Following the Herd and Sleeping with the Enemy:

Strategies for Managing Policy Uncertainty

WITOLD J. HENISZ Department of Management

The Wharton School 2021 Steinberg Hall-Dietrich Hall

University of Pennsylvania Philadelphia, PA 19104-6370

Tel: (215) 898-0788 Fax: (215) 898-0401

E-mail: henisz@wharton.upenn.edu

ANDREW DELIOS Department of Business Policy

National University of Singapore 1 Business Link, 117592

SINGAPORE Tel: 65-6874-3094 Fax: 65-6775-5059

Email: andrew@nus.edu.sg

March 31, 2004

* This research was supported by a Social Sciences and Humanities Research Council of Canada Grant (#410-2001-0143) and an NUS Academic Research Grant (R-313-000-052-112). Thanks to Ron Adner, Kimberly Bates, Joel Baum, Heather Berry, Emilio Castilla, Wilbur Chung, David Collis, Javier Gimeno, Mauro Guillén, Tarun Khanna, Bruce Kogut, Xavier Martin, Patrick Moreton, Lori Rosenkopf, Brian Silverman, Gabriel Szulanski and participants of the 2003 Strategic Research Forum for comments on previous drafts.

2

Following the Herd and Sleeping with the Enemy: Strategies for Managing Policy Uncertainty

Abstract: We examine the extent to which two sources of policy uncertainty influence foreign-owned subsidiary exit rates. We find that prior peer exits and a firm’s own experience under the current political regime both have a strong influence on subsidiary exit rates, particularly in the presence of policy uncertainty resulting from the existing structure of a host country’s political institutions. A firm's own experience under the current political regime of a host country, however, enhances exit rates after changes in that regime. These findings point to tradeoffs between two strategies for moderating political uncertainty. Influence strategies can provide substantial gains to firms in uncertain policy environments so long as those environments themselves are not at risk of radical flux. As the probability that a policy environment could potentially enter a period of radical flux increases, a follow-the-herd strategy becomes more prominent as this strategy avoids the negative consequences of the rapid depreciation of the value of past organizational experience, in the event of a new political regime.

3

Uncertainty is a prominently featured construct in neoinstitutional research and international

business. A primary focus of neoinstitutional research is the mimetic response of an organization

to cues in its inter-organizational environment under conditions of uncertainty (Haunschild,

1994, Haunschild & Miner, 1997, Haveman, 1993). A parallel line of research in international

business examines how the accumulation of organizational experience reduces a firm’s exposure

to the potentially adverse influences of uncertainty (Barkema, Bell, & Pennings, 1996, Delios &

Henisz, 2003, Henisz & Delios, 2001).

We seek to integrate these two bodies of research by first unpacking the construct of uncertainty

to see whether uncertainty emerges from a lack of information about an external environment, or

from an underlying dynamism in that external environment (Duncan, 1972). We then examine

the conditions under which either a mimetic response to stimuli in the inter-organizational

environment, or a reliance upon organizational experience, is an effective strategy in the face of

these sources of uncertainty.

In making this examination, we highlight the idea that organizational experience can generate

signals and pressures for managers who seek to fashion strategic responses to uncertainty. This

experience, however, has its greatest utility for addressing uncertainty that stems from

information shortfalls, in stable environments. In sharp contrast to its positive effect in stable

environments, experience can become a liability for addressing information shortfalls, in a

dynamic or radically changed external environment. Meanwhile, the signals and pressures

generated by the inter-organizational environment, although less helpful than a firm’s own

experience in stable environments, do not transform into liabilities in dynamic or radically

changed external environments.

4

To test these ideas, we focus on a specific dimension of a firm’s external environment, namely

the political environment. We look at two sources of uncertainty in the policies generated by

that political environment. The first source of policy uncertainty, political hazards, stems from

the structure of a nation’s existing political institutions that either provides substantial discretion

to policymakers or limits their discretion. The second, regime instability, takes the form of

widespread protests or concerted efforts to overthrow a nation’s existing political institutions.

Our empirical models examine the exit rates of 2,283 international expansions made by 642

Japanese manufacturing firms into 53 countries that differ in their levels of political hazards and

regime instability.

BACKGROUND

Environmental Uncertainty

Environmental uncertainty is among the foremost challenges with which an organization’s

management must contend (Thompson, 1967). A core tenet emerging from research in this area

is that organizations seek to avoid uncertainty (Cyert & March, 1963). The pervasiveness of the

phenomenon of uncertainty led to the extensive use of the aggregate construct of environmental

uncertainty in early studies on organizations (Downey & Slocum, 1975). This research pointed

to the idea that organizations differ in their reactions, and susceptibility, to environmental

uncertainty (DiMaggio & Powell, 1983 , Levitt & March, 1988). Organizational variance in

susceptibility to uncertainty stems from heterogeneous organizational resources (Miner,

Amburgey, & Stearns, 1990) and heterogeneous organizational information.

Research on organizations has begun to emphasize how different organizational information

sources can help to mitigate the uncertainty a firm encounters in its external environments

5

(Haunschild, 1994, Haunschild & Miner, 1997), but it has yet to explore variations in the types

of uncertainty a firm faces in its environments, when making market entry decisions (Henisz &

Delios, 2001). The benefits of unpacking uncertainty can be profound if researchers strive for a

better understanding of the relative efficacy of a particular organizational response to

environmental uncertainty.

The idea of examining different types of uncertainty is not particularly novel as there is a long

tradition in organization theory research that has placed the construct of uncertainty at its focus

(Boyd, Dess, & Rasheed, 1993, Dess & Beard, 1984, Milliken, 1987). Aside from identifying

the importance of uncertainty in the external environment on decision-making in a firm, research

on organizations has helped to identify several definitional characteristics of perceived

uncertainty (Boyd, Dess, & Rasheed, 1993). Uncertainty concerns the lack of information in an

environment, given a specific decision-making scenario. This lack of information translates into

difficulties in accurately assessing the losses or gains associated with correct or incorrect

decisions (Duncan, 1972). Uncertainty also concerns rates and unpredictability of change in an

environment (Dess & Beard, 1984). As the results of dynamic change are hard to predict, it

heightens uncertainty for decision makers.

These characterizations of environmental uncertainty highlight two points of relevance for

differentiating among types of uncertainty. First, uncertainty can emerge from a lack of

information about a particular environment. Second, uncertainty can arise from instability in that

environment. If we examine organizational responses to these two broad types of uncertainty,

we can explore the relative efficacy of two key organizational responses to each type. In the first

organizational response, information derived from an organization’s experience in an

environment has been shown to reduce the influence of uncertainty in that environment for that

6

organization (Barkema, Bell, & Pennings, 1996, Delios & Henisz, 2000, Lu, 2002). A second

organizational response is grounded in research in neoinstitutional theory. Neoinstitutional

scholars emphasize that managers can seek cues for responses to uncertainty from the inter-

organizational environment, or the actions of a reference group of organizations (Haunschild,

1994, Haunschild & Miner, 1997, Haveman, 1993). These cues lead firms to imitate the

decisions, strategies and structures of one another, and the prevalence of mimetic behavior tends

to increase in the level of environmental uncertainty (Haunschild & Miner, 1997). In our

hypotheses, we discuss how organizational responses to uncertainty can vary depending on

whether the uncertainty comes from a lack of information about an environment, or from

environmental dynamism.

Two Sources of Policy Uncertainty

The international environment is a setting with copious sources of uncertainty. Scholars have

long recognized that cultural, economic, legal, social and political variance by nation injects

substantial amounts of uncertainty in a firm’s decisions when it undertakes an international

expansion (Barkema, Bell, & Pennings, 1996, Eriksson, Johanson, Majkgard, & Sharma, 1997,

Henisz, 2000b). An international expansion is a form of organizational growth in which a firm

undertakes a foreign direct investment to form a foreign subsidiary in a country other than the

one in which it is domiciled.

When expanding internationally, a firm’s management often must contend with a new culture, a

new language, a new social system, new market structures, and a new political system. To

manage an expansion in a new political environment, a firm’s management should understand

the current and future policies that define the rules and regulations for a firm’s operations (e.g.,

7

taxes or quantitative restrictions on production, foreign trade and employment and any

exemptions thereto; regulations in the realm of employee hiring and firing, benefits, health and

safety and the local or national environment) and thereby influence its profitability.

The status quo policies and any changes therein are the outcome of a policymaking process in

which various interest groups seek to lobby or influence policymakers who interact within

formal or informal policymaking structures. The level of uncertainty over future policies is

therefore a function of both the current policymaking structure (i.e., the extent to which it

provides checks and balances against the discretion of existing policymakers) and the stability of

the existing political structure.

