Openness and Internal Conflict
Christopher S. P. Magee
Department of Economics
Bucknell University
Lewisburg, PA 17837
[email protected]
Tansa George Massoud
Department of Political Science
Bucknell University
Lewisburg, PA 17837
[email protected]
Openness and Internal Conflict
Abstract
This paper examines the relationship between economic openness and internal conflict.
The different theoretical perspectives on how openness affects the internal stability of a country
are discussed and then empirical estimates of the relationship between conflict and openness are
presented. The correlation between openness and conflict in the data is negative: more open
countries tend to have less internal conflict by several different measures. Internal conflict
affects the level of openness, however, which suggests that openness should be treated as an
endogenous variable. When the effect of openness on conflict is estimated using instrumental
variable or fixed effects regressions to control for endogeneity, openness significantly inhibits
the most intense civil wars but it has no significant impact on the prevalence of less severe civil
wars or on other measures of a country’s internal stability. There is robust evidence, on the other
hand, that any type of conflict within a country reduces its international trade. Internal conflict is
found to significantly reduce a country’s level of openness in most cases using ordinary least
squares, instrumental variables, or fixed effects regressions. The results are similar whether
openness is measured using trade flows or foreign direct investment and when internal conflict is
measured over only political, economic, or military issues.
1. INTRODUCTION
Since the end of the Cold War, civil wars have become the dominant form of conflict,
making an understanding of their origins extremely significant (Gleditsch et al. 2002; Erikson
and Wallensteen 2004). Although researchers have been studying for some time the causes of
civil war from a state level of analysis, it is only recently that they have begun to highlight the
link between globalization and internal conflict.
As in the debate about the impact of trade on interstate conflict, the relationship between
openness and internal conflict is explained theoretically by the liberal and structural paradigms.
The liberal model highlights the pacifying effects that economic growth and development have
on the likelihood of civil war. The structural model, on the other hand, singles out inequality as
the basic reason for conflict within society. Most researchers testing the conflicting hypotheses
from these models examine the correlation between economic openness and the onset or
presence of civil wars.
This paper contributes to the debate about globalization and peace in several ways.
Using a new events data set, we examine the impact of trade openness on the overall level of
cooperation within a country and on the level of cooperation for political, economic, and military
events. We also test the influence of trade openness on a constructed country stability measure
and on a more traditional indicator of civil war. In this way, we extend the literature on civil
wars by investigating whether openness affects broader measures of internal conflict. In doing
so, we are one of only a few papers to account for the reciprocal effect between trade and
internal conflict.
The variables and measures are drawn from an events data set covering the period 19902004. The events data include not only overt military disputes but also more subtle degrees of
1
political, military, and economic cooperation and conflict. Our goal is to extend the empirical
literature by investigating whether the relationship between trade and domestic cooperation
varies for lower levels of conflict and different categories of interaction.
The empirical results in this paper reveal that open countries have less internal conflict –
they are less likely to have a civil war, have more cooperative internal relationships, and have
higher levels of country stability. The negative correlation between openness and internal
conflict is similar to what previous researchers have found. We show in this paper, however,
that the negative correlation between openness and conflict arises primarily because more
internal stability leads to greater openness rather than because higher levels of openness cause
greater stability. When we perform instrumental variables regression (two-stage least squares) or
fixed effects regressions, trade openness has no significant impact on internal stability in seven
of the eight models we estimate. The one exception is that more open countries are significantly
less likely to experience the most intense civil wars. Higher levels of internal stability, on the
other hand, are consistently estimated to have a significant effect on openness regardless of
whether we use ordinary least squares, two-stage least squares, or fixed effects regressions.
Thus, there seems to be strong evidence that greater stability allows countries to open their
borders but only weak evidence that more openness leads to better internal relations.
2. TRADE OPENNESS AND CIVIL WAR
Recent empirical studies have yielded much insight about the effect of trade on military
conflict. From an interstate level, papers such as Oneal and Russett (1999) find that increased
trade flows reduce the likelihood of militarized interstate disputes. Other studies, such as
2
Barbieri (2002), however, find a positive relationship between trade and conflict.1 While the
effects of trade on militarized interstate disputes have been extensively studied, there is less
research on the impact of globalization on internal conflict. Most studies linking the impact of
trade on internal conflict focus on the onset or prevalence of civil wars. Due to the use of
different data sets and variations in the definitions of economic openness and of what constitutes
a civil war, the results of the existing quantitative studies are mixed. Sambanis (2004) combines
the results of quantitative studies on greed with evidence from case studies and recommends that
researchers come up with better measures for some of the empirical variables used to define the
link between economics and civil wars.
To explain the connection between trade and civil wars, most researchers root their
explanation in either the liberal or structural paradigm.
Liberal Model
The same logic underlying the relationship between trade and peace on the international
level is said to hold when we examine the impact of trade on internal conflict. Open economies
have more trading relationships that can be destroyed by a civil war, and thus the opportunity
cost of internal conflict is higher. The logic of the liberal paradigm suggests that we should be
able to observe greater levels of cooperation within states for countries that have high levels of
economic interaction. The major hypothesis is that trade promotes development and
development is in turn linked to peace (Hegre, Gissinger and Gleditsch 2003; Sachs and Warner
1995). In addition to development, Barbieri and Reuveny (2005) cite a reduction in inequality
and state control, and an increase in communication and information flows as conducive to
1
Useful reviews of the literature on the relationship between trade and conflict include the volumes by Mansfield
and Pollins (2003) and Schneider, Barbieri and Gleditsch (2003).
3
peace. The posited relationship found in the liberal model is supported by the rationality model
which argues that fear of economic and social welfare losses from war have a deterrent effect on
conflict (Polachek 1980). Thus, the anticipated costs of lost trade are likely to prohibit highly
interdependent states from becoming embroiled in civil wars.
Barbieri and Reuveny (2005) test several theories about trade and civil wars and find that
globalization, as measured by several indicators, decreases the presence or prevalence of civil
wars. Trade and foreign direct investment (FDI) are associated with economic growth and
therefore peace. De Soysa (2002) also shows that trade as a percentage of GDP has a strong
negative impact on conflict. Although a high level of economic openness is associated with a
low probability of civil war, the steps toward globalization (measured by the changes in the level
of economic openness) may increase the risk of armed conflict (Bussmann and Schneider 2007).
Fearon and Laitin (2003), on the other hand, suggest that trade has no effect on the onset of civil
wars.
Economic openness improves economic development and this in turn diminishes the
motivation behind greed or predation as possible causes of internal wars (Collier and
Hoeffler 1998; Fearon and Laitin 2003; Ross 2004). The greed explanation for civil wars
emphasizes the financing opportunities for rebel groups through access to natural resources such
as oil, diamonds or drugs. The emphasis in such studies is on the opportunity to form a rebel
group rather than political science explanations that highlight grievances as the cause for
rebellions. An implication of the greed literature is that civil wars are more likely to occur in
countries that depend heavily on primary commodities exports, since the presence of such
resources makes them an attractive target for rebels. Thus, improvement of economic conditions
will decrease the incentives for individuals to participate in rebellions, making it difficult for
4
leaders to recruit rebels. Collier and Hoeffler (2004) show that the opportunity model of civil
wars performs better than the grievance model. However, Elbadawi and Sambanis (2002)
provide evidence that grievance issues represented by such variables as democracy and ethnic
diversity are significant in explaining the prevalence of civil wars. Furthermore, economic
growth through trade is likely to provide the governments of such countries with enough strength
to deter or put down rebellions (Fearon and Laitin 2001). The greed argument also has been
extended to explain the duration of civil wars. There seems to be a consensus that access to
financing prolongs wars. For example, Fearon (2004) shows that civil wars tend to be long if
rebels can obtain their funding from contraband such as opium, diamonds or coca. Thus, trade
can reduce conflict by raising the capacity of government and reducing the opportunity costs to
act peacefully due to alternative income earning activities (de Soysa 2002).
