diva portalsu.diva-portal.org/smash/get/diva2:1115208/fulltext01.pdfeast and southeast asia by...
TRANSCRIPT
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-
12 CHAPTER 2. MANUFACTURING REVOLUTIONS
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,
2.1. INTRODUCTION 13
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.
14 CHAPTER 2. MANUFACTURING REVOLUTIONS
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].
16 CHAPTER 2. MANUFACTURING REVOLUTIONS
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 .
2.2. INSTITUTIONAL CONTEXT 17
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.
18 CHAPTER 2. MANUFACTURING REVOLUTIONS
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. INSTITUTIONAL CONTEXT 19
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].
20 CHAPTER 2. MANUFACTURING REVOLUTIONS
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
2.2. INSTITUTIONAL CONTEXT 21
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].
22 CHAPTER 2. MANUFACTURING REVOLUTIONS
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
2.2. INSTITUTIONAL CONTEXT 25
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].
26 CHAPTER 2. MANUFACTURING REVOLUTIONS
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.
28 CHAPTER 2. MANUFACTURING REVOLUTIONS
(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).
30 CHAPTER 2. MANUFACTURING REVOLUTIONS
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.
32 CHAPTER 2. MANUFACTURING REVOLUTIONS
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.
34 CHAPTER 2. MANUFACTURING REVOLUTIONS
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].
36 CHAPTER 2. MANUFACTURING REVOLUTIONS
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.
2.5. DIRECT EFFECTS OF INDUSTRIAL POLICY 37
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).
38 CHAPTER 2. MANUFACTURING REVOLUTIONS
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
2.5. DIRECT EFFECTS OF INDUSTRIAL POLICY 39
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).
42 CHAPTER 2. MANUFACTURING REVOLUTIONS
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.
44 CHAPTER 2. MANUFACTURING REVOLUTIONS
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
2.5. DIRECT EFFECTS OF INDUSTRIAL POLICY 45
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).
50 CHAPTER 2. MANUFACTURING REVOLUTIONS
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-
52 CHAPTER 2. MANUFACTURING REVOLUTIONS
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.
2.6. NETWORK EXTERNALITIES 53
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 .
54 CHAPTER 2. MANUFACTURING REVOLUTIONS
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.
2.6. NETWORK EXTERNALITIES 55
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:
2.6. NETWORK EXTERNALITIES 57
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.
2.6. NETWORK EXTERNALITIES 59
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.
60 CHAPTER 2. MANUFACTURING REVOLUTIONS
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),
62 CHAPTER 2. MANUFACTURING REVOLUTIONS
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
64 CHAPTER 2. MANUFACTURING REVOLUTIONS
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
68 CHAPTER 2. MANUFACTURING REVOLUTIONS
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.
70 CHAPTER 2. MANUFACTURING REVOLUTIONS
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
72 CHAPTER 2. MANUFACTURING REVOLUTIONS
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-
74 CHAPTER 2. MANUFACTURING REVOLUTIONS
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
76 CHAPTER 2. MANUFACTURING REVOLUTIONS
Frictions. Econometrica, (Forthcoming).
Janne E. Nolan. (1986). Military Industry in Taiwan and South Korea. London,
England: Palgrave Macmillan UK.
Johnson, C. (1982). Miti and the Japanese Miracle: The Growth of Industrial
Policy : 1925-1975. Stanford, California: Stanford University Press.
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.
Kang, D. C. (2002). Crony Capitalism: Corruption and Development in South
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-
78 CHAPTER 2. MANUFACTURING REVOLUTIONS
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;
New York, New York: Cambridge University Press.
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
80 CHAPTER 2. MANUFACTURING REVOLUTIONS
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,
England: Harvard University Press.
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
82 CHAPTER 2. MANUFACTURING REVOLUTIONS
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:
84 CHAPTER 2. MANUFACTURING REVOLUTIONS
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,
England: Harvard University Press.
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-
86 CHAPTER 2. MANUFACTURING REVOLUTIONS
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-
88 CHAPTER 2. MANUFACTURING REVOLUTIONS
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
90 CHAPTER 2. MANUFACTURING REVOLUTIONS
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.
92 CHAPTER 2. MANUFACTURING REVOLUTIONS
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.
94 CHAPTER 2. MANUFACTURING REVOLUTIONS
Table 2.6: Differences in Industrial Growth Relative to 1972, 1970-1986
Dependent Variable (IHS) :
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.
96 CHAPTER 2. MANUFACTURING REVOLUTIONS
Table 2.8: Differences in Labor Productivity, Before-After 1973, 1970-1986
Dependent Variable (IHS) :
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
Baseline Controls 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.
