at-L

on

No

No

No

Co

nsi

st.

Tri

m2

5to

Lat

-Lo

nB

nd

.&

Dis

t.U

rban

HC

MP

rov.

No

On

lyP

rov.

Fo

r1

00

All

Po

lyn

om

ial

HC

MC

Pro

v.C

ap.

Riv

erB

nd

.F

EM

igr.

Km

SV

N

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

(9)

(10

)(1

1)

(12

)

Dai

Vie

t0

.25

60

.23

40

.24

20

.24

70

.17

70

.33

80

.31

10

.34

00

.30

10

.21

00

.22

10

.35

1

(0.0

43

)(0

.05

9)

(0.0

61

)(0

.04

4)

(0.0

64

)(0

.05

5)

(0.0

65

)(0

.08

4)

(0.0

60

)(0

.04

0)

(0.0

68

)(0

.02

6)

Ob

s4

,31

94

,31

94

,31

93

,48

32

,56

52

,86

63

,59

77

22

4,3

19

4,2

40

6,7

89

25

,61

7

Clu

ster

s4

50

45

04

50

36

22

58

31

23

74

76

45

04

50

67

02

58

1

The

unit

of

anal

ysi

sis

the

house

hold

.C

olu

mns

(1)

and

(3)

thro

ugh

(11)

incl

ude

ali

nea

rpoly

nom

ial

inla

titu

de

and

longit

ude,

and

colu

mns

(2)

and

(3)

incl

ud

ea

lin

ear

po

lyn

om

ial

ind

ista

nce

toth

eb

ou

nd

ary.

All

colu

mn

sin

clu

de

aco

ntr

ol

for

dis

tan

ceto

Ho

Ch

iM

inh

Cit

y,d

emo

gra

ph

ic

contr

ols

for

the

num

ber

of

infa

nts

,ch

ildre

n,an

dad

ult

sin

the

house

hold

,an

dyea

rfi

xed

effe

cts.

Colu

mns

(1)

thro

ugh

(11)

incl

ude

boundar

y

segm

ent

fixed

effe

cts,

and

colu

mn

(9)

incl

udes

consi

sten

tpro

vin

cefi

xed

effe

cts.

Robu

stst

andar

der

rors

,cl

ust

ered

atth

evil

lage

level

,ar

ere

port

ed

inp

aren

thes

es.

A. APPENDIX 227

Tabl

eA

.4:7

5K

ilo

met

erB

ou

nd

ary

Seg

men

tF

ixed

Eff

ects

Dep

end

ent

var

iab

leis

log

ho

use

ho

ldex

pen

dit

ure

.S

pec

ifica

tio

nis

:

Dis

t.L

at-L

on

No

No

No

Co

nsi

st.

Tri

m2

5to

Lat

-Lo

nB

nd

.&

Dis

t.U

rban

HC

MP

rov.

No

On

lyP

rov.

Fo

r1

00

All

Po

lyn

om

ial

HC

MC

Pro

v.C

ap.

Riv

erB

nd

.F

EM

igr.

Km

SV

N

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

(9)

(10

)(1

1)

(12

)

Dai

Vie

t0

.30

70

.24

90

.25

40

.30

40

.23

40

.37

80

.31

10

.34

00

.36

30

.25

20

.22

90

.35

1

(0.0

45

)(0

.06

3)

(0.0

62

)(0

.04

7)

(0.0

63

)(0

.04

9)

(0.0

65

)(0

.08

4)

(0.0

58

)(0

.04

2)

(0.0

81

)(0

.02

6)

Ob

s4

,31

94

,31

94

,31

93

,48

32

,56

52

,86

63

,59

77

22

4,3

19

4,2

40

6,7

89

25

,61

7

Clu

ster

s4

50

45

04

50

36

22

58

31

23

74

76

45

04

50

67

02

58

1

The

unit

of

anal

ysi

sis

the

house

hold

.C

olu

mns

(1)

and

(3)

thro

ugh

(11)

incl

ude

ali

nea

rpoly

nom

ial

inla

titu

de

and

longit

ude,

and

colu

mns

(2)

and

(3)

incl

ud

ea

lin

ear

po

lyn

om

ial

ind

ista

nce

toth

eb

ou

nd

ary.

All

colu

mn

sin

clu

de

aco

ntr

ol

for

dis

tan

ceto

Ho

Ch

iM

inh

Cit

y,d

emo

gra

ph

ic

contr

ols

for

the

num

ber

of

infa

nts

,ch

ildre

n,an

dad

ult

sin

the

house

hold

,an

dyea

rfi

xed

effe

cts.

Colu

mns

(1)

thro

ugh

(11)

incl

ude

boundar

y

segm

ent

fixed

effe

cts,

and

colu

mn

(9)

incl

udes

consi

sten

tpro

vin

cefi

xed

effe

cts.

Robu

stst

andar

der

rors

,cl

ust

ered

atth

evil

lage

level

,ar

ere

port

ed

inp

aren

thes

es.

228 REFERENCES

Tabl

eA

.5:5

0K

ilo

met

erB

ou

nd

ary

Seg

men

tF

ixed

Eff

ects

Dep

end

ent

var

iab

leis

log

ho

use

ho

ldex

pen

dit

ure

.S

pec

ifica

tio

nis

:

Dis

t.L

at-L

on

No

No

No

Co

nsi

st.

Tri

m2

5to

Lat

-Lo

nB

nd

.&

Dis

t.U

rban

HC

MP

rov.

No

On

lyP

rov.

Fo

r1

00

All

Po

lyn

om

ial

HC

MC

Pro

v.C

ap.

Riv

erB

nd

.F

EM

igr.

Km

SV

N

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

(9)

(10

)(1

1)

(12

)

Dai

Vie

t0

.32

00

.25

20

.25

50

.31

70

.30

90

.36

80

.31

10

.34

00

.41

50

.26

90

.23

60

.35

1

(0.0

51

)(0

.06

2)

(0.0

64

)(0

.05

1)

(0.0

63

)(0

.05

4)

(0.0

65

)(0

.08

4)

(0.0

59

)(0

.04

8)

(0.0

74

)(0

.02

6)

Ob

s4

,31

94

,31

94

,31

93

,48

32

,56

52

,86

63

,59

77

22

4,3

19

4,2

40

6,7

89

25

,61

7

Clu

ster

s4

50

45

04

50

36

22

58

31

23

74

76

45

04

50

67

02

58

1

The

unit

of

anal

ysi

sis

the

house

hold

.C

olu

mns

(1)

and

(3)

thro

ugh

(11)

incl

ude

ali

nea

rpoly

nom

ial

inla

titu

de

and

longit

ude,

and

colu

mns

(2)

and

(3)

incl

ud

ea

lin

ear

po

lyn

om

ial

ind

ista

nce

toth

eb

ou

nd

ary.

All

colu

mn

sin

clu

de

aco

ntr

ol

for

dis

tan

ceto

Ho

Ch

iM

inh

Cit

y,d

emo

gra

ph

ic

contr

ols

for

the

num

ber

of

infa

nts

,ch

ildre

n,an

dad

ult

sin

the

house

hold

,an

dyea

rfi

xed

effe

cts.

Colu

mns

(1)

thro

ugh

(11)

incl

ude

boundar

y

segm

ent

fixed

effe

cts,

and

colu

mn

(9)

incl

udes

consi

sten

tpro

vin

cefi

xed

effe

cts.

Robu

stst

andar

der

rors

,cl

ust

ered

atth

evil

lage

level

,ar

ere

port

ed

inp

aren

thes

es.

A. APPENDIX 229

Tabl

eA

.6:1

0K

ilo

met

erB

ou

nd

ary

Seg

men

tF

ixed

Eff

ects

Dep

end

ent

var

iab

leis

log

ho

use

ho

ldex

pen

dit

ure

.S

pec

ifica

tio

nis

:

Dis

t.L

at-L

on

No

No

No

Co

nsi

st.

Tri

m2

5to

Lat

-Lo

nB

nd

.&

Dis

t.U

rban

HC

MP

rov.

No

On

lyP

rov.

Fo

r1

00

All

Po

lyn

om

ial

HC

MC

Pro

v.C

ap.

Riv

erB

nd

.F

EM

igr.

Km

SV

N

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

(9)

(10

)(1

1)

(12

)

Dai

Vie

t0

.31

00

.21

70

.22

50

.35

10

.23

50

.41

30

.31

10

.34

00

.39

70

.26

70

.33

10

.35

1

(0.0

59

)(0

.06

2)

(0.0

62

)(0

.05

8)

(0.0

61

)(0

.06

1)

(0.0

65

)(0

.08

4)

(0.0

59

)(0

.05

6)

(0.0

87

)(0

.02

6)

Ob

s4

,31

94

,31

94

,31

93

,48

32

,56

52

,86

63

,59

77

22

4,3

19

4,2

40

6,7

89

25

,61

7

Clu

ster

s4

50

45

04

50

36

22

58

31

23

74

76

45

04

50

67

02

58

1

The

unit

of

anal

ysi

sis

the

house

hold

.C

olu

mns

(1)

and

(3)

thro

ugh

(11)

incl

ude

ali

nea

rpoly

nom

ial

inla

titu

de

and

longit

ude,

and

colu

mns

(2)

and

(3)

incl

ud

ea

lin

ear

po

lyn

om

ial

ind

ista

nce

toth

eb

ou

nd

ary.

All

colu

mn

sin

clu

de

aco

ntr

ol

for

dis

tan

ceto

Ho

Ch

iM

inh

Cit

y,d

emo

gra

ph

ic

contr

ols

for

the

num

ber

of

infa

nts

,ch

ildre

n,an

dad

ult

sin

the

house

hold

,an

dyea

rfi

xed

effe

cts.

Colu

mns

(1)

thro

ugh

(11)

incl

ude

boundar

y

segm

ent

fixed

effe

cts,

and

colu

mn

(9)

incl

udes

consi

sten

tpro

vin

cefi

xed

effe

cts.

Robu

stst

andar

der

rors

,cl

ust

ered

atth

evil

lage

level

,ar

ere

port

ed

inp

aren

thes

es.

230 REFERENCES

Tabl

eA

.7:C

on

tem

po

rary

Ho

use

ho

ldC

on

sum

pti

on

20

02

-20

08

Dep

end

ent

var

iab

leis

log

ho

use

ho

ldex

pen

dit

ure

.S

pec

ifica

tio

nis

:

Dis

t.L

at-L

on

No

No

No

Co

nsi

st.

Tri

m2

5to

Lat

-Lo

nB

nd

.&

Dis

t.U

rban

HC

MP

rov.

No

On

lyP

rov.

Fo

r1

00

All

Po

lyn

om

ial

HC

MC

Pro

v.C

ap.

Riv

erB

nd

.F

EM

igr.