Structural Model
The structural model argues that trade leads to inequality and conflict. The trigger
mechanism for conflict is the increase in inequality that is created through a country’s openness
to globalization. Marxists and neo-Marxists argue that development can increase conflict in
society through a variety of means, including exploitation, class conflict, and economic crises
(Boswell and Dixon 1993). Muller and Seligson (1987) provide strong support for the link
between income inequality and political violence. The rate of economic growth in countries can
also be associated with greater conflict due to feelings of relative deprivation on the part of some
groups or classes, since economic growth is not evenly distributed. This is particularly true for
countries that depend on the export of commodities.
5
Hegre, Gissinger, and Gleditsch (2003) show that countries that export primary
commodities benefit less from economic openness. One general implication of this literature is
that we can expect to see greater conflict in the developing world since inequality has been
increasing for these countries. Reuveny and Li (2003) show that trade openness reduces income
inequality, particularly for LDCs; however, FDI inflows increase the level of income inequality.
Other studies have found no relationship between inequality and civil war (Collier and Hoeffler
2001).
Effects of Civil Wars on Trade
Estimates of the effect of openness on internal conflict are complicated by the results
from many studies showing that civil wars affect external trade flows. At a basic level, civil
wars increase the costs and risks associated with trading due to domestic insecurity and
disruptions of trade routes (Collier 1999; Stewart, Huang and Wang 2001; Blomberg and Hess
2002). Civil wars also tend to negatively influence short-run growth within countries and in
surrounding neighbors (Murdoch and Sandler 2002). Furthermore, civil wars are likely to drive
away foreign direct investments and encourage capital flight for the same reasons. Collier
(1999) argues that civil wars damage investment as well as future economic growth and that the
pace of economic recovery is greater after long civil wars rather than short civil wars. In one
study, civil wars decreased bilateral trade by one-third (Bayer and Rupert 2004). The decrease in
growth and income is also true for political violence short of war (Sambanis 2004). In a
comprehensive model, Blomberg and Hess (2002) show that internal conflict, external conflict,
and the economy are linked. Martin, Mayer, and Thoenig (2008) find that severe civil wars
6
reduce trade by about 25% in the first year of conflict and the destruction of trade persists so that
trade remains 40% lower than normal even 25 years after the start of the conflict.
While the vast majority of studies focus on bloody civil wars as a measure of internal
conflict, some recent scholars such as Hegre, Gissinger, and Gleditsch (2003) have argued that
we should test different models of trade using data on levels of violence short of civil war. The
next section presents a model in which we take such suggestions seriously by moving beyond
looking at a dichotomous variable that defines the existence or absence of civil war and instead
use broader measures of internal stability. In doing so, we test the predictions of the liberal and
structural models about the impact of economic openness on domestic stability to answer the
following questions. What is the effect of economic openness on internal political, economic,
and military cooperation? How does trade openness influence the overall level of domestic
stability of a country? What are the reciprocal effects of trade and domestic conflict?
3. DATA AND EMPIRICAL MODEL
The main hypothesis we examine is that a country’s economic openness affects its
internal stability. The liberal model suggests that a more open economy may lead to more rapid
economic growth and a higher level of development, which can pacify relations within the
country. Higher levels of development are also associated with greater democracy, which can
reduce military conflict by allowing adversarial groups in society to contend with each other in
the political arena rather than on the battlefield. Furthermore, since conflict threatens the gains
from trade, there is a higher cost of conflict in more open countries. Thus, we expect the logic
underlying a decrease in civil wars to also apply to a reduction in overall internal conflict short
of civil war.
7
If globalization leads to greater inequality, on the other hand, as the structural model
suggests, it could raise internal conflict. The structural model identifies several reasons why
globalization may promote inequality, including uneven distribution of economic gains,
exploitation, class conflict, and economic crises. The effect of openness on internal conflict also
depends on how quickly a country changes its level of integration. Bussman and Schneider
(2007) show that global economic integration should reduce civil wars but that the process of
liberalization can increase the prospects for domestic violence.
Equation (1) provides the basic empirical model we estimate in the paper.
(1)
Conflict = F ( past conflict , openness, control variables) + u
The control variables in equation (1) include the level of economic development and growth in a
country, the type of government, the population and population density, the fraction of
population in prison, and ethnic fractionalization among the citizens. We also add continent
dummy variables to measure differences in internal conflict across regions and year dummy
variables to control for changes in worldwide conflict over time. A lagged conflict variable is
included to account for the likelihood that conflict in one time period may persist and result in
higher conflict in future years.
For a measure of each country’s openness, we use total exports plus imports as a share of
GDP from the Penn World Tables. In later estimates, we define openness as inward FDI as a
share of GDP. The Penn World Tables provide country population, real GDP, and real GDP per
capita information. Population is measured in billions of people, GDP in billions of 2000
dollars, and GDP per capita in thousands of 2000 dollars. A chain index of prices is used to
convert GDP and GDP per capita from nominal to real values. Population density is measured as
thousands of people per square kilometer. We also include a measure of ethnic fractionalization
8
for each country from Alesina (2003). Prison population rates are measured as prisoners per
thousand people in the country, and the data come from the World Prison Brief published by the
International Center for Prison Studies. The country’s level of democracy is from the Polity IV
dataset, and it ranges from -10 for autocracies to 10 for pure democracies. Hegre, Gissinger, and
Gleditsch (2003) hypothesize that civil wars are unlikely under pure authoritarian regimes, which
can effectively suppress opposition, and under democracies, where opposition can be expressed
through the political system. Internal conflict is likely to be highest in regimes that lie between
the two extreme cases. Thus, we rescale the polity variable so that it runs from 0 to 20 and
include both the polity and polity squared as explanatory variables in estimating equation (1).
Most previous studies examining the link between openness and conflict define conflict
as the presence or onset of a civil war. The Correlates of War data on intra-state conflict and
Uppsala/PRIO Armed Conflict Dataset provide two sources of data on civil wars, and each of
these data sets includes indicators of the intensity of the conflict as measured by the number of
battle deaths. In one set of regressions, we present results using a dummy variable that equals
one if a country is defined as having any civil war in year t with at least 25 battle-related deaths
in the Uppsala/PRIO Armed Conflict Dataset. We also present regression results using the
cutoff of 1000 battle-related deaths to count as a civil war. The last variable is referred to by
Barbieri and Reuveny (2005) as the presence of a civil war or civil war prevalence, as opposed to
civil war onset which would be coded as one only in the first year of the war. We refer to a civil
war with at least 1000 deaths as an intense civil war. Since the civil war variables are
dichotomous, we estimate equation (1) using a probit model when a civil war is the measure of
internal conflict.
