Breaking the curse of Sisyphus: an empirical analysis of post-conflict economic transitions.

Cevik, Serhan ; Rahmati, Mohammad

INTRODUCTION

Since World War II, intra-state armed conflicts have become the most common type of aggression across the world according the UCDP/PRIO Armed Conflict Dataset (Figure l). (1) Countries emerging from such a conflict struggle to break the curse of Sisyphus, with 20% relapsing into a subsequent conflict in the first year and 40% within 5 years after the end of conflict. Hence, understanding the causes and consequences of civil conflict remains an important objective in the quest to achieve sustainable sociopolitical reconciliation and economic development. This paper makes a novel contribution to the literature by estimating a forward-looking measure of conflict recurrence risk and including it as an explanatory variable in analyzing post-conflict growth dynamics. We adopt a logistic regression approach and model the probability of conflict recurrence with the latest available information in each period, while controlling for unobserved fixed effects in a non-linear probability model for the first time in the conflict literature.

[FIGURE 1 OMITTED]

According to the UCDP/PRIO Armed Conflict Dataset covering 146 episodes of civil conflict in 94 countries over the period 1960-2010, real GDP per capita growth accelerated after the end of a conflict, with an average of 1.5% a year over the subsequent 5 years, and shows significant cross-country variation, ranging from--10% in Liberia to 20% in Bosnia and Herzegovina. Our empirical model explains these variations in post-conflict growth through a broad set of explanatory variables including the estimated risk of conflict recurrence, duration and intensity of the conflict, demographic characteristics, measures of human capital, geographic variables and natural resource endowments, macroeconomic conditions, external factors, and measures of political regime and governance quality. The key empirical finding presented in this paper is that the estimated risk of conflict recurrence has a statistically significant and economically large effect on post-conflict growth dynamics. A 1% increase in the probability of conflict recurrence in the subsequent period is associated with lower real GDP per capita growth by 10%.

The other main findings are broadly in line with the literature. Human capital and institutions have positive and statistically significant effects on post-conflict growth, while natural resource dependence appears to cause a double curse of higher risk of conflict recurrence and lower growth, compared with other countries with a more diversified economic structure. Ethnic fractionalization is another significant factor with negative influence on the strength of post-conflict growth. Global economic conditions matter for post-conflict recovery, especially for countries that are more open and integrated with the rest of the world. Finally, contrary to popular expectations, we do not find UN peacekeeping operations, foreign aid and IMF-supported economic programs to have significant effect on post-conflict growth. Altogether, the empirical results presented in this paper call for a framework centered on three main pillars, implementing growth-enhancing policies, reforming dysfunctional institutions, and addressing urgent needs, to reduce the risk of conflict recurrence and pave the way to broad-based, sustainable economic growth after the end of a conflict.

A BRIEF OVERVIEW OF THE LITERATURE

This paper connects two broad strands of the literature: (1) the cross-country differences in economic development and (2) the causes and consequences of civil conflict. The vast literature on the determinants of economic growth concludes that there is no single cause of cross-country variations in growth performance. Instead, a wide spectrum of factors, from geography to factors of production to natural resources to institutions, and the interactions between various factors influence economic growth. In particular, Knack and Keefer (1995), Hall and Jones (1999), Engerman and Sokoloff (2000), Acemoglu et al. (2001), Acemoglu et al. (2004), and Acemoglu and Johnson (2005), among other scholars, demonstrate that institutions are one of the most important sources of cross-country differences in economic growth. However, while the quality of institutional arrangements is among the fundamental determinants of cross-country differences in macroeconomic performance, institutions may well be shaped by a broad range of factors including geographic conditions, demographic composition, and the structure of factor endowments. A number of studies, led by Sachs and Warner (1997) and Gallup et al. (1998), take into account geographic variables and find that countries located in the tropics tend to grow at a slower pace than those in more temperate climates.

The theoretical literature has focused mainly on the pivotal question of why civil conflict occurs in spite of the prohibitively high cost of rebellion, and draws attention to explanations like information asymmetries and the absence of the rule of law (Grossman, 1991; Skaperdas, 1992; Hirshleifer, 1995; Sandler, 2000; Sambanis, 2001; Fearon and Laitin, 2003; Collier and Hoeffler, 2004). On the empirical side, a burgeoning range of studies has aimed to discover the causes of civil conflict that have persistent and damaging effects on economic development, mainly through the destruction of institutional and physical infrastructure, the loss of human capital, and heightened uncertainty making investment riskier (Collier, 1999; Nafzinger et al, 2000; Addison et al, 2002; Collier and Hoeffler, 2002; Elbadawi et al, 2008).

There is also a large literature on the relationship between political institutions and civil conflict. Sambanis (2001), Hegre et al. (2001) and Reynal-Querol (2002, 2005), for example, find that partial democracies are more prone to civil conflict than are full democracies and autocracies. Ethnic fractionalization is also considered to be an important factor in explaining the cross-country differences in growth as well as the variation in social tensions heightening the risk of conflict. While Fearon and Laitin (2003) argue that ethnic, or religious, heterogeneity has no bearing on the incidence of civil conflict, after controlling for per capita income, Esteban et al (2012) show that measures of ethnic and social fragmentation are significant correlates of the occurrence of civil conflict. These issues are also intimately connected to the issue of trust and its impact on post-conflict economic activity (Cassar et al, 2013; Grosjean, 2014). On the whole, the literature links the incidence of civil conflict to low levels of income with a skewed distribution, slow and volatile growth, dysfunctional institutions, ethnic fragmentation, and demographic and geographic conditions. (2)