Where political institutions include many checks and balances, such as among the executive,

legislative, judicial and sub-federal branches, policymakers are constrained in their choice of

policies by other political actors who must approve any change from the status quo policy

(Henisz, 2000a, North, 1990). Countries with extensive checks and balances in the formal

policymaking apparatus, such as the United States, Germany and Switzerland, tend to have the

lowest levels of policy uncertainty as the multiple veto players involved in the policy making

process, find it difficult to agree to change the status quo policy. Policy uncertainty increases as

the number of veto players declines or as their preferences become more homogeneous, such as

is the case in moving to a mixed Parliamentary-Presidential system, as typified by France or

Brazil, to heavily fractionalized Parliamentary systems like Belgium, Israel or the Netherlands to

Westminster Parliamentary systems with winner-take-all districts like the United Kingdom.

Transitional and non-democratic states have the highest levels of policy uncertainty as the formal

institutional structure in these states provides tremendous discretion to policymakers.

8

A second source of policy uncertainty is the threat of the emergence of a new political regime.

Where disaffected interest groups are militating for fundamental political reform, the very rules

by which policies are made are called into question. The identity of the future political

leadership, the process of policymaking and the interest groups allotted voice or excluded from

the political process are each uncertain when there are calls by interest groups for radical

political change. Any of these changes would increase the level of policy uncertainty and

simultaneous changes in all would have an even greater effect.

The Relative Efficacy of Two Strategies for Reacting to Policy Uncertainty

Given that increases in policy uncertainty generate corresponding increases in the variance of a

subsidiary’s expected profitability, effective strategies for reacting to policy uncertainty can

generate substantial value for a firm. Firms seek to identify a country’s level of policy

uncertainty prior to investing and will only enter a country where the expected returns justify the

additional variance in those returns. One would not therefore expect an independent effect of

policy uncertainty on subsidiary performance. The exit rates of subsidiaries in a country with

high policy uncertainty, however, may vary according to the efficacy of the strategies that these

subsidiaries take to moderate the effect of policy uncertainty on their operations particularly

where those strategies involve tradeoffs that can not readily be assessed at the time of entry.

A basic strategy for a firm to pursue to moderate future policy uncertainty is to devote resources

to the identification of political risks and opportunities and to the influence of the policymaking

process. If a firm pursues this strategy, it needs to identify the pivotal political actors that

influence the policies of interest to them (Holburn & Vanden Bergh, 2002, Krehbiel, 1999) and

to develop influence strategies to guard against inimical policy change by those actors that serve

9

the interests of the broader polity, their buyers or suppliers, their competitors or political actors at

the firms’ expense (Henisz & Zelner, 2004). These influence strategies include the development

of or participation in a lobbying coalition to deliver a message to these pivotal actors and the

tailoring of that message to maximize the likelihood that those who receive the message will act

upon it.

The magnitude of a firm’s resource commitments to these influence strategies depends upon

their relative efficacy in reducing policy uncertainty. As the checks and balances within a

country’s political institutions diminish (i.e., the political hazards increase), policymakers’

discretion increases and influence strategies can offer more substantive reductions in policy

uncertainty.

The nature of the lobbying coalition and the message that they deliver to that pivotal actor is also

a function of the pivotal actor’s preferences and position in the institutional structure. Changes

in the political regime (Feng, 2001, Kobrin, 1979) which alter the interest groups represented in

government and the nature of interactions among political institutions may therefore reduce the

value of past investments in cultivating contacts, developing lobbying coalitions and delivering

messages, for reducing uncertainties about future policies in a given nation. Given the potential

depreciation of past investments and the large fixed costs to begin anew under a new political

regime, the efficacy of influence strategies is likely a decreasing function of regime instability.

A second strategy available to managers facing policy uncertainty, is to examine the strategic

decisions of their peers and infer unobservable calculations on the expected level of profitability

based on the observable entry and exit decisions of peers (Baum, Li, & Usher, 2000, Miner &

Haunschild, 1995). Based on their assessment of the accuracy of the information signal in these

10

observations, managers could then update their prior beliefs regarding profitability levels.

Managers could also conclude that despite any internal reservations they had about the strategies

adopted by their peers, it could possibly pay to imitate their peers decisions rather than deviate

from these decisions (Abrahamson & Rosenkopf, 1993). Finally, managers could eschew

individual analysis of the policy environment or augment these analyses by imitating the

strategic decisions of peers whose actions are perceived as legitimate (DiMaggio & Powell,

1983, Haunschild & Miner, 1997), by following rules-of-thumb (March, 1988) or by simply

following habits (Geertz, 1978).

Although directly observing a firm’s investments in political risk identification and management

is difficult, it is possible to link observed variation in organizational performance under varying

degrees of policy uncertainty to organizational characteristics that are associated with the

adoption of these two strategies for managing uncertainty in the political environment. In our

hypotheses, we will argue that firms with greater levels of organizational experience under the

current political regime have an advantage over competitors in mounting an influence strategy.

This advantage should be manifest in better performance, which we measure by subsidiary exit

rates, in political environments where such strategies are likely to be more efficacious, that is,

where policy uncertainty that derives from political hazards is high. In contrast, these same firms

should have higher subsidiary exit rates in political environments where such strategies are likely

to be less efficacious. We also directly examine the strategy of imitating peer exit under policy

uncertainty of various types.

11

HYPOTHESES

Organizational Experience and Influence Strategies

A manager’s own experiences condition responses to policy uncertainty emanating from political

hazards. This influence emerges from the process by which heterogeneous experience profiles

help generate differential capabilities across organizations to identify and mitigate uncertainty

that emanates from the political environment. These capabilities stem from routine-based

systems that accumulate over the course of experiential learning (Cyert & March, 1963,

Levinthal & March, 1981).

An organization’s routines are its encoded actions that encapsulate the knowledge and

capabilities developed in its experiential learning (Nelson & Winter, 1982). Routines develop

through an incremental process, in which an organization’s existing routines are updated based

upon how managers interpret signals generated by their current decisions and associated

outcomes. Chang (1995) and Pennings, Barkema and Douma (1994) provide empirical evidence

consistent with such learning processes in their respective studies of expansion into new

geographic and product markets. Similarly, Delios and Henisz (2000) and Barkema, Bell and

Pennings (1996) find evidence that firms with extensive experience are better able to manage

relationships with joint venture partners than inexperienced firms, even in the presence of policy

or cultural uncertainty.

Experiential learning that can lead to the development of appropriate routines can enhance the

efficacy of organizational strategies in the face of uncertainty emanating from a lack of

information about an environment. In the case of policy uncertainty that comes from political

hazards, prior experience in a given policymaking structure provides information that helps

12

managers to improve their forecasts of future policies. Managers with relevant experience in a

given policy setting can develop a good understanding of the nature of coalitions that could be

formed, the preferences of policymakers, the compromises that a given political party or interest

group are willing to make, the length of time it will take for a policy innovation to be enacted or

other context-specific information that facilitates the delimitation of a choice set for their

analysis (Geertz, 1978). Managers can use this information to identify those actors for whom a

shift in preferences will generate the largest change in the final policy outcomes and focus their

lobbying or informational strategies upon those pivotal actors (Holburn & Vanden Bergh, 2002).

These actors may include politicians, interest groups or alliance partners with strong political

connections (Hitt, Dacin, Levitas, Arregle, & Borza, 2000). Not only will prior experience

facilitate the identification of these pivotal actors but it also facilitates the identification and

implementation of an effective lobbying strategy (Henisz & Zelner, 2004).

A lobbying strategy is undertaken in a context in which cognitively limited and time-constrained

political actors are typically overburdened with multiple appeals for assistance from conflicting

interest groups. These actors must balance whatever desires they have to serve the public

welfare of their constituents or their nation, with the burdens of maintaining office through

soliciting campaign contributions, votes or interest group support (Kingdon, 1984). As a result,

political actors respond more strongly to appeals that are closely linked to large blocks of

political or financial support (Hilgartner & Bosk, 1988). Such appeals are rarely esoteric or

extremely technical in nature but rather appeal to broad-based beliefs or fit within existing

“master frames” (Benford & Snow, 2000) that conform to preexisting heuristics, beliefs and

biases (Hilgartner & Bosk, 1988). Frames allow for the simplification of complex debate into a

limited set of readily identifiable dimensions that reduce the burden for participation and

13

decision (Benford & Snow, 2000). For example, lobbying on the rate of telecommunications

interconnection based on technical arguments surrounding actual costs or opportunity costs is

less likely to be effective than appeals for free and fair competition. The latter “frame” appeals to

an existing heuristic instantly recognizable by actors who have no inherent preference on

interconnection pricing and no desire to enmesh themselves in the intricacies of the debate.

Experiential learning within a given structure of political institutions can thus aid an organization

grappling with policy uncertainty emanating from that structure by reducing its lack of

information that is the foundation for this form of uncertainty, and by extending its competence

base as captured in routines that identify likely future policies, the pivotal actors that determine

those policies and the lobbying or informational strategies that can most cost-effectively alter the

opinion of these actors and, thereby, policy outcomes. The accumulation of relevant experience

provides an organization with the opportunity to exploit its existing knowledge and routines in

the current political environment, while also providing an opportunity to expand upon that

knowledge and further refine its routines.

Hypothesis 1 (H1): As policy uncertainty emanating from political hazards increases,

an organization’s experience under the current political regime has an increasingly

negative effect on its subsidiaries’ exit rates.