9
There are many levels of disputes that fall short of military conflicts, however, and the
armed conflict data sets miss these lower-level disputes. In order to examine lower levels and
types of conflict other than purely military battles, we also provide measures of conflict using an
events data set from 1990 to 2004 based on machine-coded readings of Reuters news reports.
King and Lowe (2003) compared the computer program used to read the news reports with
human coders and found that both methods were equally accurate ways of coding the news
reports. Each event has a source and target country identified and is given one of 157 possible
Integrated Data for Events Analysis (IDEA) codes. Bond, et. al. (2003) describe the IDEA
framework. King and Lowe (2003) provide a mapping from IDEA codes to the conflictcooperation scale described in Goldstein (1992), in which negative values represent acts of
conflict and positive values indicate cooperation. In order to focus on internal conflict or
cooperation within a country, we restrict our data set to include only events in which both the
source and target country are the same.
The data set contains 9,507,513 events that are internal to a country during the period
1990 – 2004. One measure of the cooperation level within a country in year t is the average
Goldstein conflict-cooperation score over all the events:
∑ (Goldstein j )
(2)
Average Goldstein jt =
year t events
Nt
,
where N t is the number of events reported for country j in year t, and Goldstein j is the conflictcooperation score for each event from Goldstein (1992). These scores range from -10 for
military engagements to 8.3 for extending military aid, and thus the average score for a country
will be negative if its internal actions are primarily conflictual and it will be positive if its actions
are mostly cooperative. While the average Goldstein score is limited to being within a certain
10
range, none of the observations included in the regressions lie on the boundary, and thus we can
use a linear regression when this score is our measure of internal conflict.
There are many advantages to the events data. They provide greater flexibility in
measuring conflict and a more accurate indication of lower level disputes. The events data also
allow researchers to examine political or economic disputes within a country as well as military
conflict. One concern with using machine-coded events data is that a single event may be
reported several times in news reports and thus it may show up as more than one event in the
data set. To ensure that our results are not being driven by repeated news stories, we have re-run
each of the estimates below after excluding any event observations that might be a result of
repeat news reports (observations with variable values identical to an event occurring in the
previous seven days). The results we report below were unchanged when these potentially
repeat news reports were excluded.
A second concern in using events data is that many more news reports are issued about
some countries than about others, which may affect the event data conflict measures. As a
robustness check on our later results, we have included in the conflict equation a variable
measuring the number of newspapers per capita in the country’s population. This news variable
has no significant effect on the average Goldstein conflict-cooperation measure from equation
(2), and including the news variable does not change any of the results we describe below.
A second measure of internal conflict is a variable designed to quantify country stability.
This variable is a modified version of a country’s conflict-carrying capacity, which Jenkins and
Bond (2001, 4) define as “the ability of the state to regulate intense internal conflict without loss
of system integrity.” The idea is that countries are unstable if both the government and the civil
sector participate in contentious actions and if force is often used rather than merely being
11
threatened. Contentious actions are defined as those in which an actor threatens, demonstrates,
sanctions, expels, seizes, or uses force against another.
(3)
CS = 1 − {(
cont g
(1 + cont g ) / 2
)2 + (
cont c − cont g 2
cont c
force
)2 + (
) }×
,
(1 + cont c ) / 2
2
violence
where cont g and cont c are the proportions of government and civil actions that are contentious,
force is the number of events in which force is used, and violence is the number of events in
which force is used, threatened, or mobilized, or an ultimatum is issued. The country stability
variable ranges from zero to one with higher numbers indicating greater internal stability for the
country. While no countries have a value of zero for country stability, nearly 10 percent of
countries have the maximum possible value of one. In estimating regressions using country
stability as our measure of internal conflict, then, we use a Tobit model specifying an upper limit
of one on the dependent variable.
Estimating equation (1) alone provides an indication of how openness and conflict are
correlated with each other after controlling for each of the other explanatory variables.
Unfortunately, we can not interpret the coefficient on openness as measuring the effect of
openness on conflict. First, internal conflict within a country is likely to reduce international
trade flows and thus reduce the measure of openness. The governments in countries that have
had less internal conflict may also have found it easier to open their borders. In addition, there
are likely to be omitted variables that are correlated with both openness and internal conflict in a
country. Governments that have good policies in areas other than trade may also tend to have
more open economic policies. In both cases, the problem is that the openness variable is
correlated with the error term (i.e. it is endogenous). If so, we can observe a positive correlation
between openness and internal cooperation even if openness has no impact on cooperation.
12
There are two ways to solve the problem of omitted variables or endogeneity. The first is
to include country fixed effects. If the unobserved variables affecting conflict are constant over
the period of our sample (1990-2004), then country fixed effects will capture the impact of the
time-invariant omitted factors on conflict.
A second solution is to use instrumental variables regression (two-stage least squares in a
linear model). First we specify a model determining a country’s openness.
(4)
Openness = F ( past openness, conflict , control variables ) + u
The control variables affecting a country’s openness include its GDP, population, population
density, government type, distance from other economies, and level of infrastructure. Equation
(4) also includes time and continent dummy variables as well as lagged measures of the
country’s openness.
In a two-stage least squares regression, openness is first regressed on all of the exogenous
variables in equations (1) and (4), and then the predicted value from this regression is used in
estimating equation (1). To identify the effect of openness on conflict, it is necessary to include
variables in equation (4) that are excluded from equation (1). These are the instrumental
variables that affect the ratio of trade to GDP (openness) but do not directly influence the
cooperation or conflict within a country. The instruments we use for openness measure a
country’s trade infrastructure and how close geographically it is to the other major economies in
the world. There are four infrastructure variables that affect the cost of transporting goods within
the country: the navigable waterways per land area, railways per land area, highways per land
area, and airports per land area. The other instrumental variable quantifies the country’s
remoteness, which is a weighted average of the country’s distance from potential trading
partners, where the weight assigned to each trading partner is that country’s GDP as a share of
13
world GDP. None of the five instrumental variables has a statistically significant impact on
conflict if they are included in equation (1).
Equation (4) can also be estimated to determine the impact of internal conflict on a
country’s level of openness but it will suffer the same problem as equation (1) in that conflict is
likely to be endogenous. Again, however, we can solve this problem by including country fixed
effects or by using instrumental variables regression, where the instruments are the variables that
affect conflict but do not directly affect openness. The instruments we use for conflict are the
country’s prison population per 1,000 citizens, ethnic fractionalization, and the growth rate of
GDP per capita.
4. RESULTS
Table 1 shows the impact of a country’s openness on internal conflict and cooperation.
The level of conflict or cooperation within a country in columns 1 and 2 is measured using
events data. In the first column, the dependent variable is the average Goldstein score for all
events occurring within the country in each year. The dependent variable in the second column
is the country stability measure from Bond, Bond, Jenkins, and Taylor (2001). In the third and
fourth columns, the dependent variable equals one if the country was in a civil war in which
there were at least 25 battle-related deaths (column 3) or 1000 battle-related deaths (column 4) in
the year. Each of the explanatory variables is lagged by one year to prevent civil conflict in the
current year from affecting the explanatory variables in the regression. We also include a lagged
dependent variable in each of the regressions to account for the fact that conflict can persist over
multiple years so that past levels of conflict influence current levels.
14
The top half of the table treats the level of openness as an exogenous variable while the
bottom half presents instrumental variable regression results for the four models. The top half of
Table 1 shows that countries with greater openness tend to have lower levels of internal conflict.