DATA OVERVIEW

The empirical analysis utilizes a comprehensive set of conflict data--encompassing 146 cases of civil conflict in 94 countries across the world over the period 1960-2010--extracted from the latest version of the UCDP/PRIO Armed Conflict Dataset. The UCDP/PRIO dataset characterizes four different types of conflicts according to location and participants: (1) extra-systemic armed conflicts, (2) inter-state armed conflicts, (3) intra-state armed conflicts, and (4) internationalized intra-state armed conflicts. In this paper, we concentrate on episodes of internal armed conflict including all three categories based on the level of intensity: low-intensity conflict with at least 25 battle-related deaths a year and fewer than 1,000 battle-related deaths during the course of the conflict; cumulatively high-intensity conflict with an accumulated total of at least 1,000 battle-related deaths, but fewer than 1,000 in any given year; and high intensity conflict with at least 1,000 battle-related deaths a year. Although low-intensity conflicts may not necessarily have a measurable impact on economic activity, we still include such episodes in the sample to develop a broader understanding of civil conflicts.

In addition to our estimated measure of the risk of conflict recurrence, the growth regressions consider a broad range of variables: these include the initial level of real GDP per capita at the end of the conflict; human capital as measured by educational attainments; demographic variables such as population density, the size of the youth bulge, and ethnic fractionalization; geographic variables and climatic conditions; natural resource dependence; terms of trade; global real GDP growth; degree of trade openness; international factors such as foreign aid, IMF-supported programs, and UN peacekeeping operations; and institutional variables such as the type of political regime and the number of cabinet changes.

The majority of the tests for unit root or stationarity in panel data that are widely used in the empirical literature assume a balanced panel dataset (Levin et al, 2002; and Im et al, 2003). In our study, however, the panel dataset of post-conflict growth episodes is unbalanced because of frequent occurrences of civil conflict, making most of these tests for unit root in panel data inapplicable. One prominent exception is the approach proposed by Choi (2001) that introduce a Fisher-type unit root test for even unbalanced panel datasets. The results, available upon request, show that there is no evidence for non-stationarity in our panel dataset.

EMPIRICAL METHODOLOGY AND RESULTS

To avoid the potential endogeneity of the explanatory variables and unobserved country-specific effects, Holz-Eakin et al. (1988) and Arellano and Bond (1991) developed the generalized method of moments (GMM) estimator for dynamic panel regressions using the lagged levels of the regressors as instruments for the equation in differences. This approach, however, suffers from small sample bias when the time series is short and exhibits high persistency over time, which makes the lagged levels weak instruments for subsequent differences. These shortcomings of the difference-GMM estimator--documented by Arellano and Bover (1995) and Blundell and Bond (1998)--are more pronounced when a approaches one and/or [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] high. Although using multiple-year intervals or averages has become a standard procedure in the empirical literature, this reduces the number of observations and does not effectively mitigate persistency in the data. Furthermore, if the variance of country-specific characteristics increases relative to transitory shocks, the standard method produces weak results. Because of the dominant impact of unobserved heterogeneities in growth during and after a conflict, it is likely that [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] will be greater than [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII]. In addition, post-conflict data are relatively short, partly because of a series of conflict recurrences.

As the difference-GMM estimator is not a suitable approach to studying post-conflict economic growth, we follow the system-GMM approach proposed by Arellano and Bover (1995) and Blundell and Bond (1998), which brings together the equation in differences and the equation in levels in a system of equations and utilizes both lagged levels and differences of the regressors as instruments. Accordingly, the system-GMM estimator performs better with highly persistent data and allows us to obtain more precise estimates. We operationalize a modified version of the standard neoclassical growth model in the following form:

[DELTA][y.sub.it] = [g.sub.it] = [alpha][g.sub.it-1] + [beta]C[R.sub.it] + [gamma][Z.sub.it] + [[eta].sub.i] + [[epsilon].sub.it] (1)

where [DELTA][y.sub.it] is the growth rate of real GDP per capita for country i and t periods after the end of the conflict (ie time t = 0 denotes the end of conflict); [g.sub.i,t-1] lis the lagged growth rate at the time t--1; C[R.sub.it] is the risk of conflict recurrence over the subsequent period after the end of the conflict, which we estimate according the below-outlined approach; [Z.sub.it] represents a vector of variables; [[eta].sub.i] are unobserved country-specific effects; and [[epsilon].sub.it] is the error term. The empirical method used in this paper makes following assumptions:

E[[[eta].sub.i]] = 0, E[[[epsilon].sub.it]] = 0, E[[[eta].sub.i][[epsilon].sub.it]] = 0 (2)

We also assume that transitory shocks are serially uncorrelated (ie E[[[epsilon].sub.it][[epsilon].sub.i,s]] = 0 for all s[not equal to]t) and initial conditions are predetermined (ie E[[g.sub.i1][[epsilon].sub.it]] = 0). These assumptions, in turn, provide L= (T-l)(T-2)/2 moments conditions to estimate Equation 1:

E[[g.sub.i,t-s][DELTA][[epsilon].sub.it]] = O for t = 3, ..., T and s [greater than or equal to] 2 (3)

The assumption of E([u.sub.i3][DELTA][g.sub.i2]) = 0 improves the performance of the estimator. Using the standard GMM assumption, this condition can be reduced to E([[eta].sub.i] [DELTA][g.sub.i2])--0, which is satisfied for stationary series as long as E([g.sub.iI]\[[eta].sub.i]) = [c.sub.i]. This implies that deviations from the initial conditions are uncorrelated with [c.sub.i], itself. Although Bond et al. (2001), among others, employ this methodology to estimate the level of GDP, assuming E([[eta].sub.i][DELTA][y.sub.i2]) = 0, we believe this condition may be violated in estimating post-conflict growth dynamics, as it implies that the initial rate of growth ([DELTA][y.sub.i2]) is unrelated to unobserved fixed effects. On the other hand, assuming the change in post-conflict growth is orthogonal to [[eta].sub.i] provides a weaker assumption for post-conflict recoveries. We therefore use the above-outlined moment conditions and employ the system-GMM methodology to reduce the potential bias in estimation and generate consistent and efficient parameter estimates.