Km

SV

N

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

(9)

(10

)(1

1)

(12

)

Dai

Vie

t0

.42

00

.33

40

.36

10

.40

70

.33

80

.38

90

.40

50

.37

30

.49

90

.37

90

.31

90

.34

7

(0.0

72

)(0

.07

6)

(0.0

79

)(0

.07

3)

(0.0

88

)(0

.07

4)

(0.0

86

)(0

.09

4)

(0.0

82

)(0

.06

7)

(0.1

01

)(0

.03

2)

Ob

s3

,01

13

,01

13

,01

12

,41

11

,80

61

,95

82

,50

15

10

3,0

11

2,9

86

5,1

74

19

,10

9

Clu

ster

s2

51

25

12

51

20

31

52

17

02

07

44

25

12

51

42

41

59

2

The

unit

of

anal

ysi

sis

the

house

hold

.C

olu

mns

(1)

and

(3)

thro

ugh

(11)

incl

ude

ali

nea

rpoly

nom

ial

inla

titu

de

and

longit

ude,

and

colu

mns

(2)

and

(3)

incl

ud

ea

lin

ear

po

lyn

om

ial

ind

ista

nce

toth

eb

ou

nd

ary.

All

colu

mn

sin

clu

de

aco

ntr

ol

for

dis

tan

ceto

Ho

Ch

iM

inh

Cit

y,d

emo

gra

ph

ic

contr

ols

for

the

num

ber

of

infa

nts

,ch

ildre

n,an

dad

ult

sin

the

house

hold

,an

dyea

rfi

xed

effe

cts.

Colu

mns

(1)

thro

ugh

(11)

incl

ude

boundar

y

segm

ent

fixed

effe

cts,

and

colu

mn

(9)

incl

udes

consi

sten

tpro

vin

cefi

xed

effe

cts.

Robu

stst

andar

der

rors

,cl

ust

ered

atth

evil

lage

level

,ar

ere

port

ed

inp

aren

thes

es.

A. APPENDIX 231

Tabl

eA

.8:C

on

tem

po

rary

Ho

use

ho

ldC

on

sum

pti

on

20

10

-20

12

Dep

end

ent

var

iab

leis

log

ho

use

ho

ldex

pen

dit

ure

.S

pec

ifica

tio

nis

:

Dis

t.L

at-L

on

No

No

No

Co

nsi

st.

Tri

m2

5to

Lat

-Lo

nB

nd

.&

Dis

t.U

rban

HC

MP

rov.

No

On

lyP

rov.

Fo

r1

00

All

Po

lyn

om

ial

HC

MC

Pro

v.C

ap.

Riv

erB

nd

.F

EM

igr.

Km

SV

N

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

(9)

(10

)(1

1)

(12

)

Dai

Vie

t0

.19

70

.14

40

.15

80

.23

30

.31

70

.21

40

.17

60

.21

70

.25

60

.13

00

.36

70

.36

4

(0.0

76

)(0

.09

3)

(0.0

94

)(0

.07

8)

(0.1

20

)(0

.07

9)

(0.0

92

)(0

.12

2)

(0.0

83

)(0

.07

1)

(0.1

20

)(0

.02

7)

Ob

s1

,30

81

,30

81

,30

81

,07

27

59

90

81

,09

62

12

1,3

08

1,2

54

1,6

15

6,5

08

Clu

ster

s2

82

28

22

82

23

11

65

19

52

37

45

28

22

82

35

81

42

6

The

unit

of

anal

ysi

sis

the

house

hold

.C

olu

mns

(1)

and

(3)

thro

ugh

(11)

incl

ude

ali

nea

rpoly

nom

ial

inla

titu

de

and

longit

ude,

and

colu

mns

(2)

and

(3)

incl

ud

ea

lin

ear

po

lyn

om

ial

ind

ista

nce

toth

eb

ou

nd

ary.

All

colu

mn

sin

clu

de

aco

ntr

ol

for

dis

tan

ceto

Ho

Ch

iM

inh

Cit

y,d

emo

gra

ph

ic

contr

ols

for

the

num

ber

of

infa

nts

,ch

ildre

n,an

dad

ult

sin

the

house

hold

,an

dyea

rfi

xed

effe

cts.

Colu

mns

(1)

thro

ugh

(11)

incl

ude

boundar

y

segm

ent

fixed

effe

cts,

and

colu

mn

(9)

incl

udes

consi

sten

tpro

vin

cefi

xed

effe

cts.

Robu

stst

andar

der

rors

,cl

ust

ered

atth

evil

lage

level

,ar

ere

port

ed

inp

aren

thes

es.

232 REFERENCES

Table A.9: Household Expenditure: Placebo Boundaries

Sample is:

Placebo Boundaries

River Province Expansion

(1) (2) (3)

Dai Viet -0.080 0.091 -0.060

(0.096) (0.109) (0.061)

Obs 1,607 1,535 5,270

Clusters 165 160 397

Mean 9.06 8.84 8.58

The unit of analysis is the household. All

columns include a linear RD polynomial in lat-

itude and longitude, a control for distance to

Ho Chi Minh City, demographic controls for

the number of infants, children, and adults in

the household, year fixed effects, and boundary

segment fixed effects. Robust standard errors,

clustered by village, are reported in parenthe-

ses.

A. APPENDIX 233

Tabl

eA

.10:

Eco

no

mic

Ou

tco

mes

19

69

-19

73

Co

ntr

oll

ing

for

Po

pu

lati

on

Dep

enden

tvar

iable

is:

Log

Non-r

ice

Man

uf.

Surp

lus

House

hold

s%

HH

Lan

d

Fam

ily

Eco

nF

ood

Goods

Goods

Req

uir

eA

cces

sU

nfa

rmed

Pop

Inco

me

LC

AA

vai

l.A

vai

l.P

roduce

dA

ssis

t.V

ehic

.B

adS

ec.

Gro

wth

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

(9)

Dai

Vie

t0.1

51

0.1

35

0.2

50

0.1

56

0.1

52

-0.0

58

0.1

29

0.0

18

-0.0

05

(0.0

44)

(0.0

54)

(0.0

58)

(0.0

61)

(0.0

50)

(0.0

39)

(0.0

19)

(0.0

47)

(0.0

06)

Obs

5,9

26

2,2

85

388

388

388

2,3

30

2,3

32

330

2,2

76

Clu

ster

s172

392

388

388

388

397

396

330

396

Mea

n9.7

20.8

20.7

10.6

30.4

40.6

10.3

40.2

60.0

1

The

unit

of

anal

ysi

sis

the

house

hold

,ham

let,

or

vil

lage.

All

regre

ssio

ns

incl

ude

ali

nea

rR

Dpoly

nom

ial

inla

titu

de

and

lon

git

ud

e,a

con

tro

lfo

rd

ista

nce

toH

oC

hi

Min

hC

ity,

and

bo

un

dar

yse

gm

ent

fixed

effe

cts.

Ro

bu

stst

and

ard

erro

rs,cl

ust

ered

atth

evil

lage

level

,ar

ere

port

edin

par

enth

eses

.

234 REFERENCES

Tabl

eA

.11:

Eco

no

mic

Ou

tco

mes

19

69

-19

73

:N

oH

oC

hi

Min

hC

ity

Dep

enden

tvar

iable

is:

Log

Non-r

ice

Man

uf.

Surp

lus

House

hold

s%

HH

Lan

d

Fam

ily

Eco

nF

ood

Goods

Goods

Req

uir

eA

cces

sU

nfa

rmed

Pop

Inco

me

LC

AA

vai

l.A

vai

l.P

roduce

dA

ssis

t.V

ehic

.B

adS

ec.

Gro

wth

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

(9)

Dai

Vie

t0.1

58

0.1

28

0.2

56

0.1

59

0.1

71

-0.1

32

0.1

18

0.0

09

-0.0

00

(0.0

41)

(0.0

55)

(0.0

60)

(0.0

64)

(0.0

52)

(0.0

39)

(0.0

19)

(0.0

47)

(0.0

06)

Obs

5,9

15

1,5

57

327

327

327

1,5

51

1,5

53

324

1,5

32

Clu

ster

s166

335

327

327

327

336

335

324

338

Mea

n9.7

20.7

50.6

60.5

70.4

10.5

60.2

60.2

60.0

1

The

unit

of

anal

ysi

sis

the

house

hold

,ham

let,

or

vil

lage.

All

regre

ssio

ns

incl

ude

ali

nea

rR

Dpoly

nom

ial

inla

titu

de

and

lon

git

ud

e,a

con

tro

lfo

rd

ista

nce

toH

oC

hi

Min

hC

ity,

and

bo

un

dar

yse

gm

ent

fixed

effe

cts.

Ro

bu

stst

and

ard

erro

rs,cl

ust

ered

atth

evil

lage

level

,ar

ere

port

edin

par

enth

eses

.

A. APPENDIX 235

Tabl

eA

.12:

Eco

no

mic

Ou

tco

mes

19

69

-19

73

:N

oP

rov

inci

alC

apit

als

Dep

enden

tvar

iable

is:

Log

Non-r

ice

Man

uf.

Surp

lus

House

hold

s%

HH

Lan

d

Fam

ily

Eco

nF

ood

Goods

Goods

Req

uir

eA

cces

sU

nfa

rmed

Pop

Inco

me

LC

AA

vai

l.A

vai

l.P

roduce

dA

ssis

t.V

ehic

.B

adS

ec.

Gro

wth

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

(9)

Dai

Vie

t0.1

58

0.1

19

0.2

74

0.1

69

0.1

83

-0.1

22

0.1

15

0.0

03

-0.0

01

(0.0

41)

(0.0

54)

(0.0

59)

(0.0

63)

(0.0

52)

(0.0

40)

(0.0

19)

(0.0

47)

(0.0

06)

Obs

5,9

13

1,5

06

313

313

313

1,5

01

1,5

03

309

1,4

81

Clu

ster

s164

331

313

313

313

333

332

309

334

Mea

n9.7

20.7

40.6

50.5

50.4

00.5

60.2

60.2

60.0

1

The

unit

of

anal

ysi

sis

the

house

hold

,ham

let,

or

vil

lage.

All

regre

ssio

ns

incl

ude

ali

nea

rR

Dpoly

nom

ial

inla

titu

de

and

lon

git

ud

e,a

con

tro

lfo

rd

ista

nce

toH

oC

hi

Min

hC

ity,

and

bo

un

dar

yse

gm

ent

fixed

effe

cts.

Ro

bu

stst

and

ard

erro

rs,cl

ust

ered

atth

evil

lage

level

,ar

ere

port

edin

par

enth

eses

.

236 REFERENCES

Table A.13: Economic Outcomes Following Reunification


Share

State Priv. Paddy Irrig. Mechan.