Although refining and exploiting organizational routines in a specific political context, reinforces

an organization’s competences to manage policy uncertainty, it is important to note that such

refinement and exploitation carries the potential of maladaptation should the political context be

dynamic or in flux (Amburgey & Miner, 1992). In this regard, policy uncertainty that emerges

from environmental dynamism, which in the political context takes the form of regime

14

instability, concerns managers as these changes can alter success criteria for managing the

political environment.

As the success criteria can change, the organizational experience under the current political

regime that may alleviate the influence of policy uncertainty provides little benefit to the

management of policy uncertainty that emanated from a change in the political regime.

Resources such as cultivated contacts and learned routines regarding lobbying processes

developed for one political regime are imperfectly redeployable to its successor. Learned

routines for mitigating policy uncertainty can even potentially become a liability in the aftermath

of a regime transition. A firm that made an international expansion under one political regime,

and accumulated experience relevant to the policy making of that particular regime, could find

itself maladapted or even at a disadvantage under its successor. This maladaptation can enhance

rates of exit in the presence of instability of the current regime or after a transition to a new

regime (Baum & Ingram, 1998, Ingram & Baum, 1997). A straightforward example of this is the

strategy of partnering with a Suharto family member in Indonesia. Such partnerships greatly

enhanced performance up to 1997, but these same partnerships then greatly enhanced exit rates

after the change in the regime led to Suharto’s successors launching a campaign against

corruption, cronyism and nepotism.

Hypothesis 2a (H2a): As policy uncertainty emanating from regime instability

increases, an organization’s experience under the current political regime has an

increasingly positive effect on its subsidiaries’ exit rates.

Hypothesis 2b (H2b): The greater an organization’s experience under previous

political regimes, the higher its subsidiaries’ exit rates.

15

Imitation of Peer Exits

The second strategy for managing policy uncertainty follows from the idea that a firm’s

managers can imitate the decisions of peer firms to contend with uncertainty from the political

environment. This idea has strong roots in neoinstitutional theory, which has a perspective on

organizational action that views inter-organizational linkages (Haunschild, 1994) and reference

groups (Greve, 2000, Haunschild & Miner, 1997) as influencing organizational exit rates by

providing an organization with support or legitimacy (Ingram & Baum, 1997).

Under conditions of uncertainty, organizations turn to the behavior of peers, which drives the

process of mimetic isomorphism, or the process by which organizations become more similar

over time (DiMaggio & Powell, 1983). Imitation can emerge as an organizational strategy

because repeated actions by other organizations convey legitimacy and pressures on

organizational actors to adopt similar decisions, thus inducing a spread of a decision, structure or

strategy across a set of organizations (Fligstein, 1985).

An emerging view within the neoinstitutional literature represents a transition from the viewpoint

that imitation is a pure social response (DiMaggio & Powell, 1983), to the idea that imitation

may involve a technical rationale (Abrahamson & Rosenkopf, 1993). Research points to the idea

that the strength of the technical rationale is related to the salience of an imitable organizational

structure or strategy. Where an organizational action has an observable outcome (Haunschild &

Miner, 1997), or where an organization’s environment is somewhat predictable (Argote,

Beckman, & Epple, 1990), salience is greater, as is the strength of the technical rationale. This

latter point relates to the idea that imitative behavior can be a result of vicarious learning about

organizational actions in unequivocal settings in which the results of a strategy are observable

16

(Baum, Li, & Usher, 2000, Levitt & March, 1988). Prior work in the Japanese banking industry

highlights the role of such learning processes in market niche entry decisions (Greve, 2000).

Research on rational bandwagons conveys a similar prediction. Organizations can exhibit

similarity and clustering in decisions and outcomes, such as organizational adoption, based on a

concern that deviating from the practices of early adopters will carry negative performance

consequences (Abrahamson & Rosenkopf, 1993). Not all bandwagon effects are strictly

technical, however. Bandwagon imitation can occur even in the presence of negative information

on the performance of early adopters, or where there is uncertainty about the long-term

performance implications of a strategy (Rosenkopf & Abrahamson, 1999).

Another technical rationale for the adoption of strategies previously used by one’s peers relies on

the assumption of incomplete information, which can lead managers to infer profitable strategies

for their own organization based on the behavior of other organizations that share similar traits

(Bikhchandani, Hirshleifer, & Welch, 1998). Note that in parallel to the arguments of social

legitimacy, arguments that hinge on vicarious learning or rational bandwagon explanations also

conclude with the point that uncertainty enhances the tendency to imitate other organizations. In

the case of vicarious learning and rational bandwagons, however, imitation occurs because of its

impact on incomplete information. Given that these arguments are parallel, we expect that

imitation will be an effective response to uncertainty, whether that uncertainty emerges from a

lack of information about an environment, as in the case of policy uncertainty, or from a dynamic

change in the environment, as in the case of regime instability.

Hypothesis 3a (H3a): As policy uncertainty emanating from political hazards increases,

prior peer exits have an increasingly positive effect on subsidiary exit rates.

17

Hypothesis 3b (H3b): As policy uncertainty emanating from regime instability

increases, prior peer exits have an increasingly positive effect on subsidiary exit rates.

METHODS

Data Sources and Sample

We test these hypotheses using longitudinal data on the foreign subsidiaries of Japanese firms.

Japan has been a leading source of FDI, which has flowed to an extensive number of countries.

According to the data we compiled, by 2000, more than 120 countries had received Japanese

FDI, with 54 countries possessing at least 30 Japanese subsidiaries. This large country spread

provides the variance we require on our measures of policy uncertainty and regime instability, to

test this study’s hypotheses, while controlling for other nation-level influences on exit rates.

We derived our sample from the list of subsidiaries provided in Toyo Keizai’s compendium of

FDI, Japanese Overseas Investment. We used each annual edition from 1992 to 2001 to

construct longitudinal profiles of Japanese subsidiaries for the 1991 to 2000 period. These

profiles included information on the country and date of subsidiary establishment and the year of

subsidiary exit, if an exit occurred. This process identified 28,525 subsidiaries that existed in the

1991-2000 period, of which 34.5 percent (9,859 subsidiaries) had exited by 2000.

We focused on the analysis of manufacturing subsidiaries of manufacturing firms, as the capital

costs involved in building a manufacturing plant are central to the theoretical arguments

surrounding the effect of policy uncertainty on exit rates. We limited our sample to subsidiaries

formed in the 1991-2000 period to remove left-censored cases from our analysis. Our

independent measures of investment experience, density and the like are, where appropriate,

18

drawn from the full sample of observations including those formed prior to 1991. Our sample

comprised 2,283 manufacturing subsidiaries, formed in the 1991-2000 period by 642 Japanese

manufacturing firms in 53 countries of which 17.7 percent (405 subsidiaries) had exited by 2000.

Dependent Variable

Our dependent variable, Exit, was an indicator variable, Ext, that took a value of 1 if subsidiary x

exited at time t. Observations started in the year 1992, continued until an exit occurred, or were

right-censored in 2000, if the indicator variable Ext was zero in each year t for subsidiary x. Exits

occurred in 38 of the 53 countries with the highest count of exits in China (105), the United

States (88), Thailand (29), Taiwan (20), the United Kingdom (19), Indonesia (19), Malaysia (19),

Hong Kong (14), France (13) and Germany (12).

Independent Variables

Political Hazards. The political hazards index (political hazards) measures the extent to which a

change in the preferences of any one actor may lead to a change in government policy (Henisz,

2000a). The first step in the construction of this index is the identification of the number of

independent branches of government (executive, lower and upper legislative chambers, judiciary

and states or provinces) with veto power over policy change. The preferences of each of these

branches and the status-quo policy were then assumed to be independently and identically drawn

from a uniform, unidimensional policy space. This assumption allows for the derivation of a

quantitative measure of policy uncertainty using a simple spatial model of political interaction.

The initial measure is then modified to take into account the extent of alignment across branches

of government using data on the party composition of the executive and legislative branches.

Alignment across branches increased the feasibility of policy change. The measure is then

19

further refined to capture the extent of preference heterogeneity within each legislative branch.

Greater within-branch heterogeneity increases the costs of overturning policy for aligned

branches.

The main results of the derivation are that (1) each additional veto point (a branch of government

that is both constitutionally effective and controlled by a party different from other branches)

provides a negative but diminishing effect on the total level of hazards and (2) homogeneity (or

heterogeneity) of party preferences within an opposed (or aligned) branch of government is

negatively correlated with the level of hazards. Scores for political hazards for a given country in

a given year ranged from 0.1 (minimal hazards) to 1.0 (extremely hazardous).1 The ten countries

with the highest policy uncertainty that received Japanese foreign direct investment were

Cambodia, China, Indonesia, Iran, Mexico, Myanmar, Pakistan, Saudi Arabia, Sri Lanka and

Vietnam. The ten countries with the lowest policy uncertainty were Australia, Belgium, Canada,

Chile, Denmark, Finland, Israel, Slovak Republic, Switzerland and the United States.