ESTIMATING THE RISK OF CONFLICT RECURRENCE

The principal contribution of this paper is the estimation of a forward-looking measure of conflict recurrence risk and its use as an explanatory variable in estimating the determinants of post-conflict economic performance. Quantifying the relationship between the probability of conflict recurrence and the strength of post-conflict growth is a challenge. There are different modeling approaches in the early warning/leading indicators literature; these include linear regression and limited dependent variable probit/ logit techniques developed to test the statistical significance of various indicators in determining the incidence of an event risk. (3) In this paper, we adopt a logistic regression approach and model the risk of conflict recurrence with the latest available information in each period, while controlling for unobserved fixed effects in a nonlinear probability model for the first time in the literature.

The risk of an event at time t+1 can be measured by estimating its distribution given the available information at time t. Measuring the risk of conflict recurrence over the subsequent years by using proxy variables would yield biased estimates mainly because there is no good proxy to convey forward-looking risk expectations, especially in many developing and low-income countries that have experienced episodes of civil conflict. Therefore, we model the risk of conflict recurrence with the latest available information in each period by imposing more structure on a general model of conflict and defining risk as the probability of conflict recurrence in the next period. Assuming conflict is a latent variable, the model of hostility in country i at time t ([h.sub.it]) becomes as a function of:

[h.sub.it] = [phi][X.sub.it] + [v.sub.i] + [[epsilon].sub.it] (4)

where [X.sub.it] is a set of independent variables contributing to the level of enmity at time t in country i and [v.sub.i] represents fixed effects for country i. Although the level of hostility is unobservable, we can still estimate Equation 6 since conflict incidence is an observable variable, assuming that it happens when the level of hostility passes a certain threshold: (4)

[cr.sub.it] = 1 if [h.sub.it] > 0 and [cr.sub.it] = 0 if [h.sub.it]<0 (5)

If [epsilon] has a distribution of [DELTA], a conflict occurs with the probability of 1-[DELTA] ([phi][X.sub.it]+[v.sub.i]). These binary observations provide enough variation to estimate [beta] in Equation 1 - a well-established exercise in the literature for characterizing the causes of civil conflict. At the same time, the forward-looking probability of conflict recurrence can be defined as:

E[[r.sub.it]] = probability ([cr.sub.i,t+1] = 1|[X.sub.it]: [phi], [v.sub.i]) (6)

On the basis of this framework, the perceived risk of conflict recurrence can be calculated as:

[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] (7)

Therefore, risk can be measured by the probability of a conflict at the next period given the information set available at time t in country i ([X.sub.it]), and the above equation can be estimated by using the following likelihood estimator:

[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] (8)

One major advantage of our method, compared to earlier studies, is that we control for unobserved fixed effects ([v.sub.i]) and therefore deal with unobserved variables that may contribute to the incidence of relapsing into subsequent conflict. Following Wooldridge (2005), assuming that the distribution of unobserved fixed effects [v.sub.i] is by a distribution [OMEGA](.;[pi]) the likelihood estimator becomes a function of:

[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] (9)

For this exercise, we assume [bar.[DELTA]] has the logistic functional form as exp ([phi][X.sub.it] + [[upsilon].sub.i])/1 + exp([phi][X.sub.it] + [[upsilon].sub.i]) and [[upsilon].sub.i] is distributed N(0, [[sigma].sup.2.sub.[upsilon]]). It is worth mentioning that a significant and economically large [[sigma].sub.v] shows that ignoring these unobserved fixed effects may lead to biased results.

Our baseline model for the risk of conflict recurrence utilizes the level and growth rate of real GDP per capita and whether the country had a civil conflict before. The estimated coefficients for Equation 9, presented in the first column of Table 1, indicate that the variance of unobserved fixed effects ([[sigma].sub.[upsilon]) is significant in all cases. In other words, country-specific effects are crucial in measuring conflict recurrence risk. Another important conceptual distinction between this paper and earlier studies is the timing of the model, as we infer the risk of conflict recurrence at time t+1 by using the available information set at time t, which is an abstraction and not burdened by the problem of endogeneity often present in conflict regressions. To check the robustness of our estimations, we develop an alternative measure of conflict recurrence risk without relying on the level and growth of domestic per capita income, presented in the second column of Table 1, and find that the estimated risk of conflict recurrence remains robust.

According to the baseline model, the coefficients on the level and growth rate of real GDP per capita are -0.32 and -1.72, respectively, which are equivalent to the marginal effects of -0.02 and -0.11. In other words, a 1% increase in the level of real GDP per capita or in its rate of growth would lower the probability of conflict recurrence in the subsequent period by 2 % and 11%, respectively, in that country. One of the key factors in determining the risk of conflict recurrence, however, is whether the country had a conflict before. The coefficient on the 'had conflict' variable is statistically significant across all specifications. The marginal effect of having a history of conflict increases the probability of conflict recurrence in the subsequent period by 35%. (5)

The impact of human capital on conflict recurrence risk is statistically significant and exhibits a non-linear behavior. The squared term of secondary school enrollment rate indicates that the risk of conflict recurrence starts to increase once a country reaches a certain threshold in human capital. (6) In other words, raising educational attainments from a low base has a stabilizing effect, as it helps to increase the productivity of labor and lower the opportunity cost of rebellion. But when a country is relatively well-endowed with human capital, further gains would add to labor market pressures in a fragile post-conflict economy and consequently increase the risk of conflict recurrence.