Land Land Land Paddy Paddy

(1) (2) (3) (4) (5)

Dai Viet 0.119 -0.124 -0.113 0.067 0.244

(0.067) (0.067) (0.076) (0.037) (0.142)

Obs 91 91 73 73 73

Clusters 91 91 73 73 73

Mean 0.40 0.60 0.37 0.05 0.71

The unit of analysis is the district. All columns include a linear

RD polynomial in latitude and longitude and a control for dis-

tance to Ho Chi Minh City. Robust standard errors are reported

in parentheses.

A. APPENDIX 237

Tabl

eA

.14:

Civ

ilS

oci

ety

:C

on

tro

llin

gfo

rP

op

ula

tio

nD

epen

den

tvar

iab

leis

:

Civ

il%

Ho

use

ho

lds

%H

HS

elf-

Dev

.C

ou

nci

lO

rg.

%H

HR

D%

HH

Ho

use

ho

lds

Civ

.S

oc.

So

ciet

yP

arti

cpat

ein

Act

ive

Pro

ject

Dis

cuss

esY

ou

thA

tten

dC

adre

inP

art.

Req

uir

eP

rov

ides

LC

AC

ivic

Org

Eco

n.

Tra

in.

inP

SD

FU

nd

erw

ayG

riev

.A

ctiv

.G

ov

t.M

tgs.

Ham

let

RD

Cad

reA

ssis

tan

ce

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

(9)

(10

)(1

1)

(12

)

Dai

Vie

t0

.16

30

.26

20

.22

00

.06

70

.09

7-0

.01

1-0

.05

30

.10

60

.02

10

.17

4-0

.05

80

.16

2

(0.0

35

)(0

.02

8)

(0.0

27

)(0

.02

8)

(0.0

24

)(0

.02

1)

(0.0

33

)(0

.02

9)

(0.0

30

)(0

.03

6)

(0.0

39

)(0

.04

3)

Ob

s2

,28

52

,32

52

,34

82

,33

03

88

38

43

88

2,3

31

2,3

37

2,3

14

2,3

30

2,2

06

Clu

ster

s3

92

39

73

99

39

73

88

38

43

88

39

73

97

39

63

97

38

8

Mea

n0

.76

0.3

70

.22

0.6

20

.83

0.9

30

.78

0.3

70

.76

0.5

20

.61

0.2

4

Th

eu

nit

of

anal

ysi

sis

the

ham

let

or

vil

lag

e.A

llre

gre

ssio

ns

incl

ud

ea

lin

ear

RD

po

lyn

om

ial

inla

titu

de

and

lon

git

ud

e,a

con

tro

lfo

rd

ista

nce

toH

oC

hi

Min

hC

ity,

and

bo

un

dar

y

seg

men

tfi

xed

effe

cts.

Ro

bu

stst

and

ard

erro

rs,

clu

ster

edat

the

vil

lag

ele

vel

,ar

ere

po

rted

inp

aren

thes

es.

238 REFERENCES

Tabl

eA

.15:

Civ

ilS

oci

ety

:N

oH

oC

hi

Min

hC

ity

Dep

end

ent

var

iab

leis

:

Civ

il%

Ho

use

ho

lds

%H

HS

elf-

Dev

.C

ou

nci

lO

rg.

%H

HR

D%

HH

Ho

use

ho

lds

Civ

.S

oc.

So

ciet

yP

arti

cpat

ein

Act

ive

Pro

ject

Dis

cuss

esY

ou

thA

tten

dC

adre

inP

art.

Req

uir

eP

rov

ides

LC

AC

ivic

Org

Eco

n.

Tra

in.

inP

SD

FU

nd

erw

ayG

riev

.A

ctiv

.G

ov

t.M

tgs.

Ham

let

RD

Cad

reA

ssis

tan

ce

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

(9)

(10

)(1

1)

(12

)

Dai

Vie

t0

.12

10

.22

40

.19

80

.03

70

.09

0-0

.00

6-0

.06

90

.07

5-0

.02

60

.12

9-0

.13

20

.14

5

(0.0

30

)(0

.02

4)

(0.0

28

)(0

.02

5)

(0.0

24

)(0

.02

0)

(0.0

33

)(0

.02

6)

(0.0

30

)(0

.03

2)

(0.0

39

)(0

.04

3)

Ob

s1

,55

71

,54

61

,56

81

,55

13

27

32

33

27

1,5

52

1,5

58

1,5

35

1,5

51

1,4

67

Clu

ster

s3

35

33

63

38

33

63

27

32

33

27

33

63

36

33

53

36

33

1

Mea

n0

.68

0.2

60

.19

0.5

80

.82

0.9

30

.74

0.3

00

.69

0.4

30

.56

0.1

8

Th

eu

nit

of

anal

ysi

sis

the

ham

let

or

vil

lag

e.A

llre

gre

ssio

ns

incl

ud

ea

lin

ear

RD

po

lyn

om

ial

inla

titu

de

and

lon

git

ud

e,a

con

tro

lfo

rd

ista

nce

toH

oC

hi

Min

hC

ity,

and

bo

un

dar

y

seg

men

tfi

xed

effe

cts.

Ro

bu

stst

and

ard

erro

rs,

clu

ster

edat

the

vil

lag

ele

vel

,ar

ere

po

rted

inp

aren

thes

es.

A. APPENDIX 239

Tabl

eA

.16:

Civ

ilS

oci

ety

:N

oP

rov

inci

alC

apit

als

Dep

end

ent

var

iab

leis

:

Civ

il%

Ho

use

ho

lds

%H

HS

elf-

Dev

.C

ou

nci

lO

rg.

%H

HR

D%

HH

Ho

use

ho

lds

Civ

.S

oc.

So

ciet

yP

arti

cpat

ein

Act

ive

Pro

ject

Dis

cuss

esY

ou

thA

tten

dC

adre

inP

art.

Req

uir

eP

rov

ides

LC

AC

ivic

Org

Eco

n.

Tra

in.

inP

SD

FU

nd

erw

ayG

riev

.A

ctiv

.G

ov

t.M

tgs.

Ham

let

RD

Cad

reA

ssis

tan

ce

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

(9)

(10

)(1

1)

(12

)

Dai

Vie

t0

.11

90

.22

60

.20

80

.03

10

.08

7-0

.00

5-0

.06

30

.07

6-0

.03

00

.12

5-0

.12

20

.14

9

(0.0

31

)(0

.02

4)

(0.0

27

)(0

.02

5)

(0.0

24

)(0

.02

0)

(0.0

33

)(0

.02

7)

(0.0

30

)(0

.03

2)

(0.0

40

)(0

.04

3)

Ob

s1

,50

61

,49

61

,51

71

,50

13

13

30

93

13

1,5

02

1,5

08

1,4

85

1,5

01

1,4

16

Clu

ster

s3

31

33

33

34

33

33

13

30

93

13

33

33

33

33

23

33

32

7

Mea

n0

.68

0.2

50

.19

0.5

70

.83

0.9

30

.74

0.3

00

.69

0.4

30

.56

0.1

8

Th

eu

nit

of

anal

ysi

sis

the

ham

let

or

vil

lag

e.A

llre

gre

ssio

ns

incl

ud

ea

lin

ear

RD

po

lyn

om

ial

inla

titu

de

and

lon

git

ud

e,a

con

tro

lfo

rd

ista

nce

toH

oC

hi

Min

hC

ity,

and

bo

un

dar

y

seg

men

tfi

xed

effe

cts.

Ro

bu

stst

and

ard

erro

rs,

clu

ster

edat

the

vil

lag

ele

vel

,ar

ere

po

rted

inp

aren

thes

es.

240 REFERENCES

Tabl

eA

.17:

Lo

cal

Ad

min

istr

atio

n:

Co

ntr

oll

ing

for

Po

pu

lati

on

Dep

end

ent

var

iab

leis

:

Lo

cal

Gov

t.V

ilg

.V

ilg

.H

amle

tP

oli

ceC

hie

fG

ov

t.L

ack

Tec

h.

Pro

v.L

and

Ad

min

.S

yst

.C

om

m.

Ch

ief

Ch

ief

Reg

ula

rly

Co

ntr

ols

Pro

vid

esP

rov.

Tec

h.

Per

s.A

ffai

rs

LC

AT

axes

Fil

led

Pre

sen

tR

DC

adre

Ass

ist.

Ass

ist.

Vis

itV

isit

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

(9)

(10

)(1

1)

Dai

Vie

t0

.03

60

.05

80

.05

70

.05

30

.02

40

.13

10

.08

40

.14

30

.01

0-0

.00

8-0

.00

2

(0.0

17

)(0

.03

6)

(0.0

28

)(0

.03

3)

(0.0

23

)(0

.04

6)

(0.0

20

)(0

.04

7)

(0.0

31

)(0

.04

8)

(0.0

55

)

Ob

s2

,28

53

88

38

83

88

2,3

17

2,3

39

38

22

,22

13

87

38

63

08

Clu

ster

s3

92

38

83

88

38

83

96

39

73

82

39

03

87

38

63

08

Mea

n0

.98

0.8

40

.87

0.9

30

.92

0.5

60

.88

0.3

00

.18

0.5

30

.72

The

unit

of

anal

ysi

sis

the

ham

let

or

vil

lage.

All

regre

ssio

ns

incl

ude

ali

nea

rR

Dpoly

nom

ial

inla

titu

de

and

longit

ude,

aco

ntr

ol

for

dis

tance

to

Ho

Ch

iM

inh

Cit

y,an

db

ou

nd

ary

seg

men

tfi

xed

effe

cts.

Ro

bu

stst

and

ard

erro

rs,

clu

ster

edat

the

vil

lag

ele

vel

,ar

ere

po

rted

inp

aren

thes

es.

A. APPENDIX 241

Tabl

eA

.18:

Lo

cal

Ad

min

istr

atio

n:

No

Ho

Ch

iM

inh

Cit

y

Dep

end

ent

var

iab

leis

:

Lo

cal

Gov

t.V

ilg

.V

ilg

.H

amle

tP

oli

ceC

hie

fG

ov

t.L

ack

Tec

h.

Pro

v.L

and

Ad

min

.S

yst

.C

om

m.

Ch

ief

Ch

ief

Reg

ula

rly

Co

ntr

ols

Pro

vid

esP

rov.

Tec

h.

Per

s.A

ffai

rs

LC

AT

axes

Fil

led

Pre

sen

tR

DC

adre

Ass

ist.

Ass

ist.

Vis

itV

isit

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

(9)

(10

)(1

1)

Dai

Vie

t0

.03

40

.07

50

.05

20

.05

60

.02

10

.08

50

.07

50

.12

3-0

.00

8-0

.02

00

.00

8

(0.0

16

)(0

.03

6)

(0.0

28

)(0

.03

3)

(0.0

23

)(0

.03

9)

(0.0

20

)(0

.04

9)

(0.0

31

)(0

.04

6)

(0.0

56

)

Ob

s1

,55

73

27

32

73

27

1,5

38

1,5

60

32

11

,48

23

29

32

73

04

Clu

ster

s3

35

32

73

27

32

73

35

33

63

21

33

33

29

32

73

04

Mea

n0

.97

0.8

50

.84

0.9

20

.88

0.3

60

.87

0.3

00

.17

0.5

20

.73

The

unit

of

anal

ysi

sis

the

ham

let

or

vil

lage.