Regime Instability. Our measure of regime instability is an annual count of government crises,

constitutional changes and successful coups as provided by the Cross-National Time Series

Database. This measure was time-varying and lagged by one year. The ten countries with the

highest regime instability that received Japanese foreign direct investment were the Russian

Federation, India, Italy, Pakistan, Turkey, Venezuela, Thailand, Romania, Indonesia and Israel.

The countries with the lowest regime instability were Australia, Bolivia, Cambodia, Chile,

China, Denmark, Finland, France, Greece, Hungary, Iran, Korea, Luxembourg, Malaysia,

Mexico, Myanmar, the Philippines, Portugal, Saudi Arabia, Sri Lanka, Sweden, Switzerland, the

1 The data and additional detail on its construction can be downloaded from http://www-management.wharton.upenn.edu/henisz/.

20

United States and Vietnam. Note that these 24 countries include some with very high and some

with very low policy uncertainty indicating the independent nature of these two constructs.

Organizational experience. Using the political history for each country in our sample as provided

in the Cross-National Time Series Database, the Polity dataset (Gurr, 2001) and The Statesman’s

Yearbook (Turner, various), we calculated Host country experience under the current political

regime and Host country experience under other political regimes as the log of the years of

investment history a firm possessed under the current structure of political institutions (those in

existence on January 1 of the reporting year) and under its predecessors. The most embedded

firm-country pairs were Panasonic, Sanyo, Nissan, Toyota and Honda in the United States. The

most embedded non-US pairs were Panasonic and Sanyo in China.

Prior Exits. To measure the prevalence of mimetic isomorphism in exit behavior (Greve, 1995)

we constructed, percentage of peer subsidiaries that exited in the prior year, which was the

percentage of Japanese subsidiaries in the same 3-digit SIC industry and country as a focal

subsidiary that exited in a given year. This measure was time-varying and lagged by one year.

There were 41 instances in 28 countries in which all of a firm’s peers in the same 3-digit industry

exited in a given year.

Industry, Firm and Subsidiary Controls. To capture national population-level effects of

legitimation and competition (Hannan & Freeman, 1989), we measured the density of Japanese

firms’ activities in the host country of a focal subsidiary. Given the absence of comparable data

on the multinational spread of firms from other nations, we follow previous work in the

international arena and operationalize this construct using the population of Japanese subsidiaries

alone (Mitchell, Shaver, & Yeung, 1994). Density was the number of Japanese subsidiaries

21

operating in a given year in a focal subsidiary’s industry and country. We introduced both raw

counts and a quadratic term.

To account for the levels of flexibility a firm has to shift production from one location to the next

in response to shocks (Bartlett & Ghoshal, 1990, Kogut, 1983) we included Other subsidiaries,

world. We defined this measure as the number of foreign subsidiaries owned by a firm in a

given year. Other subsidiaries, host country was the number of foreign subsidiaries owned by a

firm in a host country in a year. To examine the role of own firm exits in other countries to

control for firm-specific factors that could lead to a global retrenchment independent of

environmental uncertainty and to control for unobserved firm-level factors that could lead to a

pattern of global exits by a firm, we included a count of Exits, rest of world, which was the

number of subsidiaries of a firm that exited in a given year.

To account for the effect of organizational ties (Pfeffer & Salancik, 1978), we developed two

measures of a firm’s business group affiliations: horizontal business group and vertical business

group. These indicator variables respectively marked whether a firm had an affiliation with any

of the horizontal groups, or any of the vertical groups, in Japan. The source for these measures

was Dodwell’s publication, Industrial Groupings in Japan: The Anatomy of Keiretsu.

Finally, we included firm age and subsidiary age and their square terms in all models to examine

for liabilities of newness (Freeman, Carroll, & Hannan, 1983), adolescence (Bruderl &

Schussler, 1990) and age (Barron, West, & Hannan, 1994). We measured firm size with the

logarithm of firm employment (Freeman, Carroll, & Hannan, 1983). We sourced these time-

varying measures from the Nikkei NEEDS tapes and Japanese Overseas Investment.

22

Country controls. Country-level determinants of subsidiary exit rates could include market

demand and market potential (Carroll & Hannan, 1989). We included the log of two time-

varying, lagged measures: Log (Gross Domestic Product per capita) and Log (population), as

sourced from the World Development Indicators 2002 of the World Bank.

Fixed Effects. We used regional indicator variables as proxies for transportation costs and

cultural differences and, to some extent, for differences in investment motivations that may

influence exit rates, as well as to capture time-invariant differences across world regions: Africa,

Asia, Central and Eastern Europe, Central America and the Caribbean, former British colonies,

the Middle East, South America and Western Europe. Annual fixed effects capture variation in

exchange rates and global economic conditions. Industry fixed effects capture sectoral variation

in exit prevalence at the Japanese equivalent of a 2-digit SIC code.

Summary Statistics. Table 1 provides a correlation matrix. Include in Table 1 are descriptive

statistics for the variables included in the full sample and the subsamples in which exit did and

did not occur.

- Insert Table 1 about here -

Modeling Procedure

We estimated exit rates using event history analysis, as implemented by an exponential model.

Event history analysis uses a longitudinal record of events in a sample from a population to

examine the influences that a set of covariates have on an event. Our focal event is an exit by a

subsidiary. In the analysis, each subsidiary x is at risk of exit from country i in each time period t,

23

or until its exit occurs. This technique models the rate of a transition from an origin state to a

destination state (exit) as a function of the covariates. Its general form is:

rjk = exp (αjk0 + Ajk1αjk1 + Ajk2αjk2…)

where rjk is the transition rate from origin state j to destination state k, with the observed

covariate vector Ajk, parameters to be estimated αjk and constant αjk0. The duration of an event is

described by an exponential distribution. The relationship between the covariates and the

transition rate is specified as log-linear to ensure transition rate estimates are not negative. The

estimation uses the maximum likelihood method (Blossfeld & Rohwer, 1995). In the log-relative

hazard parameterization we employ, hazard ratios greater (less) than one indicate exit rates

increase (decrease) when the associated covariate increases in value. To estimate this model, we

expanded the base sample into multiple spells that included all subsidiary-country-year

combinations among the subsidiaries, countries and annual time periods in which an exit could

occur. In each spell, a subsidiary was at risk of exiting and was treated as right censored unless

an exit occurred. Once we divided the data into annual spells, we had as many as 12,206

observations but this dropped to 9,831 after casewise deletion of observations with missing

independent variables.

RESULTS

Table 2 presents the results of our analysis, incrementally adding the theoretical variables of

interest. Models 1 and 2 present hazard ratios for the national, industry, firm and subsidiary-level

control variables. Statistically significant hazard ratios of less than one indicate a reduction in

exit rates whereas ratios greater than one indicate an increase in exit rates. Firm age initially

reduces but subsequently increases exit rates. Subsidiary age has the opposite effect: first

24

increasing but subsequently reducing exit rates. Membership in a horizontal business group

reduces subsidiary exit rates. Firms with a larger global scope of operations experienced lower

subsidiary exit rates and firms with a larger local scope of operations had higher subsidiary exit

rates, consistent with the notions that other international operations increase a firm’s leverage in

negotiating with the government, while larger local operations enhance a firm’s flexibility to

relocate production across local plants.

We observed a strong positive relationship between a firm’s exits in the rest of the world and the

focal country suggesting that corporate strategies of downsizing have a significant effect across

multiple host country markets independent of any country-specific considerations. Size had no

independent effect on subsidiary exit rates nor did country-industry density or our country-level

macroeconomic variables. The year, region and industry indicator variables improved the

model’s explanatory power. Subsidiary exit rates were higher in Europe and South America and

in 1996-99, than in the excluded region-years of Asia in 1992-93.

-- Insert Table 2 About Here --

Model 3 adds the main effects for the two determinants of policy uncertainty: political hazards

and regime instability. As we expected, we observed no direct relationship between either of our

determinants of policy uncertainty and subsidiary exit rates. Model 4 adds the main effects for

the percentage of peer subsidiaries that exited in the prior year and organizational experience

both under the current political regime and those that preceded it. Here, we observe a significant

and substantial effect of prior peer exits and a liability of experience that is larger for experience

under preceding regimes than that for experience under the current regime. We tested for the

presence of a quadratic relationship, but the effect of host country experience was adequately

25

captured by the logarithmic measure introduced here. Models 5 and 6 individually add the sets of

interaction effects that are the focus of our empirical tests and Model 7 includes both sets.

In models 3-7, the hazard ratios from Model 2 are qualitatively unchanged although the hazard

ratios for subsidiary age and horizontal business groups are no longer statistically significant in

the latter models. Country-industry density becomes weakly significant indicating some evidence

for legitimation and competition effects.