We also include a measure of the youth bulge and find that it is a significant factor in explaining the risk of conflict recurrence.

The magnitude of natural resource dependence has a significant effect on the risk of conflict recurrence, in line with previous studies (Collier and Hoeffler, 2004; Ross, 2006). In other words, the larger the extent of natural resource rents in a country, the higher the risk of conflict recurrence after the cessation of violence. (7) We also include the squared term of natural resource rents, which is weakly significant only in one specification but still provides evidence that an increase in natural resource rents reduces the probability of future conflicts in higher-income countries. Likewise, external factors, such as changes in the terms of trade, appear to have a significant influence on the risk of conflict recurrence.

The type of political regime has a weakly significant effect that tends to increase the risk of conflict recurrence in democratic countries, but the relationship is not linear. The negative coefficient on the squared term of political regime type indicates that a competitive election process and executive constraints reduce the probability of conflict recurrence in more democratic societies, while a high degree of institutionalized autocracy would also lower the risk of conflict recurrence. In other words, countries in the middle of the political regime spectrum appear to have a higher perceived risk of conflict recurrence. On the other hand, the coefficient on the number of cabinet changes is negative and highly significant, implying that frequent changes in government contribute to a higher risk of conflict recurrence.

Using a binary variable for UN peacekeeping operations, we identify a positive and statistically significant relation with the risk of conflict recurrence. In other words, the presence of UN peacekeeping forces appears to increase the risk of conflict recurrence. This finding remains robust even if alternative measures of UN interventions, such as the number of UN troops in the country, are included instead of a binary variable. Although, it appears to be counterintuitive, this result can be explained by the endogeneity of peacekeeping operations: the UN initiates an operation when risk is apparent in generating chaos in an attempt to support the development of an environment of economic and political stability and social cohesion, confirming the perceived risk of conflict recurrence. We therefore exclude this variable from the benchmark model, since the endogeneity problem can result in biased estimates. Finally, we include decade dummies to capture shifting political relationships between countries during and after the Cold War, and find that decade dummies, except for the 1960s, are statistically significant in most specifications of the model.

[FIGURE 2 OMITTED]

According to our estimations, the probability of recurrence exceeds 20% in the first year after the end of conflict and exceeds 40% in the 5-year period. While the risk of relapsing into conflict declines with a lasting peace, it is remains higher in the subsequent years than the pre-conflict probability. As shown in the right panel of Figure 2, the cumulative distribution of conflict recurrence by region indicates extensive heterogeneity among countries and highlights the importance of taking into account country-specific effects in the empirical analysis. In addition, there is a systematically higher risk of conflict recurrence in Africa, Asia, and the Middle East than in the Americas and Europe. In particular, African and Asian countries start peace with a significantly higher risk of relapsing into conflict, although the probability declines over time with a lasting peace. In contrast, however, countries in the Middle East appear to suffer from a high risk of conflict recurrence without much decline over subsequent years.

In Figure 3, the average risk of conflict occurrence in countries with at least one conflict (left panel) and no history of civil conflict (right panel) across the world is depicted. The probability of civil conflict peaked for all countries in the 1990s, increasing from an average of 4% in the 1960s to 10% in the 1970s, 15% in the 1980s, and 16% in the 1990s. (8) However, the risk of civil conflict incidence is more than 100 times greater in countries with at least one past episode of civil conflict: it was 22% in the 1990s for countries with a history of civil conflict versus 0.06% for countries with no history of civil conflict. Although, the risk of conflict occurrence declined over the past decade for all countries across the world, the probability of relapsing into a new bout of conflict remains elevated for countries with a history of conflict. As illustrated in Figure 4, the average risk of conflict recurrence in the 2000s is highly correlated with that in the 1960s, indicating a highly persistent risk of conflict recurrence in countries with a history of conflict.

[FIGURE 3 OMITTED]

[FIGURE 4 OMITTED]

ESTIMATING POST-CONFLICT ECONOMIC PERFORMANCE

We present the results of the cross-country growth regressions in Table 2 for the baseline model and in Table 3 for the model with the alternative definition of conflict recurrence risk, which is discussed in detail in assessing the robustness of the results. Our findings show that the risk of conflict recurrence has a statistically significant and economically large effect on post-conflict growth. According to our estimations, a 1% increase in the probability of conflict recurrence would lower real GDP per capita growth by 10%. (9) The magnitude of this effect is substantial, especially considering that a country with a history of conflict would be 30% riskier than a country with no history of conflict. Similarly, alternative specifications of the risk of conflict recurrence, presented in the second column of Table 1, also have a negative effect of the same magnitude on post-conflict growth.

Conflict duration has a significant negative effect on post-conflict growth, which may reflect an amalgamation of interacting factors. On the one hand, a long conflict tends to destroy the stocks of human and physical capital, the restoration of which would lead to rapid recovery because of higher marginal productivity of capital. On the other, a long-lasting conflict can damage the institutional infrastructure and undermine post-conflict recovery. The results support the latter hypothesis. That is, a long conflict, controlling for conflict recurrence risk, undermines post-conflict growth. (10)

Human capital has a positive effect on post-conflict growth, but secondary school enrollment is not significant in all specifications. With the alternative definition of conflict recurrence risk as presented in the second column of Table 1, the impact of secondary school enrollment rates becomes statistically insignificant. This may be because of the asymmetric destruction of different types of capital during a conflict. If the conflict destroys the stocks of human capital more than that of physical capital, then, in the aftermath of the conflict, human capital becomes scarce compared with other factors of production. The lower the ratio of human to physical capital, the higher the marginal product of investing in human capital, and consequently the higher the growth.