More open countries have significantly higher average Goldstein scores among all internal
events, greater country stability, and a smaller chance of having a civil war during the year. A
one standard deviation increase in the log of openness variable raises the average Goldstein score
by 0.05 standard deviations. The final two columns show that openness is estimated to
significantly reduce the probability that a country is involved in a civil war during the year.
The conclusions change, however, when we use two-stage least squares or instrumental
variables to estimate the impact of openness on internal conflict. The bottom half of the table
shows that the coefficient on openness dramatically drops in magnitude and becomes statistically
insignificant in the first three regressions once we treat openness as endogenous. Treating
openness as endogenous reveals that openness reduces the likelihood that a country will have an
intense civil war but it does not significantly affect broader measures of stability or lower levels
of conflict within a country. These results are similar to those in Martin, Mayer, and Thoenig
(2008), who find that openness reduces the likelihood of intense civil wars but may actually
increase the probability of lower levels of internal conflict.
The last row in the table presents a test of whether it is appropriate to treat openness as
exogenous. The null hypothesis is that the variable is exogenous, so a rejection of it provides
significant evidence that instrumental variables should be used. Here the evidence is mixed.
When country stability is the dependent variable, the hypothesis that trade is exogenous is
rejected at the 10% significance level. For the other columns, however, the hypothesis that trade
is exogenous can not be rejected. If the lagged dependent variable is not included in the model,
15
however, the hypothesis that trade is exogenous is rejected at the 1% level in three of the four
regressions.
The first column also presents two tests of the instrumental variables used. The weak
identification test presents the Cragg-Donald (1993) Wald statistic. Weak instruments can create
a bias in IV estimation, and Stock and Yogo (2002) derive critical values for the Cragg-Donald
Wald statistic. In the case of column 1, a Cragg-Donald Wald statistic over 18.37 indicates less
than a 5% maximal relative bias in the instrumental variables estimation. Since the weak
instrument statistic is over 36, it easily passes the test for weak instruments. The
overidentification test presents the Sargan statistic, which tests the assumption that the
instruments are uncorrelated with the error term in the cooperation equation (a necessary
condition for the instruments to be valid). The Sargan statistic for the model in column 1 is not
significant, which means that the model passes the overidentification test.
The estimates in Table 1 suggest that richer countries and those with large prison
populations are more stable and have higher average Goldstein internal cooperation scores. The
significant negative coefficient on the polity variable and significant positive coefficient on
polity squared in the country stability regressions provides some evidence that mixed
government types are less stable than pure autocracies or pure democracies, as Hegre, Gissinger,
and Gleditsch (2003) argue. In the top half of Table 1, country stability is minimized at a polity
score of slightly less than halfway between pure autocracy and pure democracy. The large and
significant positive coefficient on the lagged dependent variable shows that internal conflict is
highly persistent over time.
Table 2 presents estimates of the effect of openness on internal conflict after controlling
for country fixed effects. The fixed effect captures the impact on conflict of any time-invariant
16
variable, so the continent dummy variables and our measure of ethnic fractionalization, which
are constant within each country, are dropped from the regression. The estimates in Table 2
confirm the results from the bottom half of Table 1. Trade does not have a significant impact on
any of the measures of internal conflict when omitted variables are controlled for using country
fixed effects. The impact of trade on intense civil wars is larger in magnitude than the OLS
estimate, however, so its statistical insignificance is caused by the large standard error (a result
of the relatively few cases of intense civil wars).
It may be reasonable to find no significant effect of openness on conflict after we have
controlled for per-capita income since one common argument is that trade leads to higher
incomes, which are associated with less internal conflict. The results in Tables 1 and 2 that
openness does not significantly affect internal conflict in the instrumental variables and fixed
effects regressions are unchanged, however, if the per-capita income and per-capita income
change variables are dropped from the regressions.
The relationship between openness and internal stability may be different in developing
and developed countries since the former tend to be less stable and less open. When the models
in Tables 1 and 2 are re-estimated using only developing countries (those not in the OECD), the
conclusions remain unchanged. Openness significantly raises internal cooperation or stability
measured using events data and lowers the chances of a civil war if openness is treated as an
exogenous variable in the regression. When instrumental variables or fixed effects regression is
used, however, openness has no significant effect on any of the measures of internal conflict in
developing countries.
While trade is the variable most commonly used in the existing literature, there are other
measures of a country’s openness. Table 3 investigates the relationship between openness and
17
internal conflict using foreign direct investment as a share of GDP as the measure of how open
an economy is. The results are similar to those in Table 1. Countries that receive larger shares
of foreign direct investment are significantly more stable and have higher average Goldstein
scores in the events data set when FDI is treated as exogenous. Unlike the results with trade
flows, however, countries with more inward FDI are not significantly less likely to have civil
wars (although the coefficients on FDI in the civil war regressions are negative). When
instrumental variables estimation is used, however, FDI openness only has a statistically
significant impact (at the 10% level) on the probability of an intense civil war occurring. If
country fixed effects are included in the regressions, then FDI has no significant impact on any
of the internal conflict measures. These results suggest that the relationship between FDI and
internal conflict is similar, but perhaps slightly weaker, than the relationship between trade
openness and internal conflict.2
One advantage of the events data is that we can examine the relationship between
openness and different types of internal conflict or cooperation. Table 4 presents estimates of the
impact of openness on the average Goldstein score among political events, military events, and
economic events. Political events relate to government actors and diplomacy. The military
category includes armed groups, weapons, and military or violent actions. Economic events are
largely policy changes or disputes on economic issues. The table shows that openness has no
significant impact on the level of political or economic conflict within the country but the OLS
estimates in column 2 show that openness significantly reduces internal military conflict. The
impact of openness on military conflict remains similar in the two-stage least squares regression,
2
We have investigated creating an index of openness that combines FDI and trade flows as a share of GDP, and we
find that the results in Tables 1-3 are largely unchanged by using this index rather than using trade or FDI measures
separately. Such an index would force the impact of a dollar in FDI on conflict to be the same as the impact of a
dollar in trade flows, however, which may not be accurate. The OLS estimates suggest that trade has an impact on
internal conflict that is between two and seven times larger than the impact of foreign direct investment.
18
although it is less precisely estimated and is no longer statistically significant. A one standard
deviation increase in logged openness raises the average Goldstein score among military events
in the country by about 0.1 standard deviations.
The results in Table 4 are consistent with the conclusion from the earlier estimates that
openness reduces only the most severe internal conflict (military clashes), but that it has no
significant effect on the less severe forms of conflict such as political or economic disputes.
Table 5 addresses the question of causality between openness and conflict by estimating
the impact of internal stability measures on the country’s openness. Once again, each
explanatory variable is lagged one year to reduce the problem that the dependent variable may
have an impact on the explanatory variables. The top half of the table presents OLS regressions
and the bottom half uses instrumental variables to estimate the effect of internal cooperation on
openness (the dependent variable is the natural log of openness). The variables used as
instruments are the lagged change in the country’s GDP per capita, ethnic fractionalization, and
prison population per capita. The table includes estimates using four measures of internal
stability: the average Goldstein score, Country Stability, and dummy variables indicating
whether or not the country is in any civil war or an intense civil war. In the first three columns,
the exogeneity test statistic is significant at the 5% level, which means that it is necessary to use
two-stage least-squares to estimate the regression.