All

regre

ssio

ns

incl

ude

ali

nea

rR

Dpoly

nom

ial

inla

titu

de

and

longit

ude,

aco

ntr

ol

for

dis

tance

to

Ho

Ch

iM

inh

Cit

y,an

db

ou

nd

ary

seg

men

tfi

xed

effe

cts.

Ro

bu

stst

and

ard

erro

rs,

clu

ster

edat

the

vil

lag

ele

vel

,ar

ere

po

rted

inp

aren

thes

es.

242 REFERENCES

Tabl

eA

.19:

Lo

cal

Ad

min

istr

atio

n:

No

Pro

vin

cial

Cap

ital

s

Dep

end

ent

var

iab

leis

:

Lo

cal

Gov

t.V

ilg

.V

ilg

.H

amle

tP

oli

ceC

hie

fG

ov

t.L

ack

Tec

h.

Pro

v.L

and

Ad

min

.S

yst

.C

om

m.

Ch

ief

Ch

ief

Reg

ula

rly

Co

ntr

ols

Pro

vid

esP

rov.

Tec

h.

Per

s.A

ffai

rs

LC

AT

axes

Fil

led

Pre

sen

tR

DC

adre

Ass

ist.

Ass

ist.

Vis

itV

isit

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

(9)

(10

)(1

1)

Dai

Vie

t0

.03

50

.07

50

.05

40

.05

80

.01

70

.06

90

.07

40

.13

40

.00

8-0

.01

10

.02

9

(0.0

16

)(0

.03

7)

(0.0

28

)(0

.03

4)

(0.0

24

)(0

.03

8)

(0.0

21

)(0

.04

8)

(0.0

30

)(0

.04

7)

(0.0

55

)

Ob

s1

,50

63

13

31

33

13

1,4

88

1,5

10

30

71

,43

13

14

31

32

92

Clu

ster

s3

31

31

33

13

31

33

32

33

33

07

32

93

14

31

32

92

Mea

n0

.97

0.8

40

.84

0.9

20

.87

0.3

40

.87

0.3

10

.17

0.5

10

.73

The

unit

of

anal

ysi

sis

the

ham

let

or

vil

lage.

All

regre

ssio

ns

incl

ude

ali

nea

rR

Dpoly

nom

ial

inla

titu

de

and

longit

ude,

aco

ntr

ol

for

dis

tance

to

Ho

Ch

iM

inh

Cit

y,an

db

ou

nd

ary

seg

men

tfi

xed

effe

cts.

Ro

bu

stst

and

ard

erro

rs,

clu

ster

edat

the

vil

lag

ele

vel

,ar

ere

po

rted

inp

aren

thes

es.

A. APPENDIX 243

Tabl

eA

.20:

Pu

bli

cG

oo

ds:

Co

ntr

oll

ing

for

Po

pu

lati

on

Dep

end

ent

var

iab

leis

:

Hea

lth

Gov

t.H

ealt

hH

ealt

hM

at.

Pri

mar

yS

eco

nd

ary

Att

end

.L

aw

Car

eM

ed.

Ser

v.W

krs

.V

isit

Cli

nic

Ed

uc.

Sch

oo

lS

cho

ol

Res

tr.

En

forc

ed

LC

AA

vai

l.R

eg.

inV

illa

ge

LC

AA

cces

s.C

om

ple

tio

nIn

Vil

g.

Att

end

.b

yS

ec.

Day

/Nig

ht

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

(9)

(10

)(1

1)

(12

)

Dai

Vie

t0

.12

70

.15

70

.31

90

.12

9-0

.01

00

.02

90

.04

80

.07

60

.03

40

.02

2-0

.01

50

.20

1

(0.0

43

)(0

.03

3)

(0.0

42

)(0

.05

0)

(0.0

66

)(0

.04

4)

(0.0

23

)(0

.03

1)

(0.0

53

)(0

.01

3)

(0.0

13

)(0

.04

6)

Ob

s2

,28

52

,33

92

,33

63

88

38

82

,28

52

,33

63

88

38

83

88

2,3

33

2,3

33

Clu

ster

s3

92

39

73

97

38

83

88

39

23

96

38

83

88

38

83

96

39

7

Mea

n0

.86

0.3

90

.47

0.7

90

.61

0.8

20

.90

0.6

10

.35

0.1

80

.02

0.7

9

Th

eu

nit

of

anal

ysi

sis

the

ham

let

or

vil

lag

e.A

llre

gre

ssio

ns

incl

ud

ea

lin

ear

RD

po

lyn

om

ial

inla

titu

de

and

lon

git

ud

e,a

con

tro

lfo

rd

ista

nce

toH

oC

hi

Min

h

Cit

y,an

db

ou

nd

ary

seg

men

tfi

xed

effe

cts.

Ro

bu

stst

and

ard

erro

rs,

clu

ster

edat

the

vil

lag

ele

vel

,ar

ere

po

rted

inp

aren

thes

es.

244 REFERENCES

Tabl

eA

.21:

Pu

bli

cG

oo

ds:

No

Ho

Ch

iM

inh

Cit

y

Dep

end

ent

var

iab

leis

:

Hea

lth

Gov

t.H

ealt

hH

ealt

hM

at.

Pri

mar

yS

eco

nd

ary

Att

end

.L

aw

Car

eM

ed.

Ser

v.W

krs

.V

isit

Cli

nic

Ed

uc.

Sch

oo

lS

cho

ol

Res

tr.

En

forc

ed

LC

AA

vai

l.R

eg.

inV

illa

ge

LC

AA

cces

s.C

om

ple

tio

nIn

Vil

g.

Att

end

.b

yS

ec.

Day

/Nig

ht

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

(9)

(10

)(1

1)

(12

)

Dai

Vie

t0

.11

30

.19

90

.31

30

.13

60

.00

60

.01

50

.07

00

.08

40

.05

30

.02

1-0

.02

00

.17

5

(0.0

42

)(0

.03

8)

(0.0

41

)(0

.05

0)

(0.0

69

)(0

.04

4)

(0.0

21

)(0

.03

1)

(0.0

59

)(0

.01

3)

(0.0

13

)(0

.04

3)

Ob

s1

,55

71

,56

01

,55

73

27

32

71

,55

71

,55

73

27

32

73

27

1,5

54

1,5

54

Clu

ster

s3

35

33

63

36

32

73

27

33

53

35

32

73

27

32

73

35

33

6

Mea

n0

.80

0.3

50

.40

0.7

70

.56

0.7

50

.90

0.5

90

.28

0.1

70

.03

0.6

9

Th

eu

nit

of

anal

ysi

sis

the

ham

let

or

vil

lag

e.A

llre

gre

ssio

ns

incl

ud

ea

lin

ear

RD

po

lyn

om

ial

inla

titu

de

and

lon

git

ud

e,a

con

tro

lfo

rd

ista

nce

toH

oC

hi

Min

h

Cit

y,an

db

ou

nd

ary

seg

men

tfi

xed

effe

cts.

Ro

bu

stst

and

ard

erro

rs,

clu

ster

edat

the

vil

lag

ele

vel

,ar

ere

po

rted

inp

aren

thes

es.

A. APPENDIX 245

Tabl

eA

.22:

Pu

bli

cG

oo

ds:

No

Pro

vin

cial

Cap

ital

s

Dep

end

ent

var

iab

leis

:

Hea

lth

Gov

t.H

ealt

hH

ealt

hM

at.

Pri

mar

yS

eco

nd

ary

Att

end

.L

aw

Car

eM

ed.

Ser

v.W

krs

.V

isit

Cli

nic

Ed

uc.

Sch

oo

lS

cho

ol

Res

tr.

En

forc

ed

LC

AA

vai

l.R

eg.

inV

illa

ge

LC

AA

cces

s.C

om

ple

tio

nIn

Vil

g.

Att

end

.b

yS

ec.

Day

/Nig

ht

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

(9)

(10

)(1

1)

(12

)

Dai

Vie

t0

.10

80

.20

30

.31

20

.13

7-0

.00

00

.00

20

.07

10

.08

40

.05

70

.02

2-0

.02

00

.18

0

(0.0

43

)(0

.03

8)

(0.0

41

)(0

.05

2)

(0.0

70

)(0

.04

5)

(0.0

22

)(0

.03

2)

(0.0

58

)(0

.01

4)

(0.0

13

)(0

.04

4)

Ob

s1

,50

61

,51

01

,50

73

13

31

31

,50

61

,50

73

13

31

33

13

1,5

04

1,5

04

Clu

ster

s3

31

33

33

33

31

33

13

33

13

32

31

33

13

31

33

32

33

3

Mea

n0

.80

0.3

50

.39

0.7

60

.56

0.7

40

.89

0.5

90

.27

0.1

70

.03

0.6

8

Th

eu

nit

of

anal

ysi

sis

the

ham

let

or

vil

lag

e.A

llre

gre

ssio

ns

incl

ud

ea

lin

ear

RD

po

lyn

om

ial

inla

titu

de

and

lon

git

ud

e,a

con

tro

lfo

rd

ista

nce

toH

oC

hi

Min

h

Cit

y,an

db

ou

nd

ary

seg

men

tfi

xed

effe

cts.

Ro

bu

stst

and

ard

erro

rs,

clu

ster

edat

the

vil

lag

ele

vel

,ar

ere

po

rted

inp

aren

thes

es.

246 REFERENCES

Tabl

eA

.23:

Pu

bli

cO

pin

ion

:C

on

tro

llin

gfo

rP

op

ula

tio

n

Dep

enden

tvar

iable

is:

Gvt.

Loca

lK

now

sV

ilg

LT

TN

atl.

Gvt.

Peo

ple

Act

ive

inP

eople

Res

ponsi

ve

Offi

cial

sA

dm

in.

Fai

rly

Per

form

sM

an.E

con.

Res

pons.