Institutional measures, such as the type of political regime and the number of cabinet changes in a given year, have the expected signs, but do not appear to be robustly significant, when the risk of conflict recurrence is included in the model. However, we are reluctant to draw definite conclusions about the relationship between growth and institutional variables. First, the analysis covers post-conflict periods, which may be too short for institutional change to have a meaningful impact on growth dynamics. Second, depending on the definition of conflict recurrence risk or other factors included as explanatory variables, the significance of institution variables varies across model specifications. When the estimation is conducted using a broader set of observations that includes non-conflict countries, institutional factors indeed become statistically and economically highly significant with the expected signs. In our opinion, this reflects numerous institutional similarities among conflict-prone countries; therefore, excluding observations from non-conflict countries makes the empirical identification of the impact of institutional variables on post-conflict growth weaker.

Natural resource dependence has an economically large negative effect on post-conflict growth. Post-conflict countries with a high degree of dependence on natural resources suffer from a 'double curse' of higher risk of conflict and lower growth, compared with other post-conflict countries with a more diversified economic structure. While climatic conditions, measured by temperature variability, do not appear to be a significant factor, ethnic fractionalization has a statistically significant negative effect on the pace of post-conflict growth. Moreover, the impact of ethnic fractionalization remains statistically significant and economically meaningful, even when we control for the risk of conflict recurrence.

International economic factors such as global growth and net per capita FDI are statistically significant and have an economically large positive impact on the post-conflict growth process. Although, we find a negative coefficient for trade openness, interaction terms relating trade openness to the size of the domestic economy and global real GDP growth are highly statistically significant. In other words, as expected, more open conflict-prone economies tend to more sensitive to global developments. Foreign aid, on the other hand, does not appear to be important. This may reflect the pattern experienced in many post-conflict countries: foreign aid surges after the cessation of hostilities but abates thereafter. Moreover, as Collier and Hoeffler (2002) argue, post-conflict countries tend to have low capacity to absorb foreign aid because of political and administrative constraints; although, Suhrke et al. (2005) challenge these findings. (11)

Contrary to popular expectations, UN peacekeeping operations do not appear to be a significant factor in determining the strength of post-conflict growth. This, we believe, is related to the relationship between the risk of conflict recurrence and international peacekeeping operations, which may signal a greater level of risk, which in turn depresses economic activity. As a 11 result, the coefficient of UN peacekeeping operations becomes sensitive to the definition of conflict recurrence risk and makes it difficult to disentangle these linkages and their impact on growth. We also test the impact of IMF-supported economic programs on post-conflict growth dynamics. Although, the empirical evidence does not support a positive relationship between IMF engagement and real GDP per capita growth after the cessation of violence, such lending arrangements do not necessarily deal with the underlying fragilities that remain a drag on post-conflict growth.

ROBUSTNESS CHECKS

One important concern about the empirical findings presented in this paper is related to the definition of conflict recurrence risk. The baseline model is estimated with a set of variables including the level and growth rate of real GDP per capita, which are also included in analyzing post-conflict growth dynamics. Consequently, there may be a potential problem of endogeneity. One possible way to deal with this econometric predicament is to implement an instrumental variable approach for a linear probability model, similar to the methodology outlined by Miguel et al. (2004). However, a shortcoming of this methodology is that the linear prediction for the probability of an event risk produces many negative risk observations in the growth regression and may result in biased estimates. Furthermore, while Miguel et al. (2004) use rainfall as an instrumental variable in a limited geographic scope of African countries, our sample includes a broader set of countries, making instrumental variables such as rainfall a weak option for the analysis.

Instead, we estimate alternative measures of conflict recurrence risk using a variety of variables such as a country's history of civil conflict, the level and growth rate of global real GDP, the frequency of cabinet changes, and the degree of ethnic fractionalization, without including the level and growth of domestic per capita income in the model. The results indicate that having a history of conflict is a robust factor in explaining the recurrence of intra-state wars. Furthermore, other variables such as the youth bulge, the duration of the last conflict, and the frequency of cabinet changes contribute significantly to measuring the time-varying risk of conflict recurrence. The growth regression based on this alternative specification of conflict recurrence risk, presented in Table 3, yields similar results coinciding with the benchmark model, except for foreign aid, which appears to reduce post-conflict growth. Furthermore, though slightly weaker because of the loss of valuable information related to post-conflict growth, these results, based on the alternative specification of conflict recurrence risk, have no potential problem of endogeneity or reverse causality, as the level and growth rate of per capita income are excluded in the estimation.

To check the robustness of our empirical findings based on the sample covering the entire period 1960-2010, we estimate the post-conflict growth model for different time regimes (1970-1989 and 1990-2010). The risk of conflict recurrence remains a statistically important factor with a negative coefficient across different time specifications, as shown in Table 4, although data limitations appear to reduce the level of significance. We also test whether the statistical significance of our results is independent of geographical factors by running regressions that exclude one region at a time. For example, the first column of Table 5 shows the results when European countries are excluded from the growth regression: both the benchmark and alternative definitions of conflict recurrence risk have a statistically significant and economically large negative effect on post-conflict real GDP per capita growth. Other geographic specifications yield similar results, supporting the validity of the findings based on our benchmark model.