In models 5-7, the hazard ratios on our interaction terms provide good empirical support for four

of our five hypotheses. In contrast to the lack of a direct relationship between any of the

determinants of policy uncertainty or the measure of experience under the current political

regime and subsidiary exit rates, the interactions in Model 7, which are consistent with those in

models 5 and 6, demonstrate that exit rates for firms with this type of experience that should

facilitate the implementation of influence strategies so long as the environment remains stable,

have lower exit rates as political hazards increases (H1). We also find that experience under the

current regime increases exit rates as regime instability increases (H2a) and that experience

under past regimes increases subsidiary exit rates (H2b). Consistent with Hypothesis 3a, the

predicted effect of percentage of peer subsidiaries that exited in the prior year on subsidiary exit

is greater, the higher the political hazards. Subsidiaries are more likely to exit a country when

peers exit, particularly in the presence of uncertainty over future policies generated by the

structure of a nation’s political institutions. We do not find support for H3b which predicted a

similar relationship between prior exits and regime instability.

An examination of hazard rate multipliers for various combinations of the independent variables

of theoretical interest, as depicted in Figures 1, 2, 3 and 4, demonstrates the economic

26

significance of the results reported in Table 2. In Figure 1, we plot the predicted hazard ratios

associated with setting policy uncertainty at the value given on the x-axis and experience under

the current political leadership and prior governments at one standard deviation above (high) or

below (low) the mean, as indicated in the legend. We see sharp differences between subsidiaries

with high and low levels of experience under the current political regime. As predicted by H1,

the subsidiaries with high experience display a negative relationship between subsidiary exit

rates, as plotted on the y-axis, and the level of political hazards. Subsidiaries with low

experience, in contrast, display a positive relationship. For example, if we compare firms with

no experience under prior regimes, the effect of increasing political hazards by one standard

deviation (0.4) from its mean value (0.6) is to increase exit rates for a firm with low levels of

experience under the current political regime by 9 percent. The same increase in political hazards

leads to a reduction in exit rates of 63 percent, for a firm with high levels of experience under the

current political regime. For firms with high levels of experience under prior regimes, the effect

is more muted varying from a 4 percent increase in exit rates to a 16 percent decrease.

Figures 2 and 3 demonstrate the liability of experience in the political environment by plotting

the hazard rate multipliers relative to the values plotted in Figure 1 when the same subsidiary

faces a potential or actual change in political regime (H2a) or possesses experience from prior

regimes (H2b). Consistent with H2a, Figure 2 shows that regime instability increases exit rates

particularly for those firms with high levels of experience under the current regime. For these

firms, the hazard rate of exit increases by between 225 (when political hazards are at their

minimum and the possibility of retribution is therefore reduced) and 710 percent (when political

hazards are at their maximum and the possibility of retribution is therefore enhanced). By

contrast, for firms with low experience under the current regime, the effects are minimal ranging

27

from 4 to 5 percent. Consistent with H2b, Figure 3 shows that firms with experience under

previous regimes one standard deviation above the mean have exit rates 217 to 883 percent

higher, with the greatest effect felt by those firms that possess minimal experience under the

current political regime.

In Figure 4, we examine the marginal effect on the predicted hazard ratio of the percentage of

peer subsidiaries that exited in the prior year at low (one standard deviation below the mean) and

high (one standard deviation above the mean) values of policy uncertainty. We find that when

just one percent of peer subsidiaries exited in the prior year, the predicted exit rate was 43 to 54

percent lower than at the mean value of prior peer exits. Meanwhile, when 20 percent of peer

subsidiaries exit, the effect is to increase exit rates by 872 to 1085 percent. Consistent with H3a,

the relationship between the prior exit of peer subsidiaries and the exit rate of a firm’s

subsidiaries is particularly strong where policy uncertainty is high. As compared to environments

with low policy uncertainty, those with high levels of policy uncertainty show a 24 percent

increase in the effect of prior peer exits.

-- Insert Figures 1, 2, 3 and 4 About Here --

Sensitivity Analyses

We examined the sensitivity of our results to several alternative theoretical explanations and

distributional assumptions regarding hazard rates across time. First, we examined several

definitions of the inter-organizational environment so as to insure that our definition of a peer

group at the level of the industry was not driving our results. We divided all Japanese firms into

high and low status cohorts where status was defined by size (sales, assets or employees) and

age. We found consistent evidence of imitation in both the high and low status subsamples

28

though the effects were much stronger among high status firms as suggested by the work of

Stuart, Hoang and Hybels (1999). The effects of organizational experience (both positive and

negative), were strongest for smaller, although not necessarily younger, firms.

Next, we examined if the experience measures proxied for the resource buffering effect of firm

size or diversification that would minimize the impact of uncertainty (Thompson, 1967).

Interactions of policy uncertainty and regime instability with indicator variables for horizontal or

vertical group membership and employment and sales proxies for firm size failed to generate

statistically significant coefficient estimates or to change the results in our base specification.

Given the literature on culture and subsidiary performance (Barkema, Bell, & Pennings, 1996),

we added measures of cultural uncertainty and interacted these with measures of referent and

own-firm experience in the same cultural block. These additions did not alter our results.

Prior studies have found links between corporate performance and exit rates using such proxies

as a firm’s return-on-sales or return-on-assets, but the inclusion of these variables did not

improve the fit of our model. Similarly, multiple proxies for a country’s cost of capital

(Anderson & Tushman, 2001) did not improve model fit. We examined the role of tax policies at

the aggregate level and found a positive association between the rate of capital taxation and

subsidiary exit in the reduced sample of country-years for which tax data were available. Once

again, the results of theoretical interest were unchanged. We introduced additional

macroeconomic data into the specification. When we included the change in the real effective

exchange rate, the growth in real per capita income, the growth of population, the percentage of

value-added from manufacturing, the government budget balance, and the government debt to

GDP ratio, the results did not change substantively.

29

We examined the sensitivity of our results in our base specification to various hazard functions

including the Cox proportional hazards model, a Weibull distribution and a Gamma distribution.

The results were qualitatively unchanged across these functional forms.

DISCUSSION

We investigated the phenomenon of subsidiary exit in international expansions. As these

expansions were made across clearly-defined national borders, we were able to test for the

influence of a particular type of environmental uncertainty emanating from differences in

national political structures, namely policy uncertainty, on subsidiary exit rates. Although we

observed no direct influence of two determinants of policy uncertainty – political hazards and

regime instability – on exit rates, we did find that these determinants of policy uncertainty

differentially affected exit rates depending on a firm’s level and type of experience in a host

country and the actions of its peer firms.

Despite the prominence attached to the concept of environmental influences in neoinstitutional

research, the question of the relative efficacy of various firm strategies for managing various

types of environmental uncertainty remains relatively under-explored. Further, despite the

prominent role ascribed to political change and political processes in organizational survival

(Haveman, Russo, & Meyer, 2001, Ranger-Moore, 1997), neoinstitutional research has only

recently started to exploit the natural variation in the political environment in an international

setting (Guillén, 2002, Henisz & Delios, 2001, Martin, Swaminathan, & Mitchell, 1998). Our

results address both of these limitations in previous research.

Notably, we find that two determinants of policy uncertainty, political hazards and regime

instability, did not directly lead to higher rates of subsidiary exit. This observation stands

30

somewhat in contrast to the general finding that environmental uncertainty creates differential

stresses on organizations such that when uncertainty is heightened, an organization’s ability to

adapt to its changed environment, will vary in such characteristics as age (Ranger-Moore, 1997),

size (Amburgey, Kelly, & Barnett, 1993), and inter-organizational linkages (Kraatz, 1998).

Given that the context of our study is organizational expansion into a new market, our

observation of no baseline influence of policy uncertainty on exit rates implies that organizations

self-select into environments with varying levels of uncertainty. Environments that have higher

levels of uncertainty ex ante are populated by organizations that have the capabilities to manage

this form of uncertainty or potential returns that warrant an attempt to do so. In this sense, at

least in the context of international expansion, uncertainty itself does not elevate exit rates.

Within a particular policy environment, however, we observed that subsidiaries do exhibit

differential rates of exit depending on the level and type of experience a firm possesses, as well

as the actions of peer firms. This variation is linked to the nature of policy uncertainty realized

in a given country ex post. Specifically, if the structure of a host country’s political institutions

lends a wide range of discretion to policymakers, a firm’s experience within that institutional

structure may help provide it with useful information and resources that can aid in influencing

policy outcomes. This experience can thereby reduce realized policy uncertainty. By contrast,

where political institutions themselves are subject to discontinuous change, firms face even

greater difficulties in influencing policy outcomes. Furthermore, the same information and

resources that may have been important assets in the past can be transformed into substantial

liabilities. This finding reinforces the notion that experience provides an organization with an

advantage over its competitors as long as conditions in the environment remain comparatively

stable. In the aftermath of environmental change, inexperienced organizations can be at an

31

advantage relative to their experienced counterparts. After an environmental change, an

inexperienced organization faces a single hurdle in learning about the new environment, while an

experienced counterpart faces a double-hurdle of learning about the new environment, while

trying to slough resources and routines related to the old environment.