In three of the four OLS regressions, greater internal stability leads to significantly higher
levels of openness (the coefficient on the average Goldstein score barely misses the cutoff for
statistical significance – the p-value is 0.107). The instrumental variables estimations generally
confirm this result as well: openness is significantly higher in countries with larger average
Goldstein cooperation scores, stronger country stability measures, and those without a civil war.
19
The only case in which internal conflict fails to affect openness significantly in the instrumental
variables regressions is when conflict is measured by the presence of an intense civil war. This
last result is likely caused by the weakness of the instrumental variables. Intense civil wars
occur rarely, and the three instrumental variables do a poor job of predicting their occurrence
(the low value for the weak identification test suggests that the instrumental variables estimates
suffer from bias due to weak instruments as Stock and Yogo (2002) discuss). The weak
instruments also mean that despite the fact that the coefficient on the intense civil war variable is
over three times larger in magnitude than in the OLS estimation, it is very imprecisely estimated
and is not statistically significant.
In the first three columns of the instrumental variable regressions, a rise of one standard
deviation in the internal stability measures raises a country’s openness by 0.15 to 0.18 standard
deviations. These effects are much greater in magnitude than the impact of any of the other
variables in the regressions except for the lagged dependent variable. Thus, internal conflict
measures generally have a significant and robust impact on trade as a share of countries’
economies.
The other coefficient estimates in the regressions are intuitive. Countries with more
waterways and airports per land area have greater trade shares. More populous countries are less
dependent on international trade than are smaller economies. More highways are weakly
associated with less international trade, although the coefficients are not statistically significant
in the instrumental variables regressions.
Table 6 presents estimates of the effect of internal conflict on trade openness using
country fixed effects to control for unobserved factors that influence a country’s level of
openness. As in Table 5, three of the four regressions find that greater internal conflict
20
significantly reduces a country’s openness to international trade. The estimated effect of a civil
war on openness in Table 6 is not statistically significant but is negative and very close in
magnitude to the coefficient on the civil war variable in the OLS regression in Table 5.
In all three types of regressions (OLS, instrumental variables, and fixed effects), the point
estimates reveal that more intense civil wars are associated with a larger reduction in the
country’s openness than are the less severe civil wars. The OLS estimates indicate that civil
wars with at least 25 battle-related deaths reduce external trade as a share of the economy by
about 2.4% while civil wars with at least 1000 deaths reduce openness by 4%. The instrumental
variables estimates suggest a much larger effect: any civil war is estimated to reduce openness by
24% while intense civil wars reduce it by 61%. The fixed effects estimates are close to the OLS
estimates, with any civil war reducing openness by 2.1% and intense civil wars reducing
openness by 6%.
Comparing the results in Tables 5 and 6 with those in Tables 1 and 2 suggests that the
impact of internal conflict on trade is more consistent and robust than the impact of trade
openness on conflict. Except for the case of intense civil wars, a significant effect of openness
on conflict is found only in the OLS regressions in Tables 1 and 2. When instrumental variables
or fixed effects regressions are estimated, openness is generally found to have no significant
impact on internal conflict. The estimated impact of internal conflict on trade, on the other hand,
is largely robust to the estimation method with significant coefficients on the conflict variable
found in three out of four of the regressions for each type of estimation: OLS, instrumental
variables, and fixed effects.
Another way of showing the relationship between openness and internal conflict is found
in Figures 1 and 2, which examine how changes in country stability are related to the decision by
21
countries to liberalize their economies. The liberalization variable comes from Wacziarg and
Welch (2003), who have updated the Sachs and Warner (1995) data set on dates of country
liberalization. A closed economy is defined as having at least one of the following
characteristics: an average tariff rate of 40% or more, nontariff barriers covering 40% or more of
imports, a black market exchange rate at least 20% lower than the official rate, a state monopoly
on major exports, or a socialist economic system. The Wacziarg and Welch liberalization dates
are the years in which the economy becomes permanently classified as open in the data set.
Time period zero in the figures is the most recent date at which a country liberalized its
economy, so the figure excludes the few countries that were always classified as open in the data
set and it excludes the larger number of countries that are never classified as open in the post-war
period. The line in the figure is from a nonparametric regression of country stability on the time
since liberalization.
If opening up to trade has short-run costs but long-term gains for the country in terms of
stability, as Bussman and Schneider (2007) argue, then we should observe a U-shaped curve as
country stability initially declines and then rises following liberalization. Figure 1 shows that
countries that liberalized their economies experienced a rise in stability over the first ten years
after liberalization. There follows a dip in country stability between years 10 to 17 and then a
continued increase in stability through the 28th year after liberalization. After that date, country
stability remains fairly constant. Thus, the gains from liberalization appear to be quite longlasting.
Interestingly, the countries’ stability measures begin rising about 9 years prior to its
liberalization date. This result could be explained in two ways. One is that countries liberalize
their economies gradually over time, and it takes nearly a decade on average for liberalizing
22
countries to reach the threshold set by Sachs and Warner to be counted as an open economy.
The liberalization dates may thus be marking a point in the middle of a process of opening up the
country. If the move toward openness leads to greater country stability (which is supported only
partially by the evidence in Tables 1-4), then we would see a pattern of country stability rising
before and after the official liberalization date, as in Figure 1.
A second explanation is that countries may need some level of internal stability before
they are capable of liberalizing their economies. In that case, countries that are successful in
opening up their economies will have experienced, on average, a rise in their stability in the
years preceding the liberalization. Thus, other factors which lead to a rise in stability could
allow the liberalization to take place. This explanation would be consistent with the result in
Tables 5 and 6 that more stable countries are more open.
Figure 2 shows the results of a nonparametric regression of the civil war dummy variable
on openness, and the conclusion are similar to those from Figure 1. The probability of a civil
war is estimated to drop beginning about a decade before the liberalization date until about the
eighth year after liberalization. For the next 10 years, the probability of a civil war is estimated
to rise, after which the likelihood of having a civil war declines sharply. Thus, in both figures 1
and 2 there is some evidence that the process of liberalization is a destabilizing force, as
Bussman and Schneider (2007) argue, but instability caused by liberalization does not
materialize instantly.
5. CONCLUSION
This paper has examined the relationship between openness and different levels and types
of internal conflict. We find consistent empirical evidence that instability within a country
23
reduces the international trade share of its economy. This result emerges both when we treat
internal conflict as exogenous and when we control for endogeneity using instrumental variables
or fixed effect regressions.
The impact of openness on internal conflict, on the other hand, depends on whether
openness is treated as an exogenous variable and on how the internal conflict is measured. If we
assume that openness is exogenous, as previous researchers have done, we find that openness is
associated with reduced levels of internal conflict by nearly any measure. When we use
instrumental variables or fixed effect regressions to control for endogeneity, we find no
significant impacts of openness on the probability that a country has a small civil war or on other
measures of internal stability. These conclusions do not emerge because of the specific variables
we use as instruments for openness since we get the same results using a variety of instrumental
variables and in fixed effect regressions. Examining different types of internal stability, we find
no significant impacts of openness on internal conflict over political, economic, or military
issues in instrumental variables regression.
Estimating the impact of openness on large-scale civil wars provides a different
conclusion. In both ordinary least squares and instrumental variable regression, we find that
openness significantly decreases the probability that a country will have an intense civil war.
The estimates thus support the liberal model’s prediction that more open societies are less likely
to engage in the most severe form of military conflict.