CONCLUSION

The number of civil conflicts has declined over the past decade, but countries emerging from a conflict still struggle to break the curse of Sisyphus, with 20% relapsing into a new bout of conflict in the first year and 40% within 5 years after the end of conflict. Hence, understanding the causes and consequences of civil conflict remains an important objective in the quest to achieve sustainable sociopolitical reconciliation and economic development. This paper makes a novel contribution to the literature by estimating a forward-looking measure of conflict recurrence risk and including as an explanatory variable in analyzing post-conflict growth dynamics. We adopt a logistic regression approach and model the probability of conflict recurrence with the latest available information in each period, while controlling for unobserved fixed effects in a non-linear probability model for the first time in the conflict literature.

Using a panel dataset covering 146 episodes of civil conflict in 94 countries from 1960 to 2010, we find that the estimated risk of conflict recurrence has a statistically significant and economically large effect on post-conflict growth. A 1 % increase in the probability of conflict recurrence in the subsequent period is found to lower real GDP per capita growth by 10%. The rest of our main empirical results are broadly in line with the existing literature. Human capital and institutional factors have significant positive effects on post-conflict growth, while natural resource dependence appears to cause a 'double curse' of higher risk of conflict recurrence and lower growth, compared with other countries with a more diversified economic structure. Ethnic fractionalization is another significant factor with negative influence on the strength of post-conflict growth. Global growth and capital flows matter for post-conflict recovery, especially for countries that are more open and integrated with the rest of the world. Finally, contrary to popular expectations, we do not find UN peacekeeping operations, foreign aid and IMF-supported economic programs to have significant effect on post-conflict growth.

The empirical results presented in this paper call for a framework centered on three main pillars, implementing growth-enhancing policies, reforming dysfunctional institutions, and addressing urgent needs, to reduce the risk of conflict recurrence and pave the way for broad-based, sustainable economic growth after the end of a conflict. While sound macroeconomic management is important for success, helping to sustain the pace of recovery and avoid a relapse into conflict, strengthening institutions is critical for addressing longstanding grievances and raising the economy's growth potential. Moreover, the challenge facing post-conflict countries is not just to make economic growth stronger, but also more inclusive to ensure that the benefits of increased prosperity are shared more evenly across society, especially considering the possibility that pre-conflict policies and institutional arrangements may have contributed to the outbreak of civil conflict by discriminating against particular ethnic, religious or social segments. Last but not the least, in the particular case of natural resource-dependent countries, the need for economic diversification is not just a long-term objective, but an urgent undertaking in order to break the 'double curse' of higher conflict recurrence risk and lower post-conflict growth.

The principal contribution of this paper is to take a first pass at quantifying the impact of the forward-looking risk of conflict recurrence on post-conflict economic transitions. Accordingly, a valuable extension would be an investigation of the dynamic process of how economic agents assess the risk of conflict recurrence and its evolution over time using a Bayesian methodology. Similarly, the geographic diffusion of conflict risk within and between countries is an interesting empirical question.

Acknowledgements

The authors would like to thank Josef Brada, Carlos Caceres, Paul Cashin, Victor Davies, Selim Elekdag, Samiei Hossein, Padamja Khandelwal, Tidiane Kinda, Zijun Luo, Lawrence McDonough, Kia Penso, Katerina Teksoz, Fatih Yilmaz, and two anonymous referees for their insightful comments and suggestions. The views expressed herein are those of the authors and should not be attributed to the IMF, its executive Board, or its management.

REFERENCES

Acemoglu, D, Johnson, S and Robinson, J. 2001: The colonial origins of comparative development: An empirical investigation. American Economic Review 91(5): 1369-1401.

Acemoglu, D, Johnson, S and Robinson, J. 2004: Institutions as the Fundamental Cause of Long-Run Growth. NBER Working Paper No. 10481, National Bureau for Economic Research: Cambridge, MA.

Acemoglu, D and Johnson, S. 2005: Unbundling institutions. Journal of Political Economy 113(5): 949-995.

Addison, T, Chowdhury, A and Murshed, S. 2002: By How Much Does Conflict Reduce Financial Development? WIDER Discussion Paper No. 2002/48, United Nations University World Institute for Development Economics Research: Helsinki.

Arellano, M and Bond, S. 1991: Some tests of specification for panel data: Monte carlo evidence and an application to employment equations. Review of Economic Studies 58(2): 277-297.

Arellano, M and Bover, O. 1995: Another look at the instrumental variables estimation of error components models. Journal of Econometrics 68(1): 29-51.

Blattman, C and Miguel, E. 2010: Civil war. Journal of Economic Literature 48(1): 3-57.

Blundell, R and Bond, S. 1998: Initial conditions and moment restrictions in dynamic panel data models. Journal of Econometrics 87(1): 115-143.

Bond, S, Hoeffler, A and Temple, J. 2001: GMM Estimation of Empirical Growth Models. CEPR Discussion Paper No. 3048, Center for Economic Policy Research: London.

Cassar, A, Grosjean, P and Whitt, S. 2013: Legacies of violence: Trust and market development. Journal of Economic Growth 18(3): 285-318.

Choi, I. 2001: Unit root tests for panel data. Journal of International Money and Finance 20(2): 249-272.

Collier, P. 1999: On the economic consequences of civil war. Oxford Economic Papers 51(1): 168-183.

Collier, P and Hoeffler, A. 2002: On the incidence of civil war in Africa. Journal of Conflict Resolution 46(1): 13-28.

Collier, P and Hoeffler, A. 2004: Greed and grievance in civil war. Oxford Economic Papers 56(4): 563-595.

Elbadawi, I, Hegre, H and Milante, G. 2008: The aftermath of civil war. Journal of Peace Research 45(4): 451-459.

Eichengreen, B, Rose, A and Wyplosz, C. 1995: Exchange rate Mayhem: The antecedents and aftermath of speculative attacks. Economic Policy 21 (10): 249-312.