Inte

rest

Dec

ide

Cit

izen

sS

ucc

essf

ul

Str

uct

.W

ell

Adm

inis

t.P

oorl

yP

oorl

yC

om

m.L

ife

Gro

up

SD

P

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

(9)

Dai

Vie

t0.1

13

0.1

54

0.0

73

0.3

17

0.0

92

0.0

81

0.2

71

0.3

38

0.2

08

(0.0

45)

(0.0

74)

(0.0

56)

(0.0

80)

(0.0

38)

(0.0

40)

(0.0

76)

(0.1

46)

(0.0

55)

Obs

2,7

79

3,4

87

1,4

57

999

2,8

11

5,7

78

879

243

353

Clu

ster

s190

183

89

101

182

215

106

35

53

Mea

n0.3

70.5

20.2

20.5

70.1

90.3

10.1

80.1

80.2

3

The

unit

of

anal

ysi

sis

the

indiv

idual

.A

llre

gre

ssio

ns

incl

ude

ali

nea

rR

Dpoly

nom

ial

inla

titu

de

and

longit

ude,

aco

ntr

ol

for

dis

tance

toH

oC

hi

Min

hC

ity,

and

boundar

yse

gm

ent

fixed

effe

cts.

Robust

stan

dar

der

rors

,cl

ust

ered

atth

evil

lage

level

,ar

ere

port

edin

par

enth

eses

.

A. APPENDIX 247

Tabl

eA

.24:

Pu

bli

cO

pin

ion

:N

oH

oC

hi

Min

hC

ity

Dep

enden

tvar

iable

is:

Gvt.

Loca

lK

now

sV

ilg

LT

TN

atl.

Gvt.

Peo

ple

Act

ive

inP

eople

Res

ponsi

ve

Offi

cial

sA

dm

in.

Fai

rly

Per

form

sM

an.E

con.

Res

pons.

Inte

rest

Dec

ide

Cit

izen

sS

ucc

essf

ul

Str

uct

.W

ell

Adm

inis

t.P

oorl

yP

oorl

yC

om

m.L

ife

Gro

up

SD

P

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

(9)

Dai

Vie

t0.1

10

0.1

70

0.0

96

0.3

50

0.0

93

0.1

17

0.2

25

0.3

95

0.2

22

(0.0

41)

(0.0

77)

(0.0

46)

(0.0

75)

(0.0

42)

(0.0

33)

(0.0

73)

(0.1

46)

(0.0

52)

Obs

1,5

90

1,7

50

335

999

1,4

32

2,5

58

532

243

353

Clu

ster

s141

132

43

101

131

160

70

35

53

Mea

n0.3

20.4

50.1

60.5

70.1

70.1

80.2

60.1

80.2

3

The

unit

of

anal

ysi

sis

the

indiv

idual

.A

llre

gre

ssio

ns

incl

ude

ali

nea

rR

Dpoly

nom

ial

inla

titu

de

and

longit

ude,

aco

ntr

ol

for

dis

tance

toH

oC

hi

Min

hC

ity,

and

boundar

yse

gm

ent

fixed

effe

cts.

Robust

stan

dar

der

rors

,cl

ust

ered

atth

evil

lage

level

,ar

ere

port

edin

par

enth

eses

.

248 REFERENCES

Tabl

eA

.25:

Pu

bli

cO

pin

ion

:N

oP

rov

inci

alC

apit

als

Dep

enden

tvar

iable

is:

Gvt.

Loca

lK

now

sV

ilg

LT

TN

atl.

Gvt.

Peo

ple

Act

ive

inP

eople

Res

ponsi

ve

Offi

cial

sA

dm

in.

Fai

rly

Per

form

sM

an.E

con.

Res

pons.

Inte

rest

Dec

ide

Cit

izen

sS

ucc

essf

ul

Str

uct

.W

ell

Adm

inis

t.P

oorl

yP

oorl

yC

om

m.L

ife

Gro

up

SD

P

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

(9)

Dai

Vie

t0.0

94

0.1

52

0.0

77

0.3

50

0.0

62

0.1

10

0.1

98

0.3

95

0.2

22

(0.0

42)

(0.0

80)

(0.0

49)

(0.0

75)

(0.0

40)

(0.0

33)

(0.0

75)

(0.1

46)

(0.0

52)

Obs

1,3

16

1,2

73

235

999

1,0

65

2,2

58

432

243

353

Clu

ster

s141

128

39

101

127

160

66

35

53

Mea

n0.2

80.3

60.1

10.5

70.1

40.1

70.2

30.1

80.2

3

The

unit

of

anal

ysi

sis

the

indiv

idual

.A

llre

gre

ssio

ns

incl

ude

ali

nea

rR

Dpoly

nom

ial

inla

titu

de

and

longit

ude,

aco

ntr

ol

for

dis

tance

toH

oC

hi

Min

hC

ity,

and

boundar

yse

gm

ent

fixed

effe

cts.

Robust

stan

dar

der

rors

,cl

ust

ered

atth

evil

lage

level

,ar

ere

port

edin

par

enth

eses

.

A. APPENDIX 249

Table A.26: Current Outcomes: No Urban Ho Chi Minh City




Fund Lower Sec. > 25 25-40 40-60 >60

(1) (2) (3) (4) (5) (6)

Dai Viet 0.122 0.310 0.970 0.886 1.045 1.044

(0.032) (0.076) (0.201) (0.195) (0.244) (0.240)

Obs 4,689 112 33,000 14,186 13,353 5,461

Clusters 362 112 365 364 365 354

Mean 0.69 0.79 7.28 8.26 7.50 4.16

The unit of analysis is the household, district, or individual. All columns include

a linear RD polynomial in latitude and longitude, a control for distance to Ho Chi


include year fixed effects. Robust standard errors, clustered at the village level,

are reported in parentheses.

250 REFERENCES

Table A.27: Current Outcomes: No Ho Chi Minh Province




Fund Lower Sec. > 25 25-40 40-60 >60

(1) (2) (3) (4) (5) (6)

Dai Viet 0.060 0.113 1.847 1.857 2.019 1.628

(0.051) (0.059) (0.334) (0.342) (0.411) (0.350)

Obs 3,448 100 23,420 9,939 9,520 3,961

Clusters 258 100 260 259 260 250

Mean 0.66 0.77 6.79 7.82 7.02 3.63






A. APPENDIX 251

Table A.28: Current Outcomes: No Provincial Capitals




Fund Lower Sec. > 25 25-40 40-60 >60

(1) (2) (3) (4) (5) (6)

Dai Viet 0.132 0.334 0.692 0.577 0.748 0.769

(0.036) (0.081) (0.211) (0.200) (0.257) (0.274)

Obs 3,893 106 27,545 11,861 11,054 4,630

Clusters 312 106 314 314 314 304

Mean 0.68 0.80 7.06 8.05 7.23 4.06






252 REFERENCES

Tabl

eA

.29:

Th

eV

ietn

amW

ar:

Co

ntr

oll

ing

for

Po

pu

lati

on

Dep

end

ent

var

iab

leis

:

VC

VC

Vil

g.

VC

VC

Fri

end

lyA

ir/A

rt.

U..

S.

SV

NT

erri

t.

Sec

uri

tyF

orc

esB

ase

Gu

err.

Mai

nIn

fra

VC

Fo

rces

Str

ke

Init

iate

dF

orc

es

LC

AP

rese

nt

Nea

rby

Sq

uad

Sq

uad

Act

ivit

yT

axat

ion

Nea

rby

Nea

rby

Att

ack

Pre

sen

t

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

(9)

(10

)(1

1)

(12

)

Dai

Vie

t-0

.01

5-0

.03

3-0

.00

90

.07

3-0

.10

00

.03

5-0

.01

2-0

.01

0-0

.03

7-0

.00

0-0

.09

5-0

.02

0

(0.0

42

)(0

.03

3)

(0.0

49

)(0

.03

9)

(0.0

40

)(0

.03

2)

(0.0

16

)(0

.03

7)

(0.0

32

)(0

.00

2)

(0.0

23

)(0

.02

7)

Ob

s2

,28

52

,33

53

90

39

03

90

2,3

39

38

93

89

38

82

,34

82

,34

82

,34

8

Clu

ster

s3

92

39

83

90

39

03

90

39

83

89

38

93

88

39

93

99

39

9

Mea

n0

.80

0.1

50

.49

0.2

00

.23

0.0

90

.07

0.4

90

.13

0.0

00

.71

0.2

4

The

unit

of

anal

ysi

sis

the

ham

let

or

vil

lage.

All

regre

ssio

ns

incl

ude

ali

nea

rR

Dpoly

nom

ial

inla

titu

de

and

longit

ude,

aco

ntr

ol

for

dis

tance

to

Ho

Ch

iM

inh

Cit

y,an

db

ou

nd

ary

seg

men

tfi

xed

effe

cts.

Ro

bu

stst

and

ard

erro

rs,

clu

ster

edat

the

vil

lag

ele

vel

,ar

ere

po

rted

inp

aren

thes

es.

A. APPENDIX 253

Tabl

eA

.30:

Th

eV

ietn

amW

ar:

No

Ho

Ch

iM

inh

Cit

y

Dep

end

ent

var

iab

leis

:

VC

VC

Vil

g.

VC

VC

Fri

end

lyA

ir/A

rt.

U..

S.

SV

NT

erri

t.

Sec

uri

tyF

orc

esB

ase

Gu

err.

Mai

nIn

fra

VC

Fo

rces

Str

ke

Init

iate

dF

orc

es

LC

AP

rese

nt

Nea

rby

Sq

uad

Sq

uad

Act

ivit

yT

axat

ion

Nea

rby

Nea

rby

Att

ack

Pre

sen

t

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

(9)

(10

)(1

1)

(12

)

Dai

Vie

t-0

.02

1-0

.02

40

.00

90

.07

4-0

.09

60

.02

9-0

.01

1-0

.04

6-0

.03

9-0

.00

0-0

.06

50

.03

2

(0.0

43

)(0

.03

4)

(0.0

50

)(0

.03

9)

(0.0

39

)(0

.03

3)

(0.0

16

)(0

.03

8)

(0.0

32

)(0

.00

1)

(0.0

18

)(0

.02

3)

Ob

s1

,55

71

,55

63

29

32

93

29

1,5

60

32

93

29

32

81

,57

81

,57

81

,56

8

Clu

ster

s3

35

33

73

29

32

93

29

33

73

29

32

93

28

33

93

39

33

8

Mea

n0

.73

0.2

10

.57

0.2

40

.27

0.1

10

.08

0.4

70

.15

0.0

00

.78

0.3

5

The

unit

of

anal

ysi

sis

the

ham

let

or

vil

lage.

All

regre

ssio

ns

incl

ude

ali

nea

rR

Dpoly

nom

ial

inla

titu

de

and

longit

ude,

aco

ntr

ol

for

dis

tance

to

Ho

Ch

iM

inh

Cit

y,an

db

ou

nd

ary

seg

men

tfi

xed

effe

cts.

Ro

bu

stst

and

ard

erro

rs,

clu

ster

edat

the

vil

lag

ele

vel

,ar

ere

po

rted

inp

aren

thes

es.

254 REFERENCES

Tabl

eA

.31:

Th

eV

ietn

amW

ar:

No

Pro

vin

cial

Cap

ital

s

Dep

end

ent

var

iab

leis

:

VC

VC

Vil

g.