One possible explanation is that openness has both positive effects and negative effects
on internal disputes. An increase in trade is likely to raise political and economic issues that will
bring different groups into conflict, as the radical or structural paradigm predicts. Since
instability can destroy trading relationships, on the other hand, more open economies have a
24
higher opportunity cost of internal conflict, and thus greater openness may also have a pacifying
effect as the liberal paradigm argues. If these contradictory impacts largely offset one another, it
seems plausible that openness could have only a small overall impact on low-level conflict.
Only in the most severe cases of internal conflict does the pacifying effect of openness dominate.
The different results that emerge depending on how civil wars are defined and what type of
conflict is examined suggest that researchers should not limit their research agenda to only one
definition of internal conflict.
While openness is associated with fewer intense civil wars, we present evidence that the
process of economic liberalization can be destabilizing. Investigating countries’ liberalization
histories reveals that there is an increase in a country’s internal stability for the first ten years
after liberalization but then a decrease until the 17th year, when stability again begins to rise over
time. The estimates also show that countries’ internal stability begins to rise nine years before
the country liberalizes. This result is consistent with the idea that countries usually achieve some
degree of stability before they open themselves to the global market.
25
References
Alesina, Alberto, Arnaud Devleeschauwer, William Easterly, Sergio Kurlaut, Romain Wacziarg.
2003. “Fractionalization.” Journal of Economic Growth 8(2): 155-194.
Barbieri, Katherine, 2002. The Liberal Illusion: Does Trade Promote Peace? Ann Arbor, MI:
University of Michigan Press.
Barbieri, Katherine and Rafael Reuveny. 2005. “Economic Globalization and Civil War.” The
Journal of Politics, 67(4): 1228-1247.
Bayer, Resat and Matthew Rupert. 2004. “Effects of Civil Wars on International Trade, 195092.” Journal of Peace Research, 41(6): 699-713.
Blomberg, Brock and Gregory Hess. 2002. “The Temporal Links between Conflict and
Economic Activity.” Journal of Conflict Resolution, 46(1): 74-90.
Bond, Doug, Joe Bond, Churl Oh, J. Craig Jenkins, and Charles Lewis Taylor. 2001.
“Integrated Data for Events Anaysis (IDEA): An Event Form Typology for
Automated Events Data Development.” Manuscript, Harvard University.
Bond, Doug, Joe Bond, Churl Oh, J. Craig Jenkins, and Charles Taylor. 2003. “Integrated Data
for Events Anaysis (IDEA): An Event Form Typology for Automated Events Data
Development,” Journal of Peace Research 40 (6): 733-45.
Boswell, Terry and William Dixon. 1993. “Marx’s Theory of Rebellion: A CrossNational Analysis of Class Exploitation, Economic Development, and Violen
Revolt.” American Sociological Review, 58(5): 681-702.
Bussmann, Margit and Gerald Schneider. 2007. “When Globalization Discontent Turns
Violent: Foreign Economic Liberalization and Internal War.” International Studies
Quarterly, 51: 79-97.
Collier, Paul. 1999. “On the Economic Consequences of Civil War.” Oxford Economic
Papers, 51: 168-183.
Collier, Paul and Anke Hoeffler. 1998. “On Economic Causes of War.” Oxford Economic
Papers, 50: 563-573.
Collier, Paul and Anke Hoeffler. 2004. “Greed and Grievance in Civil War.” Oxford
Economic Papers, 56: 563-595.
DeSoysa, Indra. 2002. “Paradise Is a Bazaar? Greed, Creed, and Governance in Civil War,
1989-1989. Journal of Peace Research, 39: 395-416.
26
Elbadawi, Ibrahim and Nicholas Sambanis. 2002. “How Much War Will We See?
Explaining the Prevalence of Civil War.” Journal of Conflict Resolution, 46(3): 307-34.
Fearon, James and David Laitin. 2003. “Ethnicity, Insurgency, and Civil War.” American
Political Science Review, 97: 75-90.
Gleditsch, Nils Petter, Peter Wallensteen, Mikael Eriksson, Margareta Sollenberg, and Havard
Strand. 2002. Armed Conflict1946- 2001: A New Dataset.” Journal of Peace Research,
39:615-637.
Goldstein, Joshua. 1992. “A Conflict-Cooperation Scale for WEIS Events Data.” Journal
of Conflict Resolution 36 (2): 369 – 385.
Hegre, Havard, Ranveig Gissinger, and Nils Petter Gleditsch. 2003. “Globalization and Internal
Conflict.” In Globalization and Armed Conflict, eds. Gerald Schneider, Katherine
Barbieri, and Nils Petter Gleditsch. Lanham, MD: Rowan and Littlefield.
Jenkins, Craig and Doug Bond. 2001. “Conflict-Carrying Capacity, Political Crisis, and
Reconstruction: A Framework for the Early Warning of Political System
Vulnerability.” Journal of Conflict Resolution, 45(1): 3-31.
King, Gary and Will Lowe. 2003. "An Automated Information Extraction Tool for
International Conflict Data with Performance as Good as Human Coders: A Rare Events
Evaluation Design." International Organization, Vol. 57, No. 03 (July, 2003): 617 – 642.
Mansfield , Edward D. and Brian M. Pollins. Eds. 2003. Economic Interdependence and
International Conflict: New Perspectives on an Enduring Debate. Ann Arbor, MI:
University of Michigan Press.
Martin, Philippe, Thierry Mayer, and Mathias Thoenig, 2008, “Civil Wars and International
Trade,” Journal of the European Economic Association Papers and Proceedings 6 (3): 541-550.
Muller, Edward and Mitchell Seligson. 1987. “Inequality and Insurgency.” American
Political Science Review, 82(2): 425-51.
Murdoch, James and Todd Sandler. 2002. “Economic Growth, Civil Wars, and Spatial
Spillovers.” Journal of Conflict Resolution, 46(1): 91-110.
Oneal, John R. and Bruce M. Russett. 1999. “Assessing the Liberal Peace with Alternative
Specifications: Trade Still Reduces Conflict.” Journal of Peace Research 36(4): 423-442.
Polachek, Solomon W. 1980. “Conflict and Trade.” Journal of Conflict Resolution 42(1): 57-58
Reuveney, Rafael and Quan Li. 2003. “Economic Openness, Democracy and Income
Inequality: An Empirical Analysis.” Comparative Political Studies, 36(5): 575- 601.
27
Russet, Bruce M., and John R. Oneal. 2001. Triangulating Peace. Democracy, Interdependence,
and International Organization. New York: Norton.
Sachs, Jeffrey and Andrew Warner. 1995. “Economic Reform and the Process of Global
Integration.” Brookings Papers on Economic Activity, 1:1-118.
Sambanis, Nicholas. 2004. “Using Case Studies to expand Economic Models of Civil War.”
Perspectives on Politics, 2(2): 259-279.
Schneider, Gerald , Katherine Barbieri, and Nils Petter Gleditsch. Eds. 2003. Globalization and
Armed Conflict. Lanham, MD: Rowan and Littlefield.
Staiger and Stock, 1997, Instrumental Variables Regression with Weak Instruments.
Econometrica: 65(3), 557-586.