Engerman, S and Sokoloff, K. 2000: Institutions, factor endowments, and paths of development in the new world. Journal of Economic Perspectives 14(3): 217-232.

Esteban, J, Mayoral, L and Ray, D. 2012: Ethnicity and conflict: An empirical study. American Economic Review 102(4): 1310-1342.

Estrella, A and Mishkin, F. 1998: Predicting U.S. Recessions: Financial variables as leading indicators. Review of Economics and Statistics 85(3): 629-644.

Fearon, J and Laitin, D. 2003: Ethnicity, insurgency, and civil war. American Political Science Review 97(1): 75-90.

Frankel, J and Rose, A. 1996: Currency crashes in emerging markets: An empirical treatment. Journal of International Economics 41(3): 351-366.

Gallup, J, Sachs, J and Mellinger, A. 1998: Geography and economic development. International Regional Science Review 22(2): 179-232.

Grosjean, P. 2014: Conflict and social and political preferences: Evidence from world war II and civil conflict in 35 European countries. Comparative Economic Studies 56(3): 424-451.

Grossman, H. 1991: A general equilibrium model of insurrections. American Economic Review 81(4): 912-921.

Hall, R and Jones, C. 1999: Why do some countries produce so much more output per worker than others? Quarterly Journal of Economics 114(1): 83-116.

Hegre, H, Ellingsen, T, Gates, S and Gledish, N. 2001: Towards a democratic civil peace? Democracy, political change, and civil war, 1816-1992. American Political Science Review 95(1): 22-45.

Hirshleifer, J. 1995: Theorizing about Conflict. In Handbook of Defense Economics Hartley, K and Sandler, T (eds) Elsevier Science: Amsterdam.

Holz-Eakin, D, Newey, W and Rosen, H. 1988: Estimating vector autoregressions with panel data. Econometrica 56(6): 1371-1395.

Humphreys, M. 2005: Natural resources, conflict, and conflict resolution: Uncovering the mechanisms. Journal of Conflict Resolution 49(4): 508-537.

Im, K, Pesaran, M and Shin, Y. 2003: Testing for unit roots in heterogeneous panels. Journal of Econometrics 115(1): 53-74.

Knack, S and Keefer, P. 1995: Institutions and economic performance: Cross-country tests using alternative institutional measures. Economics and Politics 7(3): 207-227.

Levin, A, Lin, CF and Chu, CS. 2002: Unit root tests in panel data: Asymptotic and finite-sample properties. Journal of Econometrics 108(1): 1-24.

Miguel, E, Satyanath, S and Sergenti, E. 2004: Economic shocks and civil conflict: An instrumental variables approach. Journal of Political Economy 112(4): 725-753.

Nafzinger, E, Stewart, F and Vayrynen, R. 2000: War, Hunger, and Displacement: The Origin of Humanitarian Emergencies. Oxford University Press: New York.

Reynal-Querol, M. 2002: Political systems, stability and civil wars. Defense and Peace Economics 13(6): 465-483.

Reynal-Querol, M. 2005: Does democracy preempt civil wars? European Journal of Political Economy 21(2): 445-465.

Ross, M. 2006: A closer look at oil, diamonds and civil war. Annual Review of Political Science 9(1): 265-300.

Sachs, J and Warner, A. 1997: Sources of slow growth in African economies. Journal of African Economies 6(3): 335-376.

Sambanis, N. 2001: Do ethnic and non-ethnic civil wars have the same causes? A Theoretical and empirical enquiry. Journal of Conflict Resolution 45(3): 259-282.

Sandler, T. 2000: Economic analysis of conflict. Journal of Conflict Resolution 44(6): 723-729.

Skaperdas, S. 1992: Cooperation, conflict, and power in the absence of property rights. American Economic Review 82(4): 720-739.

Suhrke, A, Villanger, E and Woodward, S. 2005: Economic Aid to Post-Conflict Countries: A Methodological Critique of Collier and Hoeffler. CMI Working Papers No. 2005:4, Chr. Michelsen Institute: Bergen, Norway.

Wooldridge, J. 2005: Fixed-effects and related estimators for correlated random-coefficient and treatment-effect. Panel Data Models 87(2): 385-390.

(1) This database is a joint project of the Uppsala Conflict Data Program (UCDP) at the Department of Peace and Conflict Research, Uppsala University and the Center for the Study of Civil War at the International Peace Research Institute in Oslo (PRIO).

(2) Humphreys (2005) and Blattman and Miguel (2010) provide extensive reviews of the literature on the causes and consequences of civil conflicts.

(3) There is a vast and growing literature on early warning indicators and prediction methods, led by, among others, Eichengreen et al. (1995) and Frankel and Rose (1996) on forecasting currency crises and Estrella and Mishkin (1998) on predicting economic recessions.

(4) Without loss of generality, the threshold can be 0.

(5) While the 'had conflict' variable may arguably capture country characteristics, there is no test available to address this question directly. It is also worth noting that country fixed effects are modeled separately in our estimation, based on the logit random-effect framework. Also, including a range of other explanatory variables shows that the marginal impact of having a conflict before never drops below 25 %.

(6) We test other measures of human capital such as primary, tertiary, and total school enrollment rates, which show the same pattern, but are statistically insignificant. In our view, this is because conflict has a greater effect on secondary school enrollment, causing a higher degree of volatility compared with other variables.

(7) Additional regressions show that hydrocarbon production is not significant for the occurrence of civil conflict. We also consider two other time dummies for the discovery of hydrocarbon reserves and the start of production, which are insignificant when the 'had conflict' variable is included in the regression.

(8) These simple averages, not weighted by population or GDP, better highlight the global level of conflict recurrence risk. First, as the GDP and population have a direct effect on conflict risk, averaging according to these weights can lead to biased results. Second, the sample of countries is changing over time in both panels.