Our examination of exit decisions also provides a useful extension to research on inter-

organizational influences to environmental uncertainty. Extant research has focused on the entry

decision (Guillén, 2002, Haunschild & Miner, 1997, Henisz & Delios, 2001) in corporate

diversification. Our research shows how the inter-organizational environment, as manifest in the

pattern of peer firms’ decisions, continues to influence a firm’s strategy even after it has entered

a market, which in turn leads to the idea that among organizations already present in a market

mimesis continues to drive convergence in organizational strategies and structures (Fligstein,

1985).

This convergence came from our observation that managers fashioned their responses to policy

uncertainty with reference to the actions of peer firms. The overall tendency was for a greater

rate of exit given higher levels of peer exit, but this effect was exacerbated by the level of

political hazards. In nations characterized as politically hazardous, where future policies are

uncertain, firms tend to move in concert to exit a nation. This finding reinforces the idea that

following the herd becomes a more common organizational strategy under conditions of

uncertainty, where that uncertainty stems from a lack of information about the environment.

The strategic implications of these results are that firms that readily crawl into bed with a given

political regime may enjoy a short-term advantage but that gain needs to be balanced against the

long-term costs of maladaptation to, or retribution by, succeeding regimes. Simply following the

32

herd is another strategy that may be used in place of or alongside the development of specific

information on, and ties to, an existing regime. The choice of political risk management

strategies depends on managers’ prior beliefs regarding the stability of the current regime. Where

political hazards are high but the regime appears inherently stable, building strong ties to the

existing regime enhances performance. Where political hazards coincide with regime instability,

however, a more fluid strategy of following the herd will avoid the costs of associating with a

dying regime. While this conclusion may appear somewhat intuitive, our results are the first to

show how subsidiary exit rates vary in a broad sample of countries that differ in the determinants

of policy uncertainty. At the same time, we demonstrate the types of firms and strategies under

which subsidiaries enjoyed lower exit rates, given specific types of policy uncertainty.

Our results identify that the strategies used for contending with policy uncertainty depended on

the nature of the uncertainty. Information based uncertainty could be addressed at one level by

following the actions of peer organizations, and at another level by developing experience to

overcome the information deficiency which is at the source of the uncertainty. For uncertainty

that comes from dynamic change in an environment, imitative and experience-based strategies

provided less benefit to an organization than for the case of information-based uncertainty.

Perhaps, more importantly, experience-based strategies can become a liability in a dynamic

environment, which is a finding that connects to research on the applicability and erosion of

experience over time.

Limitations and Future Research

Similar to research that has identified how the home business context can exert substantial

influences on a firm’s international expansion decisions (Guillén, 2002), we considered only the

33

case of home country competitors. In the international environment, other firms from the home

country are among the most likely to be observed (Martin, Swaminathan, & Mitchell, 1998), and

imitated. An industry is an appropriate defining context (Baum & Mezias, 1992, Delacroix &

Swaminathan, 1991) as the actions of organizations of the same type as a focal organization are

more likely to be observed closely (Haveman, 1993). In focusing on the behavior of peer firms

and on organizational experience, we did not consider firm-specific resources, entry mode choice

(Lu, 2002), subsidiary characteristics, or the identity of joint venture partners as influences on

exit rates given environmental uncertainty. An analysis of these endogenous choices would

necessitate a two-stage selection model that separates the effects of a firm’s characteristics and

its external environment on the strategy choice, from their impact on exit rates (Shaver, 1998).

The absence of a mimetic response in the case of regime instability could indicate that, contrary

to our theoretical arguments, firms are relatively capable of independently discerning the nature

of such a fundamental political transformation and thus do not need to rely upon peer behavior.

Policy uncertainty emanating from political hazards is a complex and ambiguous form of

environmental uncertainty in which many political actors interact to shape a range of future

policies of interest to the firm. When we hold the level of political hazards constant, as we do in

our empirical specification, the potential replacement of the political regime of a host country

with another is a more observable, straightforward and easy to interpret type of change than

policy uncertainty emanating from political hazards. The question of whether a new potential

regime is better or worse than its predecessor may not be one for which managers rely heavily

upon the behavior of peer firms to answer. Another explanation for the lack of support of our

hypothesis revolves around a potential benefit of regime change. If firms well-connected to the

prior regime leave, it enhances the ability of previously poorly-connected firms to become

34

relatively well-connected to the new political regime. This positive substitution effect could

offset the negative information effect leading to the lack of a statistically significant relationship.

Conclusion

We use variation in the determinants of policy uncertainty across national environments to

demonstrate that organizations rely upon cues from their peers and upon organizational

experience to manage policy uncertainty. We demonstrate that imitation is a strong influence in

the decision to exit a country particularly when political hazards are high. We call attention to

organizational experience by demonstrating that firms with relevant experience profiles are

better able to forecast or influence future policy environments and thus enjoy lower exit rates.

These same firms are, however, at a disadvantage in the event of a change in the current political

regime. We argue that these findings point to two strategies for dealing with policy uncertainty:

imitation (following the herd) or the development of specific resources and ties to the current

political regime (sleeping with the enemy). Our study thus highlights the importance for

researchers to carefully define the types of environmental uncertainty that an organization faces

and the factors that influence the relationships between these types of uncertainty and

organizational performance.

35

REFERENCES

Abrahamson, Eric & Lori Rosenkopf (1993) 'Institutional and Competitive Bandwagons: Using mathematical modeling as a tool to explore innovation diffusion', Academy of Management Review 18: 487-517.

Amburgey, Terry L., Dawn Kelly, & William P. Barnett (1993) 'Resetting the Clock: The Dynamics of Organizational Change and Failure', Administrative Science Quarterly 38: 51-72.

Amburgey, Terry L. & Anne S. Miner (1992) 'Strategic Momentum: The effects of repetitive, positional and contextual momentum on merger activity', Strategic Management Journal 13: 335-48.

Anderson, Philip & Michael L. Tushman (2001) 'Organizational Environments and Industry Exit: The Effects of Uncertainty, Munificence and Complexity', Industrial and Corporate Change 10(3): 675-711.

Argote, Linda, Sara L. Beckman, & Dennis Epple (1990) 'The Persistence and Transfer of Learning in Industrial Settings', Management Science 36(2): 140-54.

Barkema, Harry G. , John H. J. Bell, & Johannes M. Pennings (1996) 'Foreign entry, cultural barriers, and learning', Strategic Management Journal 17(2): 151-66.

Barron, David N., Elizabeth West, & Michael T. Hannan (1994) 'A Time to Grow and a Time to Die: Growth and mortality of credit unions in New York City, 1914-1990', American Journal of Sociology 100: 381-421.

Bartlett, Christopher A. & Sumantra Ghoshal (1990) 'The Multinational Corporation as an Interorganizational Network', Academy of Management Review 15(4): 603-25.

Baum, Joel A. C. & Paul Ingram (1998) 'Survival-enhancing Learning in the Manhattan Hotel Industry', Management Science 44(7): 996-1016.

Baum, Joel A. C., Stan X. Li, & John M. Usher (2000) 'Making the next move: How Experiential and Vicarious Learning Shape the Locations of Chains' Acquisitions', Administrative Science Quarterly 45(4): 766-801.

Baum, Joel A.C. & Stephen J. Mezias (1992) 'Localized Competition and Organizational Failure in the Manhattan Hotel Industry, 1898-1990', Administrative Science Quarterly 37: 580-604.

Benford, Robert D. & David A. Snow (2000) 'Framing Processes and Social Movements: An Overview and Assessment', Annual Review of Sociology 26: 611-39.

36

Bikhchandani, Sushil, David Hirshleifer, & Ivo Welch (1998) 'Learning from the Behavior of Others: Conformity, Fads and Information Cascades', Journal of Economic Perspectives 12(3): 151-70.

Blossfeld, Hans-Peter & Gotz Rohwer (1995) Techniques of event history modeling : new approaches to causal analysis. Mahwah, NJ: Erlbaum.

Boyd, Brian K, Gregory G. Dess, & Abdul MA Rasheed (1993) 'Divergence Between Archival and Perceptual Measures of the Environment: Causes and Consequences', Academy of Management Review 18(2): 204-26.

Bruderl, Josef & Rudolf Schussler (1990) 'Organizational Mortality: The Liabilities of Newness and Adolescence', Administrative Science Quarterly 35: 530-47.

Carroll, Glenn R. & Michael T. Hannan (1989) 'Density Delay in the Evolution of Organizational Populations', Administrative Science Quarterly 34: 411-30.

Chang, Sea Jin (1995) 'International expansion strategy of Japanese firms: Capability building through sequential entry', Academy of Management Journal 38(2): 383-407.

Cyert, Richard M. & James G. March (1963) A Behavioral Theory of the Firm. Cambridge, MA: Blackwell Business.

Delacroix, Jacques & Anand Swaminathan (1991) 'Cosmetic, Speculative and Adaptive Organizational Change in the Wine Industry: A longitudinal study', Administrative Science Quarterly 36: 631-61.

Delios, Andrew & Witold J. Henisz (2000) 'Japanese Firms' Investment Strategies in Emerging Economies', Academy of Management Journal 43(3): 305-23.