VC

VC

Fri

end

lyA

ir/A

rt.

U..

S.

SV

NT

erri

t.

Sec

uri

tyF

orc

esB

ase

Gu

err.

Mai

nIn

fra

VC

Fo

rces

Str

ke

Init

iate

dF

orc

es

LC

AP

rese

nt

Nea

rby

Sq

uad

Sq

uad

Act

ivit

yT

axat

ion

Nea

rby

Nea

rby

Att

ack

Pre

sen

t

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

(9)

(10

)(1

1)

(12

)

Dai

Vie

t-0

.03

0-0

.01

70

.00

90

.07

2-0

.10

00

.03

3-0

.01

2-0

.05

0-0

.04

0-0

.00

0-0

.04

80

.04

0

(0.0

44

)(0

.03

5)

(0.0

50

)(0

.04

0)

(0.0

40

)(0

.03

4)

(0.0

16

)(0

.03

9)

(0.0

32

)(0

.00

2)

(0.0

16

)(0

.02

4)

Ob

s1

,50

61

,50

63

14

31

43

14

1,5

09

31

43

14

31

31

,52

71

,52

71

,51

7

Clu

ster

s3

31

33

43

14

31

43

14

33

33

14

31

43

13

33

53

35

33

4

Mea

n0

.72

0.2

20

.59

0.2

40

.27

0.1

20

.08

0.4

80

.16

0.0

00

.79

0.3

5

The

unit

of

anal

ysi

sis

the

ham

let

or

vil

lage.

All

regre

ssio

ns

incl

ude

ali

nea

rR

Dpoly

nom

ial

inla

titu

de

and

longit

ude,

aco

ntr

ol

for

dis

tance

to

Ho

Ch

iM

inh

Cit

y,an

db

ou

nd

ary

seg

men

tfi

xed

effe

cts.

Ro

bu

stst

and

ard

erro

rs,

clu

ster

edat

the

vil

lag

ele

vel

,ar

ere

po

rted

inp

aren

thes

es.

A. APPENDIX 255

Tabl

eA

.32:

Ad

dit

ion

alM

ech

anis

ms:

No

Ho

Ch

iM

inh

Cit

y

Dep

enden

tvar

iable

is:

Agri

c.M

ain

Shar

eH

.H.

Em

plo

yed

Agri

c.L

and

Job

inA

nnual

Per

ennia

lR

esid

enti

alIn

tere

stIn

form

al

H.H

.S

ize

Indust

ryL

and

Cer

tifi

edE

xpen

ses

Sec

tor

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

Dai

Vie

t-0

.184

-0.0

10

-0.0

20

-0.1

18

-0.1

25

-0.2

18

-0.1

13

-0.0

70

(0.0

35)

(0.1

11)

(0.0

23)

(0.0

41)

(0.0

49)

(0.0

64)

(0.0

31)

(0.0

22)

Obs

13,2

05

4,4

71

16,5

18

176

173

170

3,5

90

16,5

04

Clu

ster

s365

270

365

131

129

128

203

365

Mea

n0.2

80.8

80.2

50.9

30.9

20.9

40.2

60.6

4

The

unit

of

anal

ysi

sis

the

ho

use

ho

ld,

ind

ivid

ual

,o

rco

mm

un

e.A

llco

lum

ns

incl

ud

ea

lin

ear

RD

po

lyn

o-

mia

lin

lati

tude

and

longit

ude,

aco

ntr

ol

for

dis

tance

toH

oC

hi

Min

hC

ity,

yea

rfi

xed

effe

cts,

and

boundar

y

segm

ent

fixed

effe

cts.

Robust

stan

dar

der

rors

,cl

ust

ered

atth

evil

lage

level

,ar

ere

port

edin

par

enth

eses

.

256 REFERENCES

Tabl

eA

.33:

Ad

dit

ion

alM

ech

anis

ms:

No

Ho

Ch

iM

inh

Pro

vin

ce

Dep

enden

tvar

iable

is:

Agri

c.M

ain

Shar

eH

.H.

Em

plo

yed

Agri

c.L

and

Job

inA

nnual

Per

ennia

lR

esid

enti

alIn

tere

stIn

form

al

H.H

.S

ize

Indust

ryL

and

Cer

tifi

edE

xpen

ses

Sec

tor

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

Dai

Vie

t-0

.218

-0.0

06

0.0

16

-0.2

71

-0.2

10

-0.2

30

-0.1

55

-0.0

73

(0.0

63)

(0.1

77)

(0.0

42)

(0.1

27)

(0.1

16)

(0.1

22)

(0.0

53)

(0.0

41)

Obs

9,5

50

4,2

23

12,1

97

161

160

157

2,6

62

12,1

83

Clu

ster

s260

234

260

116

116

115

152

260

Mea

n0.3

80.9

00.2

50.9

40.9

30.9

60.3

10.6

9

The

unit

of

anal

ysi

sis

the

ho

use

ho

ld,

ind

ivid

ual

,o

rco

mm

un

e.A

llco

lum

ns

incl

ud

ea

lin

ear

RD

po

lyn

o-

mia

lin

lati

tude

and

longit

ude,

aco

ntr

ol

for

dis

tance

toH

oC

hi

Min

hC

ity,

yea

rfi

xed

effe

cts,

and

boundar

y

segm

ent

fixed

effe

cts.

Robust

stan

dar

der

rors

,cl

ust

ered

atth

evil

lage

level

,ar

ere

port

edin

par

enth

eses

.

A. APPENDIX 257

Tabl

eA

.34:

Ad

dit

ion

alM

ech

anis

ms:

No

Pro

vin

cial

Cap

ital

s

Dep

enden

tvar

iable

is:

Agri

c.M

ain

Shar

eH

.H.

Em

plo

yed

Agri

c.L

and

Job

inA

nnual

Per

ennia

lR

esid

enti

alIn

tere

stIn

form

al

H.H

.S

ize

Indust

ryL

and

Cer

tifi

edE

xpen

ses

Sec

tor

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

Dai

Vie

t-0

.169

0.0

25

-0.0

32

-0.1

19

-0.1

25

-0.2

16

-0.1

07

-0.0

46

(0.0

40)

(0.1

15)

(0.0

23)

(0.0

42)

(0.0

50)

(0.0

64)

(0.0

31)

(0.0

23)

Obs

11,1

54

4,2

23

13,9

75

171

168

165

2,9

63

13,9

61

Clu

ster

s314

232

314

127

125

124

170

314

Mea

n0.3

30.8

90.2

30.9

30.9

20.9

40.2

80.6

6

The

unit

of

anal

ysi

sis

the

ho

use

ho

ld,

ind

ivid

ual

,o

rco

mm

un

e.A

llco

lum

ns

incl

ud

ea

lin

ear

RD

po

lyn

o-

mia

lin

lati

tude

and

longit

ude,

aco

ntr

ol

for

dis

tance

toH

oC

hi

Min

hC

ity,

yea

rfi

xed

effe

cts,

and

boundar

y

segm

ent

fixed

effe

cts.

Robust

stan

dar

der

rors

,cl

ust

ered

atth

evil

lage

level

,ar

ere

port

edin

par

enth

eses

.

258 REFERENCES

Table A.35: Foreign Sector Employment


Share Employment

Foreign Private State

(1) (2) (3)

Dai Viet -0.069 0.034 0.035

(0.039) (0.040) (0.013)

Obs 640 640 640

Clusters 640 640 640

Mean 0.17 0.79 0.04

The unit of analysis is the village. All

columns include a linear RD polynomial

in latitude and longitude, a control for dis-

tance to Ho Chi Minh City, and boundary

segment fixed effects. Robust standard er-

rors are reported in parentheses.

Sammanfattning

Denna avhandling består av tre fristående artiklar om utvecklingens po-

litiska ekonomi som studerar de relativa erfarenheterna i tre asiatiska länder:

Sydkorea, Filippinerna och Vietnam. Avhandlingen berättar emellertid historien

om två angränsande världar: de komparativa historiska erfarenheterna i Ost-

och Sydostasien.

De asiatiska ländernas ambition löper som en gemensam tråd genom varje

kapitel. Avhandlingen handlar om inflytandet av vad statsvetaren (och asienex-

perten) James C. Scott kallade “stora modernistiska” ambitioner (Scott 1998).

Kapitel 2 studerar en storskalig industripolitik som förts av Sydkorea under

autokraten Park Chung Hee. Kapitel 3 studerar effekten av en lika ambitiös,

men mycket distinkt, moderniseringsansträngning som drivits av Ferdinand

Marcos på Filippinerna. Kapitel 4 förenar de distinkta utvecklingsbanorna i

Ostasien och Sydostasien genom att undersöka de historiska institutionerna i

Vietnam – ett land som stod mellan de två civilisationerna. Här omvandlades

till och med de tidigaste moderna statsbyggnadsprojekten i asiatiska imperier

till nutida utfall av utvecklingen.

Statsvetaren Paul Hutchcroft har kallat sydkoreanska och filippinska dik-

taturer “spegelvända bilder” av varandra, inte minst till följd av de kraftiga

skillnaderna i deras utfall (Hutchcroft 2011). Kapitel 2 och 3, Parks Yushin

Fourth Republic och Marcos New Society, var båda allierade med västvärlden

under det kalla kriget. På 1970-talet lyckades båda dessa kraftkarlar omvandla

sina länder från demokrati till diktatur mitt under en politisk kris. Och i båda

miljöerna möjliggjordes konsolidering genom en ekonomisk elit som ansåg

envälde vara moderniseringens pris.

Den tydliga sektorpartiskheten i politiken i kapitel 2 och 3 avslöjar den

underliggande elitpolitiken i varje regim (Kang 2002a). I fallet Sydkorea så

appellerade Parks regim till industrikapitalisternas intressen, då deras intressen

låg i linje med statens under det existentiella hotet om kommunistisk invasion.

259

260 SAMMANFATTNING

Den speciella säregenheten i denna omgivning möjliggjorde Sydkoreas ambi-

tiösa industripolitik. Samtidigt utgjorde Ferdinand Marcos gröna revolution ett

projekt som syftade till att modernisera och främja de traditionella maktkällorna

på Filippinerna, nämligen jordägarna.

Viktigare är att fallen Republiken Korea och den filippinska republiken,

såväl som Vietnams tumultartade historia, är parabler för Ost- kontra Sydostasi-

en. Sydkoreas stora ansträngning när det gällde tung kemisk industri utgjorde

en reklambild för den politik som fördes över hela Ostasien efter andra världs-

kriget. Sydkorea formade sig efter Japans Meiji restauration och lockades av

dess förvandling på 1800-talet. Å andra sidan exemplifierade den gröna revo-

lutionen den dubbla drömmen hos utvecklare i Sydostasien: att modernisera

och mildra oron på landsbygden (Cullather 2004, Cullather 2013). Insatserna i

dessa genetiska innovationer var höga, från Malaysias vacklande risskål (Barker

1985) till ett krigshärjat Sydvietnam (Poppel 2015). Å andra sidan hävdar jag i

kapitel 4 att lokala statliga institutioners mönster tilltalar allmänna former av

statsbildning sett över de två regionerna: den siniska staten i Ostasien och de

indiska staterna som symboliserar sydostasiatiska politiska institutioner.