Stewart, Frances, Cindy Huang, and Michael Wang. 2001. “Internal Wars: An Empirical
Overview of the Economic and Social Consequences.” In War and Underdevelopment,
Volume 1: The Economic and Social Consequences of Conflict. Eds. Frances Stewart and
Valpy FitzGerald. Oxford: Oxford University Press, 67 – 103.
Wacziarg, Romain, and Karen Welch. 2003. “Trade Liberalization and Growth: New Evidence.”
NBER Working Paper Number 10152.
28
Table 1: Impact of trade openness on internal conflict
Assuming openness is exogenous
OLS
Variable
Average Goldstein
0.0798 **
Ln(opennesst-1)
-0.0230
Polityt-1
Polityt-1 squared
0.0011
0.1672
Populationt-1
Population densityt-1
-0.0342
0.0098 ***
GDP per capitat-1
Change in GDPPC t-1
-0.1172
0.0932
Ethnic fractionalization
Prison population ratet-1
0.0396 ***
0.5843 ***
Dependent variablet-1
America
0.0217
Europe
0.0616
Africa
-0.0563
Pacific
0.0867
Observations
R2
1835
0.4842
Tobit
Country Stability
0.0089 ***
-0.0027 **
0.0001 **
0.0185 *
-0.0018
0.0007 ***
-0.0020
0.0058
0.0047 ***
0.6250 ***
0.0031
0.0056
-0.0007
0.0084
1799
Probit
Civil War
-0.3895 ***
0.0565
-0.0019
0.3063
-0.1729
-0.0142
0.6619
0.4180 *
-0.0040
2.4662 ***
-0.6264 ***
-0.5260 **
-0.2213
-0.3915
1950
Probit
Intense Civil War
-0.4348 ***
0.0168
-0.0006
0.4731
0.0220
-0.0289
-0.5253
0.1523
0.0796
2.4535 ***
-0.3769
-0.2565
-0.0375
1890
Instrumental Variables Regression
OLS
Variable
Average Goldstein
Ln(opennesst-1)
-0.0231
Polityt-1
-0.0280 *
0.0013 *
Polityt-1 squared
Populationt-1
0.0372
-0.0065
Population densityt-1
GDP per capitat-1
0.0092 ***
-0.0919
Change in GDPPC t-1
Ethnic fractionalization
0.1038
0.0476 ***
Prison population ratet-1
Dependent variablet-1
0.5938 ***
America
-0.0052
Europe
0.0748
Africa
-0.0720
Pacific
0.0949
Observations
Weak identification test
Overidentification test
Test of trade exogeneity
1835
36.107
5.75
1.084
Tobit
Country Stability
-0.0066
-0.0035 ***
0.0002 ***
-0.0015
0.0023
0.0006 **
0.0002
0.0074
0.0059 ***
0.6445 ***
-0.0012
0.0074
-0.0036
0.0090
Probit
Civil War
-0.2362
0.0621
-0.0021
0.4851
-0.2083
-0.0136
0.6248
0.3939
-0.0172
2.4993 ***
-0.5801 ***
-0.5336 **
-0.1909
-0.4069
1799
1950
1890
3.09 *
0.16
2.68
Year dummy variables and a constant are included but not reported
*, **, *** indicate statistical significance at the 10%, 5%, and 1% levels
29
Probit
Intense Civil War
-1.3904 ***
-0.0326
0.0014
-0.7070
0.1938
-0.0289 *
-0.2527
0.2476
0.1680 **
1.8301 ***
-0.6061 **
-0.1484
-0.1761
Table 2: Impact of trade openness on internal conflict, fixed effects regressions
Linear
Variable
Average Goldstein
0.0099
Ln(opennesst-1)
0.0353
Polityt-1
Polityt-1 squared
-0.0015
0.4826
Populationt-1
Population densityt-1
-0.0941
0.0017
GDP per capitat-1
Change in GDPPC t-1
0.0497
0.0659
Prison population ratet-1
Dependent variablet-1
0.3281 ***
Observations
F-statistic for fixed effects
1885
2.64 ***
Linear
Country Stability
0.0069
-0.0002
0.0000
-0.1276
-0.0086
0.0006
0.0110
0.0077
0.3418 ***
1847
2.93 ***
Logit
Civil War
-0.0276
0.1107
-0.0039
91.6838 ***
-9.3094
0.1559
0.0656
-0.4288
1.9212 ***
693
Year dummy variables and a constant are included but not reported
*, **, *** indicate statistical significance at the 10%, 5%, and 1% levels
30
Logit
Intense Civil War
-0.6979
-0.3058
0.0117
6.5570
2.2419
-0.1833
-0.2387
-0.4935
3.1676 ***
327
Table 3: Impact of FDI on internal conflict
Assuming FDI is exogenous
OLS
Variable
Average Goldstein
0.0381 ***
Ln(FDIt-1)
-0.0333 **
Polityt-1
Polityt-1 squared
0.0015 **
0.0814
Populationt-1
Population densityt-1
-0.0188
0.0074 ***
GDP per capitat-1
Change in GDPPC t-1
-0.3375
0.0107
Ethnic fractionalization
Prison population ratet-1
0.0423 ***
0.5812 ***
Dependent variablet-1
America
-0.0156
Europe
0.0338
Africa
-0.0381
Pacific
0.0324
Observations
R2
1705
0.4825
Tobit
Country Stability
0.0036 ***
-0.0036 ***
0.0002 ***
0.0086
0.0000
0.0004 *
-0.0170
0.0001
0.0054 ***
0.6182 ***
-0.0011
0.0036
-0.0009
0.0045
1672
Probit
Civil War
-0.0574
0.0549
-0.0017
0.6996 **
-0.1619
-0.0090
0.8021
0.4236
-0.0137
2.5162 ***
-0.4152 **
-0.4229 **
-0.0917
-0.3446
1801
Probit
Intense Civil War
-0.0651
0.0490
-0.0018
0.9004 **
-0.0813
-0.0207
-1.4972
0.3649
0.0461
2.4870 ***
-0.3354
-0.2511
-0.0605
-0.0651
1743
Instrumental Variables Regression
OLS
Variable
Average Goldstein
Ln(FDI-1)
-0.0353
Polityt-1
-0.0365 **
0.0017 **
Polityt-1 squared
Populationt-1
0.0520
0.0165
Population densityt-1
GDP per capitat-1
0.0093 ***
-0.2719
Change in GDPPC t-1
Ethnic fractionalization
0.0439
0.0410 ***
Prison population ratet-1
Dependent variablet-1
0.6038 ***
America
0.0678
Europe
0.1176
Africa
-0.0076
Pacific
0.1367
Observations
Weak identification test
Overidentification test
Test of FDI exogeneity
1705
7.024
4.856
0.816
Tobit
Country Stability
-0.0079
-0.0041 ***
0.0002 ***
0.0032
0.0054
0.0007 **
-0.0070
0.0058
0.0051 ***
0.6564 ***
0.0118
0.0166
0.0034
0.0202
Probit
Civil War
-0.1041
0.0525
-0.0016
0.6853 *
-0.1382
-0.0076
0.8343
0.4496
-0.0129
2.4946 ***
-0.3661
-0.3721
-0.0773
-0.2715
1672
1801
1743
2.06
0.01
1.06
Year dummy variables and a constant are included but not reported
*, **, *** indicate statistical significance at the 10%, 5%, and 1% levels