98 CHAPTER 2. MANUFACTURING REVOLUTIONS
Table 2.10: Differences in Exports and Imports, Before-After 1973, 1970-1986
Dependent Variable (IHS) :
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)
Industry Fixed Effects X X X X X X
Year Fixed Effects X X X X X X
Baseline Controls X X X X
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)
Industry Fixed Effects X X X X X X
Year Fixed Effects X X X X X X
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.
100 CHAPTER 2. MANUFACTURING REVOLUTIONS
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)
Industry Fixed Effects X X X X X X
Year Fixed Effects X X X X X X
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
102 CHAPTER 2. MANUFACTURING REVOLUTIONS
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
Dependent Variable (IHS) :
(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.
104 CHAPTER 2. MANUFACTURING REVOLUTIONS
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
3.1. INTRODUCTION 107
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).
108 Waiting for the Great Leap Forward
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).
110 Waiting for the Great Leap Forward
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.
112 Waiting for the Great Leap Forward
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.
114 Waiting for the Great Leap Forward
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
116 Waiting for the Great Leap Forward
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
118 Waiting for the Great Leap Forward
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.
122 Waiting for the Great Leap Forward
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.
126 Waiting for the Great Leap Forward
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.
128 Waiting for the Great Leap Forward
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.
130 Waiting for the Great Leap Forward
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..
132 Waiting for the Great Leap Forward
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
134 Waiting for the Great Leap Forward
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).
136 Waiting for the Great Leap Forward
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.
138 Waiting for the Great Leap Forward
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., 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.
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.
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.
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
Rice Research Institute.
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:‘forward
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.
Nurkse, R. (1953). Problems of Capital Formation in Underdeveloped Coun-
tries. Oxford, UK: Oxford University Press.
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.
Rostow, W. W. (1960). The Stages of Economic Growth: A Non-Communist
Manifesto (3rd ed.). New York, New York: Cambridge University Press.
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 –
Vol.16. Singapore: World Scientific Publishing.
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.
4.1. INTRODUCTION 155
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
156 The Historical State in Vietnam
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
4.1. INTRODUCTION 157
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)
158 The Historical State in Vietnam
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).
4.1. INTRODUCTION 159
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.
160 The Historical State in Vietnam
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
162 The Historical State in Vietnam
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
4.2. HISTORICAL BACKGROUND 163
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
164 The Historical State in Vietnam
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).
4.2. HISTORICAL BACKGROUND 165
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
166 The Historical State in Vietnam
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
168 The Historical State in Vietnam
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
170 The Historical State in Vietnam
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.
4.4. LONG RUN EFFECTS ON ECONOMIC PROSPERITY 171
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)
172 The Historical State in Vietnam
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
4.4. LONG RUN EFFECTS ON ECONOMIC PROSPERITY 173
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.
174 The Historical State in Vietnam
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
4.4. LONG RUN EFFECTS ON ECONOMIC PROSPERITY 175
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.
176 The Historical State in Vietnam
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.
178 The Historical State in Vietnam
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.
180 The Historical State in Vietnam
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).
182 The Historical State in Vietnam
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
184 The Historical State in Vietnam
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),
186 The Historical State in Vietnam
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.
Lieberman, V. (2003). Strange Parallels: Southeast Asia in Global Context, c.
REFERENCES 193
800-1830, Volume I: Integration on the Mainland. Cambridge, U.K and
New York, New York: Cambridge University Press.
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, 1757–1954. 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,
National Bureau of Economic Research.
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,
National Bureau of Economic Research.
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
Dependent variable is:
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
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).
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
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).
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
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).
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
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).
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
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).
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
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).
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
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).
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
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).
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
0 .2 .4 .6 .8 1civil society
(b) Local Admin Index andCivil Society Index
.2.4
.6.8
1ed
ucat
ion
0 .2 .4 .6 .8 1civil society
(c) Education Index and CivilSociety Index
.2.4
.6.8
1he
alth
0 .2 .4 .6 .8 1civil society
(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
Dependent variable is:
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
Dependent variable is:
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 0.310 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
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.
250 REFERENCES
Table A.27: Current Outcomes: No Ho Chi Minh Province
Dependent variable is:
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 0.060 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
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.
A. APPENDIX 251
Table A.28: Current Outcomes: No Provincial Capitals
Dependent variable is:
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 0.132 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
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.
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
Dependent variable is:
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
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.
266 SAMMANFATTNING
Nurkse, R. (1953). Problems of Capital Formation in Underdeveloped Coun-
tries. Oxford, UK: Oxford University Press.
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.
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.
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