Delios, Andrew & Witold J. Henisz (2003) 'Political Hazards, Experience and Sequential Entry Strategies: The International Expansion of Japanese Firms, 1980-1998', Strategic Management Journal 24(12): 1153-1164.

Dess, Gregory G. & Donald W. Beard (1984) 'Dimensions of Organizational Task Environments', Administrative Science Quarterly 29(1): 52-73.

DiMaggio, Paul J. & Walter W. Powell (1983) 'The Iron Cage Revisited: Institutional Isomorphism and Collective Rationality in Organizational Fields', American Sociological Review 48(April): 147-60.

Downey, H.K. & J.W. Slocum (1975) 'Uncertainty: Measures, Research, and Sources of Variation', Academy of Management Journal 18(3): 562-78.

Duncan, Robert B. (1972) 'The Characteristics of Organizational Environments and Perceived Environmental Uncertainty', Administrative Science Quarterly 17(3): 313-26.

37

Eriksson, Kent, Jan Johanson, Anders Majkgard, & D. Deo Sharma (1997) 'Experiential Knowledge and Cost in the Internationalization Process', Journal of International Business Studies 28(2): 337-60.

Feng, Yi (2001) 'Political Freedom, Political Instability and Policy Uncertainty: A Study of Political Institutions and Private Investment in Developing Countries', International Studies Quarterly 45: 271-94.

Fligstein, Neal (1985) 'The Spread of the Multidivisional Form Among Large Firms, 1919-1979', American Sociological Review 50(3): 377-91.

Freeman, John, Glenn R. Carroll, & Michael T. Hannan (1983) 'The Liability of Newness: Age dependence in organizational death rates', American Sociological Review 48(October): 692-710.

Geertz, Clifford (1978) 'The Bazaar Economy: Information and search in peasant marketing', American Economic Review 68(Supplement): 28-32.

Greve, Henrich R. (1995) 'Jumping Ship: The diffusion of strategy abandonment', Administrative Science Quarterly 40: 444-74.

Greve, Henrich R. (2000) 'Market Niche Entry Decisions: Competition, learning and strategy in Tokyo banking, 1894-1936', Academy of Management Journal 43(5): 816-36.

Guillén, Mauro (2002) 'Structural Inertia, Imitation and Foreign Expansion: South Korean firms and business groups in China, 1987-1995', Academy of Management Journal 45(3): 509-25.

Gurr, Ted. R. (2001) 'Polity IV: Political Structures and Regime Change, 1800-2000 [computer file]'. Boulder, CO: Center for Comparative Politics [producer], Inter-University Consortium for Political and Social Research [distributor].

Hannan, Michael T. & John Freeman (1989) Organizational Ecology. Cambridge, MA: Harvard University Press.

Haunschild, Pamela R. (1994) 'How Much Is That Company Worth? Interorganizational Relationships, Uncertainty and Acquisition Premiums', Administrative Science Quarterly 39: 391-411.

Haunschild, Pamela R. & Anne S. Miner (1997) 'Models of Interorganizational Imitation: The effects of outcome salience and uncertainty', Administrative Science Quarterly 42: 472-500.

Haveman, Heather A. (1993) 'Follow the Leader: Mimetic Isomorphism and Entry into New Markets', Administrative Science Quarterly 38: 564-92.

38

Haveman, Heather A., Michael V. Russo, & Alan D. Meyer (2001) 'Organizational Environments in Flux: The Impact of Regulatory Punctuations on Organizational Domains, CEO Succession and Performance', Organization Science 12(3): 253-73.

Henisz, Witold J. & Andrew Delios (2001) 'Uncertainty, Imitation, And Plant Location: Japanese Multinational Corporations, 1990-1996', Administrative Science Quarterly 46(3): 443-75.

Henisz, Witold J. & Bennet A. Zelner (2004) 'Legitimacy, Interest Group Pressures and Change in Emergent Institutions: The Case of Foreign Investors and Host Country Governments', Academy of Management Review 29(Forthcoming).

Henisz, Witold Jerzy (2000a) 'The Institutional Environment for Economic Growth', Economics and Politics 12(1): 1-31.

Henisz, Witold Jerzy (2000b) 'The Institutional Environment for Multinational Investment', Journal of Law, Economics and Organization 16(2): 334-64.

Hilgartner, Stephen & Charles L. Bosk (1988) 'The Rise and Fall of Social Problems: A Public Arenas Model', American Journal of Sociology 94(1): 53-78.

Hitt, Michael A., MT Dacin, E Levitas, JL Arregle, & A Borza (2000) 'Partner selection in emerging and developed market contexts: Resource-based and organizational learning perspectives', Academy of Management Journal 43(3): 449-67.

Holburn, Guy & Rick Vanden Bergh (2002) 'Policy and Process: A Framework for the Design of Non-Market Strategy', Advances in Strategic Management 19: 33-64.

Ingram, Paul & Joel Baum (1997) 'Opportunity and Constraint: Organizations' learning from the operating and competitive experience of industries', Strategic Management Journal 18(Summer Special Issue): 75-98.

Kingdon, John W. (1984) Agendas, Alternatives, and Public Policies. Boston: Little, Brown.

Kobrin, Stephen (1979) 'Political Risk: A Review and Reconsideration', Journal of International Business Studies 10(Spring): 67-80.

Kogut, Bruce (1983) 'Foreign direct investment as a sequential process'in Kindleberger, C. P., (ed) The Multinational Corporations In the 1980s. 35-56. Cambridge: MIT Press.

Kraatz, M.S. (1998) 'Learning by association? Interorganizational networks and adaptation to environmental change', Academy of Management Journal 41(6): 621-43.

Krehbiel, Keith (1999) 'Pivotal Politics: A refinement of nonmarket analysis for voting institutions', Business and Politics 1(1): 63-81.

Levinthal, Dan & James G. March (1981) 'A Model of Adaptive Organizational Search', Journal of Economic Behavior and Organization 2: 307-33.

39

Levitt, Barbara & James G. March (1988) 'Organizational Learning'in Scott, W.R. & J. Blake, (eds) Annual Review of Sociology.

Lu, Jane W. (2002) 'Intra- and Inter-Organizational Imitative Behavior: Institutional Influences on Japanese Firms' Entry Mode Choice', Journal of International Business Strategy 33(1): 19-37.

March, James G. (1988) Decisions in Organizations. New York: Blackwell.

Martin, Xavier, Anand Swaminathan, & Will Mitchell (1998) 'Organizational Evolution in the Interorganizational Environment: Incentives and Constraints on International Expansion Strategy', Administrative Science Quarterly 43(September): 566-601.

Milliken, F . J. (1987) 'Three Types of Perceived Uncertainty About the Environment: State, Effect and Response Uncertainty', Academy of Management Review 12(1): 133-43.

Miner, Anne S., Terry L. Amburgey, & Timothy M. Stearns (1990) 'Interorganizational Linkages and Population Dynamics: Buffering and transformational shields', Administrative Science Quarterly 35: 689-713.

Miner, Anne S. & Pamela R. Haunschild (1995) 'Population Level Learning', Research in Organizational Behavior 17: 115-66.

Mitchell, Will, Myles Shaver, & Bernard Yeung (1994) 'Foreign Entrant Survival and Foreign Market Share: Canadian companies experience in United States medical sector markets', Strategic Management Journal 15(7): 555-67.

Nelson, Richard R. & Sidney G. Winter (1982) An Evolutionary Theory of Economic Change. Cambridge: Belknap Press.

North, Douglass C. (1990) Institutions, Institutional Change, and Economic Performance. New York: Cambridge University Press.

Pennings, Johannes M., HG Barkema, & SW Douma (1994) 'Organizational Learning and Diversification', Academy of Management Journal 37(3): 608-40.

Pfeffer, Jeffrey & Gerald R. Salancik (1978) The External Control of Organizations. New York: Harper and Row.

Ranger-Moore, James (1997) 'Bigger May Be Better, but is Older Wiser? Organizational age and size in the New York life insurance industry', American Sociological Review 62(December): 903-20.

Rosenkopf, Lori & Eric Abrahamson (1999) 'Modeling Reputation and Informational Influences in Threshold Models of Bandwagon Innovation Diffusion', Computational & Mathematical Organization Theory 5(4): 361-84.

40

Shaver, J. Myles (1998) 'Accounting for Endogeneity When Assessing Strategy Performance: Does entry mode choice affect FDI survival?' Management Science 44(4): 571-85.

Stuart, Toby E., Ha Hoang, & Ralph C. Hybels (1999) 'Interorganizational Endorsements and the Performance of Entrepreneurial Ventures', Administrative Science Quarterly 44: 315-49.

Thompson, James D. (1967) Organizations in Action: Social Science Bases of Administrative Theory. New York: McGraw-Hill Book Company.

Turner, Barry, editor. various. The Statesman's Yearbook. New York: St. Martin's Press.