31
Probit
Intense Civil War
-0.5699 *
0.0038
0.0004
0.5416
0.1789
-0.0008
-0.6971
0.5528 *
0.0478
1.5192
0.3064
0.3784
0.1228
0.0038
Table 4: Impact of trade openness on political, military, and economic conflict
Assuming openness is exogenous
Variable
Ln(opennesst-1)
Polityt-1
Polityt-1 squared
Populationt-1
Population densityt-1
GDP per capitat-1
Change in GDPPC t-1
Ethnic fractionalization
Prison population ratet-1
Dependent variablet-1
America
Europe
Africa
Pacific
Observations
R2
Political events
0.0135
0.0044
-0.0003
-0.0189
-0.0353 *
0.0043 **
-0.2297
0.0429
0.0107
0.3689 ***
-0.0417
-0.0109
-0.1043 ***
-0.0401
Military Events
0.3833 ***
-0.0928 **
0.0040 *
-0.4298
-0.1758 *
-0.0096
0.5827
0.2304
0.0165
0.2760 ***
-0.0917
-0.1814
0.3444 **
0.2884
Economic events
0.0093
-0.0118
0.0004
-0.2731
-0.0170
-0.0145 ***
-0.4760
0.2458 **
0.0032
0.1035 ***
-0.0252
-0.0358
-0.1789 **
-0.2171
1835
0.2051
1835
0.1264
1835
0.0475
Military Events
0.4099
-0.0914 *
0.0040 *
-0.3970
-0.1829
-0.0095
0.5749
0.2273
0.0141
0.2752 ***
-0.0849
-0.1852
0.3492 **
0.2863
Economic events
0.0421
-0.0100
0.0003
-0.2317
-0.0258
-0.0143 ***
-0.4859
0.2418 *
0.0003
0.1030 ***
-0.0165
-0.0404
-0.1732 **
-0.2200
Instrumental Variables Regression
Variable
Ln(opennesst-1)
Polityt-1
Polityt-1 squared
Populationt-1
Population densityt-1
GDP per capitat-1
Change in GDPPC t-1
Ethnic fractionalization
Prison population ratet-1
Dependent variablet-1
America
Europe
Africa
Pacific
Political events
0.0140
0.0044
-0.0003
-0.0183
-0.0355
0.0043 **
-0.2299
0.0428
0.0107
0.3689 ***
-0.0416
-0.0110
-0.1042 ***
-0.0402
Observations
R2
Weak identification test
Overidentification test
Test of trade exogeneity
1835
1835
1835
35.751
2.455
0.000
34.882
1.085
0.008
35.875
2.720
0.038
Year dummy variables and a constant are included but not reported
*, **, *** indicate statistical significance at the 10%, 5%, and 1% levels
32
Table 5: Impact of internal conflict on openness
Assuming conflict is exogenous
Variable
Average Goldstein scoret-1
Country Stability measuret-1
Civil wart-1
Polityt-1
Polityt-1 squared
Remotenesst-1
Population densityt-1
Waterwayst-1
Railwayst-1
Highwayst-1
Airportst-1
Ln(GDPt-1)
Ln(Populationt-1)
Ln(opennesst-1)
Observations
R2
Goldstein
0.0105
Country stability
Civil war
Intense civil war
0.1574 **
0.0030
-0.0001
-0.0009
0.0278
0.8834 **
0.1849
-0.0198 *
7.8606
0.0074
-0.0344 ***
0.8432 ***
1832
0.8559
0.0033
-0.0001
-0.0011
0.0275
0.8871 **
0.1700
-0.0196 *
7.9306
0.0068
-0.0341 ***
0.8410 ***
1808
0.8551
-0.0241 *
0.0021
-0.0001
-0.0010
0.0223
0.7976 **
0.1563
-0.0194 *
9.3522
0.0086
-0.0325 ***
0.8527 ***
-0.0401 *
0.0015
0.0000
-0.0010
0.0243
0.7644 **
0.1772
-0.0184 *
8.2136
0.0090
-0.0331 ***
0.8536 ***
1947
0.8624
1947
0.8624
Civil war
Intense civil war
Instrumental variables regression
Variable
Average Goldstein scoret-1
Country Stability measuret-1
Civil wart-1
Polityt-1
Polityt-1 squared
Remotenesst-1
Population densityt-1
Waterwayst-1
Railwayst-1
Highwayst-1
Airportst-1
Ln(GDPt-1)
Ln(Populationt-1)
Ln(opennesst-1)
Observations
Weak identification test
Overidentification test
Test of conflict exogeneity
Goldstein
0.0941 *
Country stability
1.0737 **
0.0083
-0.0003
-0.0012
0.0310
0.9796 ***
0.0975
-0.0143
8.4299
-0.0116
-0.0195
0.8174 ***
1832
11.471
1.751
3.198 *
0.0105 *
-0.0005
-0.0027
0.0231
0.9697 ***
0.0177
-0.0148
10.6137
-0.0111
-0.0215 *
0.8119 ***
-0.2679 **
0.0095
-0.0004
-0.0008
-0.0009
0.8990 **
-0.2664
-0.0171
20.9788 **
-0.0046
-0.0098
0.8250 ***
1808
17.072
0.725
3.929 **
1947
7.527
1.242
3.863 **
Year and continent dummy variables are included but not reported
*, **, *** indicate statistical significance at the 10%, 5%, and 1% levels
33
-0.9456
0.0046
-0.0002
-0.0006
0.0170
0.2417
-0.2935
0.0090
8.4570
-0.0106
0.0045
0.8102 ***
1947
0.865
1.912
2.288
Table 6: Impact of internal conflict on openness
Fixed Effects regression
Variable
Average Goldstein scoret-1
Country Stability measuret-1
Civil wart-1
Polityt-1
Polityt-1 squared
Remotenesst-1
Population densityt-1
Waterwayst-1
Railwayst-1
Highwayst-1
Airportst-1
Ln(GDPt-1)
Ln(Populationt-1)
Ln(opennesst-1)
Observations
F-statistic for all fixed effects
34
Goldstein
0.0142 **
Country stability
Civil war
Intense civil war
-0.0603 **
-0.0243 ***
0.0012 ***
0.0027
0.0831
0.8723
0.0587
0.0065
-21.8737
-0.0578 *
-0.2341 **
0.5289 ***
2019
3.19 ***
0.2012 **
-0.0220 ***
0.0011 ***
0.0018
0.0846
0.3079
0.0771
0.0060
-21.3256
-0.0721 **
-0.2470 **
0.4879 ***
-0.0226 ***
0.0011 ***
0.0053
0.0795
0.2688
0.0769
0.0068
-21.8743
-0.0505
-0.2453 **
0.4897 ***
-0.0213
-0.0235 ***
0.0012 ***
0.0030
0.0840
0.6954
0.0526
0.0051
-23.2223
-0.0536
-0.2340 **
0.5300 ***
1901
3.42 ***
1864
3.41 ***
2019
3.17 ***
Figure 1: Country stability and years since liberalization
1
0.98
Country Stability
0.96
0.94
0.92
0.9
0.88
0.86
0.84
-12 -8
-4
0
4
8
12 16
20 24 28
32 36
Years since liberalization
35
40 44
48 52
Figure 2: Civil wars and years since liberalization
0.4
Probability of a civil war
0.35
0.3
0.25
0.2
0.15
0.1
0.05
0
-12 -8
-4
0
4
8
12 16
20 24 28
32 36
Years since liberalization
36
40 44
48 52