Den sydkoreanska tillväxtperioden var en av de mest dramatiska perio-

derna i utvecklingen efter andra världskriget. När Park Chung Hee tog makten

1961 hade landet samma BNP per capita som Ghana. På 1980-talet så undergick

Sydkorea en industriomvandling som det hade tagit västerländerna ett århund-

rade att åstadkomma (Nelson 1999). Kapitel 2, “Manufacturing Revolutions –

Industrial Policy and Networks in South Korea” (Revolutioner inom tillverk-

ningsindustrin – industripolitik och nätverk i Sydkorea), studerar effekten av en

större industriell intervention under denna period: Sydkoreas Heavy Chemical

and Industry (HCI) satsning (1973-1979). Den stora kraftansträngningen med

HCI utgjorde hörnstenen i Park Chung Hees nya diktatur, ett försök att för-

ändra landet från att vara en exportör av plywood, peruker och fotbeklädnader,

till ett samhälle som en dag skulle kunna producera inhemska vapen. Denna

artikel använder de historiska omständigheterna kring Sydkoreas kraftansträng-

ning, tillsammans med nyligen digitaliserade data, för att studera effekten av

industripolitik på industriell utveckling.

Genom att studera Sydkoreas kraftansträngning ger jag tre bidrag till

forskningen. Först beräknar jag effekten av industripolitiken på de kortsiktiga

261

utfallen av industriell utveckling. Detta gör jag genom att jämföra utvecklingen

av tillverkningsindustrier som är föremål för åtgärder respektive inte föremål

för åtgärder och efter det plötsliga tillkännagivandet av politiken. Jag visar

de positiva effekterna av industripolitiken på produktionstillväxten, sysselsätt-

ningen och arbetsproduktiviteten i de sektorer som varit föremål för åtgärder

respektive de som inte varit det. För det andra så bedömer jag överspillnings-

effekterna av interventionen och följer hur politiken sprider sig genom länkar

mellan branscher. Jag reder ut effekterna genom länkar framåt och bakåt och

motiverar mina resultat genom att använda en enkel modell av det sydkoreanska

ekonomiska nätverket. På så sätt finner jag att industripolitiken främjade tillväxt,

inträde och kapitaltillväxt i sektorer som ligger högre upp i förädlingskedjan

än den bransch som varit föremål för åtgärder. Å andra sidan så visar denna

analys en minskning för branscher längre ner på förädlingskedjan som har

de starkaste direktkontakterna med branscher som varit föremål för åtgärder,

eftersom branscher som var föremål för åtgärder importerade konkurrerande

produkter.

Slutligen testar jag huruvida effekterna av satsningen kvarstod efter pla-

neringsperioden, både i sektorer som var föremål för politiken och de som

påverkades av politiken genom kopplingar. Jag finner bevis för bestående pen-

ningexternaliteter så som de som antas av utvecklingsteoretiker som studerar

den stora kraftansträngningen, t ex Albert Hirschman (1958). Med andra ord

så finner jag att Sydkoreas kontroversiella industripolitik var framgångsrik när

det gällde att skapa industriell utveckling, de fördelar som kvarstod över tiden

och i industrier som inte var direkt föremål för politiska åtgärder. I genom-

snitt så medförde HCI-politiken ca 80 procent mer tillväxt och främjade en 11

procentig minskning i tillverkningspriser för branscher som var föremål för

åtgärder respektive inte föremål för åtgärder. Sammantaget visar dessa resultat

att industripolitiken gynnade Sydkoreas flytt uppåt i utbudskedjan.

Kapitel 3, “Waiting for the Great Leap Forward – Green Revolution and

Structural Change in the Philippines” (Väntan på det stora språnget framåt –

den gröna revolutionen och strukturomvandlingen på Filippinerna), studerar en

helt annorlunda slags intervention, den gröna revolutionen på Filippinerna. Me-

dan Koreas HCI genomfördes trots de västerländska institutionerna, så var den

gröna revolutionen deras idé. Som en produkt av Ford och Rockefeller Founda-

262 SAMMANFATTNING

tion bidrag, Robert McNamaras Världsbank och den filippinska regeringen,

grundades International Rice Research Institute (IRRI) i Lagunaprovinsen 1960.

IRRI utgjorde hjärtat i den agronomiska forskningen om nya slags rishybrider

som skulle komma att definiera den gröna revolutionen; statsvetaren Lynn T.

White hänvisade till den som “teknologiforskningsprogrammet med den högsta

profilen i världen” (White 2009). 1966 upplevde Filippinerna den omfattande

introduktionen av SK IR-8 “mirakelris”-sorter – den första avgörande produkten

från IRRI – vilket betecknar revolutionens början.

Följaktligen studerar detta kapitel effekterna av den gröna revolutionens

teknologier på strukturomvandlingar i deras ursprungsland. Alltsedan uppkoms-

ten av detta underområde har utvecklingsekonomer länge teoretiserat kring att

ökande produktivitet inom jordbruket utgjorde drivkraften bakom strukturom-

vandlingarna: omallokeringen av ekonomisk aktivitet från jordbrukssektorerna

till modern tillverkningsindustri och tjänstesektorer (Nurkse 1953, Rostow

1960). Emellertid så ledde den snabba, betydelsefulla utrullningen av tidig

teknologi relaterad till den gröna revolutionen inte till någon modernisering. I

skarp kontrast till andra samtida asiatiska länder så förblev andelen arbetskraft i

tillverkningsindustrin konstant och jordbrukssektorn förblev den dominerande

sysselsättningskällan under hela 1980-talet.

Genom att använda nyligen digitaliserade data för den gröna revolutionen

så visar jag att den snabba förändringen inom jordbruket medförde en struktur-

förändring – men på sätt som ej hade förutsetts av planerare och i teoretiska

modeller. Med en nyligen konstruerad panel av filippinska kommuner, följer

jag expansionen av nya sorter som ger stor avkastning, kända som HYVs eller

moderna sorter, ökande produktiviteten inom jordbruket och omfördelad ekono-

misk aktivitet mellan sektorer – mina mått på strukturomvandling. Jag fokuserar

särskilt på hur andelen sysselsatta inom jordbruket, tillverkningsindustrin och

tjänstesektorn förändrades under de nästföljande fyra decennierna, som följde

direkt på införandet av HYVs 1966.

Jag visar att de teknologiska chockerna från gröna revolutioner hade helt

olika effekter på kortsiktiga och långsiktiga strukturförändringar, vilket ska-

pade speciellt oväntade effekter på lantarbetares arbete inom jordbruket. Jag

bekräftar först att efter 1966, till skillnad från många asiatiska (och nuvarande

afrikanska samhällen), infördes HYVs allmänt i filippinska distrikt och för-

263

knippades sedan med en snabb ökning i produktiviteten inom jordbruket. Jag

visar sedan att på kort sikt, 1970-1980, så omvandlades den gröna revolutio-

nen till en arbetskraftsabsorberande teknologisk förändring: resulterande i en

omallokering av arbetskraft till HYV-intensiva risekonomier. Dessa resultat

överensstämmer med sysselsättningsökningen inom jordbruket under decenniet

efter införandet av nya moderna rissorter. På lång sikt, 1980-2000, visar jag

emellertid att mönstret är det omvända; den gröna revolutionen omvandlades

till en teknikförändring som kom att ersätta arbetskraft. Framför allt så försköts

lönearbetare från jordbruket och motades i riktning mot okvalificerade arbeten

inom tjänstesektorn. Jag hävdar att stigande löner och fallande priser på kapital

ledde till att risgårdarna mekaniserades och sålunda främjade den långsiktiga

sysselsättningsminskningen inom jordbruket.

Kapitel 4, “The Historical State, Local Collective Action, and Economic

Development in Vietnam” (Den historiska staten, lokal kollektiv handling och

ekonomisk utveckling i Vietnam), skrivet tillsammans med Melissa Dell och

Pablo Queurubin, utforskar de tydliga utvecklingsvägarna i Ostasien och Syd-

ostasien. Hur effektiva de avancerade utvecklingsstrategierna är beror på staters

kapacitet att organisera denna politik, ofta på lokala nivåer. Emellertid så varie-

rar staters möjlighet att sprida ambitiös utvecklingspolitik mycket inom Asien,

från weberianska byråkratier i Ostasien till personliga nätverk i Sydostasien.

De avancerade byråkratierna och planeringskontoren i Ostasien har tillskrivits

äran av framgången med det ostasiatiska tillväxtmiraklet (Evans 1992). Ex-

empelvis så tillskrevs Japans MITI (Johnson 1982) och Sydkoreas Economic

Planning Board (kapitel 2) äran av att implementera avancerad industriell poli-

tik. Liknande initiativ i Sydostasiatisk politik drabbades av kompiskapitalism

och girighetspolitik – kanske främst kännetecknad av Macros New Society.

Kapitel 4 studerar den historiska roll som statliga institutioner i Ostasien kontra

Sydostasien spelade för de olika erfarenheterna i de två grannregionerna.

Specifikt så studerar kapitel 4 vilken effekt den historiska staten har på

långsiktig utveckling och använder Vietnam som ett laboratorium för erfaren-

heterna i Ostasien och Sydostasien. Som hävdas av en lång rad historiker och

antropologer, så utgjorde det tidiga moderna Vietnam skiljelinjen mellan de två

civilisationerna (Lieberman 2003). Nordvietnam, historiskt känt som Dai Viet,

styrdes av en stark centraliserad stat där byn var den fundamentala administrati-

264 SAMMANFATTNING

va enheten. Dessa institutioner var direkt tagna från det kejserliga Kina, en av

de tidigaste moderna staterna i världen. Å andra sidan var Sydvietnam ett yttre

lydrike till khmerernas (Kambodjas) imperium, vilket följde en vasallmodell

med svagare, mer personliga maktrelationer och ingen bymedling.

Genom att använda en regressionsanalys som utnyttjar en diskontinuitet i

data över den historiska Dai Viet-Khmer gränsen visar denna studie att områden

som historiskt legat under en stark stat har högre levnadsstandard idag och

bättre ekonomiskt utfall under de senaste 150 åren. Rika historiska data doku-

menterar att i byar med en stark historisk stat har medborgarna varit bättre på

att organisera för allmännytta och omfördelning genom civilsamhället och lokal

styrning. Detta tyder på att den starka historiska staten rörde sig i gemensam

aktion på bynivå och att dessa normer kvarstod långt efter att den ursprungliga

staten försvann. Vi anser att erfarenheten av utvecklingen i dessa två världar

utgör en större analogi för de olika erfarenheterna i Ostasien och Sydostasien.