41

TABLE 1: Descriptive Statistics and Correlation Matrix

Measure 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 1. Exit 2. Log (Host country experience, prior

political regimes) 0.01

3. Log (Host country experience, current political regime)

0.08 -0.08

4. Percentage of peer subsidiaries that exited in the prior year

0.43 0.01 0.10

5. Political Hazards -0.05 -0.00 -0.12 -0.07 6. Regime Instability -0.01 0.22 -0.17 -0.03 -0.14 7. Subsidiary age 0.09 -0.01 0.32 0.10 -0.00 -0.03 8. Firm age -0.01 0.06 0.13 0.01 0.04 -0.00 0.07 9. Horizontal business group -0.02 0.05 0.08 -0.01 -0.01 -0.00 -0.01 0.17 10. Vertical business group -0.01 0.07 0.14 0.01 0.02 0.01 -0.02 0.03 0.1211. Log (employment) -0.00 0.18 0.39 0.00 0.05 0.02 -0.03 0.24 0.17 0.3312. Other exits by focal firm in rest of world 0.09 0.11 0.30 0.07 0.05 -0.00 0.11 0.07 0.06 0.24 0.4513. Other subsidiaries, world 0.00 0.15 0.47 0.01 0.09 -0.01 0.02 0.14 0.06 0.40 0.74 0.6414. Other subsidiaries, host country 0.04 0.08 0.68 0.04 0.10 -0.10 0.05 0.10 0.09 0.24 0.50 0.41 0.7015. Host country – industry density 0.00 -0.04 0.20 0.00 0.10 -0.12 0.04 -0.05 -0.04 0.05 0.02 -0.01 0.02 0.1816. Log (Gross Domestic Product) 0.07 0.01 0.22 0.11 -0.83 -0.04 0.03 -0.03 0.02 -0.03 -0.05 -0.05 -0.09 -0.02 -0.0317. Log (Population) -0.03 -0.06 0.14 -0.03 0.72 -0.13 0.01 0.04 -0.02 0.04 0.06 0.04 0.11 0.27 0.30 -0.64

Descriptive Statistics 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16

17 Mean – All Subsidiaries 0.03 0.77 1.82 0.04 0.59 0.18 2.87 60.13 0.28 0.17 8.22 1.29 34.15 4.33 31.24 7.92 19.07 Mean – Surviving subsidiaries 0.00 0.78 1.80 0.03 0.60 0.18 2.85 60.15 0.28 0.17 8.22 1.24 34.12 4.29 31.26 7.90 19.08 Mean – Exiting subsidiaries 1.00 0.53 2.61 0.25 0.50 0.14 3.45 59.46 0.24 0.16 8.19 2.97 35.02 5.56 30.64 8.55 18.73 Standard deviation – All Subsidiaries 0.17 1.21 1.48 0.08 0.38 0.50 2.26 17.58 0.45 0.38 1.37 2.82 39.16 5.74 37.54 1.57 1.70 Standard deviation – Surviving Subsidiaries 0.00 1.22 1.48 0.06 0.38 0.50 2.27 17.54 0.45 0.38 1.37 2.74 39.31 5.71 37.47 1.56 1.70 Standard deviation – Exiting Subsidiaries 0.00 0.89 1.45 0.28 0.37 0.41 1.81 18.68 0.43 0.36 1.38 4.42 33.96 6.50 39.73 1.65 1.69

42

TABLE 2* : Policy Uncertainty, Regime Instability and Subsidiary Exit Rates Variable Model

1 Model

2 Model

3 Model

4 Model

5 Model

6 Model

7 Political Hazards X Log (Host country experience current political regime (H1:<1))

0.704 0.004

0.702

0.004 Regime instability X Log (Host country experience current political regime (H2a: >1))

1.285 0.004

1.306

0.003 Log (Host country experience prior political regimes (H2b>1))

1.478 0.000

1.420 0.001

1.462 0.000

1.421 0.001

Political Hazards X % of peer subsidiaries that exited in the prior year (H3a: >1)

13.355 0.002

13.278

0.002 Regime instability X % of peer subsidiaries that exited in the prior year (H3b>1)

1.166 0.571

1.430 0.168

Log (Host country experience current political regime)

1.258 0.000

1.163 0.046

1.245 0.002

1.152

0.064 Percentage of peer subsidiaries that exited in the prior year

99.2470.000

105.08 0.000

40.144 0.000

40.948

0.000 Political Hazards 0.837

0.563 0.724 0.295

0.840 0.583

0.517 0.040

0.620 0.159

Regime Instability 0.986 0.886

0.920 0.440

0.847 0.127

0.907 0.453

0.775 0.078

Subsidiary age 1.336 0.011

1.329 0.015

1.210 0.128

1.229 0.098

1.210 0.122

1.221 0.101

Subsidiary age2 0.971 0.046

0.972 0.065

0.984 0.305

0.983 0.283

0.983 0.287

0.983 0.276

Firm age 0.954 0.005

0.953 0.005

0.949 0.001

0.951 0.001

0.950 0.000

0.953 0.002

Firm age2/1000 1.352 0.007

1.357 0.007

1.382 0.001

1.367 0.002

1.375 0.001

1.356 0.002

Horizontal Business Group 0.734 0.018

0.704 0.010

0.774 0.056

0.768 0.053

0.790 0.076

0.779 0.063

Vertical Business Group 1.115 0.514

1.126 0.487

1.053 0.755

1.033 0.852

1.064 0.705

1.044 0.798

Log (Employment) 1.034 0.614

1.045 0.534

0.997 0.958

1.003 0.957

0.984 0.792

0.989 0.858

Other subsidiary exits, rest of world 1.168

0.000 1.171 0.000

1.125

0.000 1.128 0.000

1.124 0.000

1.127

0.000 Other subsidiaries, rest of world 0.979

0.000 0.978 0.000

0.981 0.000

0.981 0.000

0.982 0.000

0.981 0.000

Other subsidiaries, host country 1.085 0.000

1.088 0.000

1.055 0.000

1.076 0.000

1.057 0.000

1.078 0.000

Host country – industry density 1.0020.860

1.001 0.902

1.000 0.994

1.006 0.133

1.006 0.136

1.007 0.069

1.007 0.071

Host country – industry density2/1000 0.9940.800

0.998 0.921

0.995 0.828

0.961 0.085

0.964 0.109

0.956 0.069

0.958 0.063

Log (Gross Domestic Product) 1.2600.016

1.175 0.103

1.084 0.490

0.969 0.785

0.997 0.981

0.978 0.847

1.003 0.979

Log (Population) 1.0230.695

0.942 0.329

0.947 0.468

0.997 0.969

1.028 0.749

1.007 0.934

1.030 0.734

N 12206 10147 9831 9831 9831 9831 9831 Log-likelihood -1319 -1173 -1119 -906.2 -900.5 -899.8 -894.1* Hazard ratios are reported with p-values in italics. Hazard ratios for region, time and industry indicator variables not reported.

43

FIGURE 1As Policy Uncertainty Emanating from Political Hazards Increases, Firms with

Experience under the Current Political Regime Have Lower Subsidiary Exit Rates (Hypothesis 1)

0

0.5

1

1.5

2

2.5

3

0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1

Political Hazards

Pred

icte

d H

azar

d R

atio

High Experience Under Current Political Regime, No Prior Experience Low Experience Under Current Political Regime, No Prior Experience

Note: Hazard ratios calculated at mean levels for all other variables.

44

FIGURE 2As Policy Uncertainty Emanating from Regime Instability Increases, Firms with Experience

Under the Current Political Regime Have Higher Subsidiary Exit Rates (Hypothesis 2a)

1

2

3

4

5

6

7

8

0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1

Political Hazards

Haz

ard

Rat

io M

ultip

lier R

elat

ive

to F

igur

e 1

High Experience Under Current Unstable Regime, No Prior ExperienceLow Experience Under Current Unstable Regime, No Prior Experience

Note: Hazard ratios calculated at mean levels for all other variables.

45

FIGURE 3Firms with Experience Under Previous Political Regimes Have

Higher Subsidiary Exit Rates(Hypothesis 2b)

1

2

3

4

5

6

7

8

9

10

0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1

Political Hazards

Haz

ard

Rat

io M

ultip

lier R

elat

ive

to F

igur

e 1

High Experience Under Current Political Regime and Previous RegimesLow Experience Under Current Political Regimes, High Experience under Previous Regimes

Note: Hazard ratios calculated at mean levels for all other variables.

46

Figure 4As Policy Uncertainty Emanating from Political Hazards Increases, Prior Peer Exits Have an

Increasingly Positive Effect on Subsidiary Exit Rates (Hypothesis 3a)

0

2

4

6

8

10

12

0% 1% 2% 3% 4% 5% 6% 7% 8% 9% 10% 11% 12% 13% 14% 15% 16% 17% 18% 19% 20%

Percentage of Peer Subisdiaries that Exit in Prior Year

Mar

gina

l Effe

ct o

f Prio

r Pee

r Exi

ts o

n Pr

edic

ted

Haz

ard

Rat

io

Low Policy Uncertainty High Policy Uncertainty

Note: Exit rate multiplier calculated as compared to a subsidiary facing no prior exits while at means of all other variables.