265

Referenser








Evans, P. B. (1992). The State as Problem and Solution: Predation, Embedded

Autonomy, and Structural Change. In S. Haggard and R. R. Kaufman (Eds.),

The Politics of Economic Adjustment: International Constraints, Distributive

Conflicts and the State, Chapter 3, pp. 139–181. Princeton, New Jersey:

Princeton University Press.

Hirschman, A. O. (1958). The Strategy of Economic Development (Third ed.).

New Haven, Connecticut: Yale University Press.

Hutchcroft, P. (2011). Reflections on a Reverse Image. In The Park Chung Hee

Era, Chapter 19, pp. 542‚Äì572. Cambridge, Massachusetts and London,





Korea and the Philippines (First ed.). Cambridge Studies in Comparative

Politics. Cambridge, U.K.: Cambridge University Press.


800-1830, Volume I: Integration on the Mainland. Cambridge, U.K and New

York, New York: Cambridge University Press.

Nelson, R. R. and H. Pack (1998). The Asian Miracle and Modern Growth

Theory. The Economic Journal 109(1881), 416–436.

266 SAMMANFATTNING



Poppel, Z. D. (2015). Quick rice: international development and the Green Re-

volution in Sierra Leone, 1960-1976. In C. Helstosky (Ed.), The Routledge

History of Food, Chapter 17, pp. 332–354. New York, New York: Routledge

and Taylor & Francis Group.



Scott, J. C. (1998). Seeing Like a State: How Certain Schemes to Improve the

Human Condition Have Failed. New Haven, Connecticut: Yale University

Press.

White, L. T. (2009). Political Booms: Local Money and Power in Taiwan,

East China, Thailand, and the Philippines. Series on Contemporary China –


MONOGRAPH SERIES

1. Michaely, Michael The Theory of Commercial Policy: Trade and Protec-tion, 1973

2. Söderström, Hans Tson Studies in the Microdynamics of Production andProductivity Change, 1974

3. Hamilton, Carl B. Project Analysis in the Rural Sector with SpecialReference to the Evaluation of Labour Cost, 1974

4. Nyberg, Lars and Staffan Viotti A Control Systems Approach to Macroe-conomic Theory and Policy in the Open Economy, 1975

5. Myhrman, Johan Monetary Policy in Open Economies, 1975

6. Krauss, Melvyn International Trade and Economic Welfare, 1975

7. Wihlborg, Clas Capital Market Integration and Monetary Policy underDifferent Exchange Rate Regimes, 1976

8. Svensson, Lars E.O. On Competitive Markets and Intertemporal Re-sources Allocation, 1976

9. Yeats, Alexander J. Trade Barriers Facing Developing Countries, 1978

10. Calmfors, Lars Prices, Wages and Employment in the Open Economy,1978

11. Kornai, Janos Economics of Shortage, Vols I and II, 1979

12. Flam, Harry Growth, Allocation and Trade in Sweden. An EmpiricalApplication of the Heckscher-Ohlin Theory, 1981

13. Persson, Torsten Studies of Alternative Exchange Rate Systems. AnIntertemporal General Equilibrium Approach, 1982

14. Erzan, Refik Turkey’s Comparative Advantage, Production and TradePatterns in Manufactures. An Application of the Factor ProportionsHypothesis with Some Qualifications, 1983

15. Horn af Rantzien, Henrik Imperfect Competition in Models of WageFormation and International Trade, 1983

16. Nandakumar, Parameswar Macroeconomic Effects of Supply Side Policiesand Disturbances in Open Economies, 1985

267

268 MONOGRAPH SERIES

17. Sellin, Peter Asset Pricing and Portfolio Choice with International In-vestment Barriers, 1990

18. Werner, Ingrid International Capital Markets: Controls, Taxes and Re-sources Allocation, 1990

19. Svedberg, Peter Poverty and Undernutrition in Sub-Saharan Africa: The-ory, Evidence, Policy, 1991

20. Nordström, Håkan Studies in Trade Policy and Economic Growth, 1992

21. Hassler, John, Lundvik, Petter, Persson, Torsten and Söderlind, Paul TheSwedish Business Cycle: Stylized facts over 130 years, 1992

22. Lundvik, Petter Business Cycles and Growth, 1992

23. Söderlind, Paul Essays in Exchange Rates, Business Cycles and Growth,1993

24. Hassler, John A.A. Effects of Variations in Risk on Demand and Measuresof Business Cycle Comovements, 1994

25. Daltung, Sonja Risk, Efficiency, and Regulation of Banks, 1994

26. Lindberg, Hans Exchange Rates: Target Zones, Interventions and RegimeCollapses, 1994

27. Stennek, Johan Essays on Information-Processing and Competition, 1994

28. Jonsson, Gunnar Institutions and Incentives in Monetary and FiscalPolicy, 1995

29. Dahlquist, Magnus Essays on the Term Structure of Interest Rates andMonetary Policy, 1995

30. Svensson, Jakob Political Economy and Macroeconomics: On ForeignAid and Development, 1996

31. Blix, Mårten Rational Expectations and Regime Shifts in Macroecono-metrics, 1997

32. Lagerlöf, Nils-Petter Intergenerational Transfers and Altruism, 1997

33. Klein, Paul Papers on the Macroeconomics of Fiscal Policy, 1997

34. Jonsson, Magnus Studies in Business Cycles, 1997

35. Persson, Lars Asset Ownership in Imperfectly Competitive Markets, 1998

36. Persson, Joakim Essays on Economic Growth, 1998

37. Domeij, David Essays on Optimal Taxation and Indeterminacy, 1998

38. Floden, Martin Essays on Dynamic Macroeconomics, 1999

39. Tangerås, Thomas Essays in Economics and Politics: Regulation, Elec-tions and International Conflict, 2000

269

40. Lidbom, Per Pettersson Elections, Party Politics and Economic Policy,2000

41. Vestin, David Essays on Monetary Policy, 2001

42. Olofsgård, Anders Essays on Interregional and International PoliticalEconomics, 2001

43. Johansson, åsa Essays on Macroeconomic Fluctuations and NominalWage Rigidity, 2002

44. Groth, Charlotta Topics on Monetary Policy, 2002

45. Gancia, Gino A. Essays on Growth, Trade and Inequality, 2003

46. Harstad, Bård Organizing Cooperation: Bargaining, Voting and Control,2003

47. Kohlscheen, Emanuel Essays on Debts and Constitutions, 2004

48. Olovsson, Conny Essays on Dynamic Macroeconomics, 2004

49. Stavlöt, Ulrika Essays on Culture and Trade, 2005

50. Herzing, Mathias Essays on Uncertainty and Escape in Trade Agreements,2005

51. Bonfiglioli, Alessandra Essays on Financial Markets and Macroeco-nomics, 2005

52. Pienaar, Natalie Economic Applications of Product Quality Regulation inWTO Trade Agreements, 2005

53. Song, Zheng Essays on Dynamic Political Economy, 2005

54. Eisensee, Thomas Essays on Public Finance: Retirement Behavior andDisaster Relief, 2005

55. Favara, Giovanni Credit and Finance in the Macroeconomy, 2006

56. Björkman, Martina Essays on Empirical Development Economics: Edu-cation, Health and Gender, 2006

57. Larsson, Anna Real Effects of Monetary Regimes, 2007

58. Prado, Jr., Jose Mauricio Essays on Public Macroeconomic Policy, 2007

59. Tonin, Mirco Essays on Labor Market Structures and Policies, 2007

60. Queijo von Heideken, Virgina Essays on Monetary Policy and AssetMarkets, 2007

61. Finocchiaro, Daria Essays on Macroeconomics, 2007

62. Waisman, Gisela Essays on Discrimination and Corruption, 2008

63. Holte, Martin Bech Essays on Incentives and Leadership, 2008

270 MONOGRAPH SERIES

64. Damsgaard, Erika Färnstrand Essays on Technological Choice andSpillovers, 2008

65. Fredriksson, Anders Bureaucracy, Informality and Taxation: Essays inDevelopment Economics and Public Finance, 2009

66. Folke, Olle Parties, Power and Patronage: Papers in Political Economics,2010

67. Drott, David Yanagizawa Information, Markets and Conflict: Essays onDevelopment and Political Economics, 2010

68. Meyersson, Erik Religion, Politics and Development: Essays in Develop-ment and Political Economics, 2010

69. Klingelhöfer, Jan Models of Electoral Competition: Three Essays onPolitical Economics, 2010

70. Perrotta, Maria Carmela Aid, Education and Development, 2010

71. Caldara, Dario Essays on Empirical Macroeconomics, 2011

72. Mueller, Andreas Business Cycles, Unemployment and Job Search: Es-says in Macroeconomics and Labor Economics, 2011

73. Von Below, David Essays in Climate and Labor Economics,2011

74. Gars, Johan Essays on the Macroeconomics of Climate Change, 2012

75. Spiro, Daniel Some Aspects on Resource and Behavioral Economics,2012

76. Ge, Jinfeng Essays on Macroeconomics and Political Economy, 2012

77. Li, Yinan Institutions, Political Cycles and Corruption: Essays on Dy-namic Political Economy of Government, 2013

78. Håkanson, Christina Changes in Workplaces and Careers, 2013

79. Qin, Bei Essays on Empirical Development and Political Economics,2013

80. Jia, Ruixue Essays on the Political Economy of China’s Development,2013

81. Campa, Pamela Media Influence on Pollution, and Gender Equality, 2013

82. Seim, David Essays on Public, Political and Labor Economics, 2013

83. Shifa, Abdulaziz Essays on Growth, Political Economy and Development,2013

84. Panetti, Ettore Essays on the Economics of Banks and Markets, 2013

85. Schmitt, Alex Beyond Pigou: Climate Change Mitigation, Policy Makingand Distortions, 2014

271

86. Rogall, Thorsten The Economics of Genocide and War, 2015

87. Baltrunaite, Auding Political Economics of Special Interests and Gender,2016

88. Harbo Hansen, Niels-Jakob Jobs, Unemployment, and MacroeconomicTransmission, 2016

89. Stryjan, Miri Essays on Development Policy and the Political Economyof Conflict, 2016

90. Karadja, Mounir On the Economics and Politics of Mobility, 2016

91. Kitamura, Shuhei Land, Power and Technology: Essays on PoliticalEconomy and Historical Development, 2016

92. Malmberg, Hannes Human Capital in Development Accounting andOther Essays in Economics, 2017

93. Öberg, Erik On Money and Consumption, 2017

diva portalsu.diva-portal.org/smash/get/diva2:1115208/fulltext01.pdfeast and southeast asia by...

Documents