(9) An important feature of this result is that we use an abstract measurement of risk in our calculation (the first column of Table 1). If we include other variables, the impact of the risk of conflict recurrence would be larger.

(10) This conclusion should be treated with caution, because the direction of causality is unclear as weak institutions may also contribute to conflict duration and to slower growth afterwards. In this paper, while we do not explicitly deal with this issue, we reason that the impact of institutional factors beyond what is incorporated in our growth model is captured by fixed effects and therefore should not distort the key empirical findings.

(11) An important problem with the analysis of aid effectiveness is identification when all types of aid are bundled together. For example, humanitarian or military aid should not be expected to have the same effect on economic growth as aid adding to a country's productive capacity. Lacking disaggregated data on foreign aid, this study, too, relies on total foreign aid as measured by the OECD Development Assistance Committee.

SERHAN CEVIK [1] & MOHAMMAD RAHMATI [2]

[1] International Monetary Fund, 700 19th Street, NW, Washington DC, 20431, USA. E-mail: scevik@imf.org

[2] Sharif University of Technology, Azadi Street, 1458889694, Tehran, Iran.

Table 1: Determinants of the risk of conflict recurrence:
random-effect logistic panel model

                                            Models

Variables                   1 Baseline         2             3
                                          Alternative

Log [(real GDP per          -0.363 ***                   -0.275 **
capita).sub.t]                (-3.11)                     (-2.25)

Real GDP per capita          -1.639 **                   -1.667 **
[growth.sub.t]                (-2.22)                     (-2.24)

Had [conflict.sub.t]         5.383 ***     2.619 ***     5.238 ***
                              (9.57)        (3.49)        (9.35)

Log [(global                                 1.443
real GDP).sub.t]                            (0.87)

Global real GDP                             -4.703
[growth.sub.t]                              (-0.90)

Duration of                                0.044 ***
[conflict.sub.t]                            (3.33)

Terms of [trade.sub.t]

Natural resource
[rent.sub.t]

Natural [resource
rent.sup.2.sub.t]

Secondary school
enrollment [rate.sub.t]

Secondary school
enrollment [rate.sub.t]

Ethnic fractionalization                     0.675       1.677 **
                                            (0.87)        (2.18)

Youth [bulge.sub.t]                        0.121 ***
                                            (3.93)

Frequency of cabinet                       0.352 ***
[changes.sub.t]                             (3.62)

Type of political
[regime.sub.t]

Type of political
[regime.sup.2.sub.t]

UN peacekeeping
[operations.sub.t]

1960s                       -0.920 ***                  -0.866 ***
                              (-4.62)                     (-4.33)

1970s                         -0.196         0.103        -0.159
                              (-1.21)       (0.16)        (-0.98)

1980s                        0.423 ***       0.551       0.452 ***
                              (2.84)        (1.24)        (3.01)

1990s                        0.531 ***       0.309       0.515 ***
                              (3.83)        (1.13)        (3.68)

Constant                    -4.782 ***      -21.707     -6.038 ***
                              (-4.63)       (-1.53)       (-5.00)

[[sigma].sup.2.sub.v]        1.202 ***     1.086 ***     1.156 ***
                              (6.79)        (5.25)        (6.45)

Log Likelihood                 -1469         -1060         -1449

[chi square]                   184.9         86.17         179.7

Number of observations         7,087         2,848         6,795

Number of countries             195           105           184

                                            Models

Variables                        4             5             6

Log [(real GDP per            -0.087      -0.507 ***    -0.399 ***
capita).sub.t]                (-0.70)       (-3.20)       (-3.08)

Real GDP per capita          -1.834 **     -2.687 **     -1.988 **
[growth.sub.t]                (-2.23)       (-2.57)       (-2.44)

Had [conflict.sub.t]         4.932 ***     5.627 ***     5.407 ***
                              (8.57)        (7.70)        (8.50)

Log [(global                                 1.724
real GDP).sub.t]                            (1.05)

Global real GDP                             -3.813
[growth.sub.t]                              (-0.68)

Duration of
[conflict.sub.t]

Terms of [trade.sub.t]                     -4.630 **
                                            (-2.55)

Natural resource                                         4.294 ***
[rent.sub.t]                                              (2.72)

Natural [resource                                         -3.134
rent.sup.2.sub.t]                                         (-1.49)

Secondary school
enrollment [rate.sub.t]

Secondary school
enrollment [rate.sub.t]

Ethnic fractionalization

Youth [bulge.sub.t]

Frequency of cabinet         0.190 **      0.293 ***
[changes.sub.t]               (2.17)        (2.89)

Type of political              0.013
[regime.sub.t]                (1.05)

Type of political           -0.016 ***
[regime.sup.2.sub.t]          (-6.63)

UN peacekeeping
[operations.sub.t]

1960s                       -0.625 ***
                              (-2.83)

1970s                          0.167         0.708        -0.060
                              (0.87)        (1.03)        (-0.35)

1980s                        0.691 ***     1.215 **      0.525 ***
                              (3.94)        (2.55)        (3.37)

1990s                        0.626 ***     0.866 ***     0.621 ***
                              (4.08)        (3.27)        (4.26)

Constant                    -5.734 ***      -18.662     -4.996 ***
                              (-5.28)       (-1.33)       (-4.39)

[[sigma].sup.2.sub.v]        1.163 ***     1.443 ***     1.292 ***
                              (6.40)        (7.26)        (7.01)

Log Likelihood                 -1323         -1028         -1254

[chi square]                   205.5         133.2         134.4

Number of observations         5,409         4,564         5,340

Number of countries             159           165           168

                                            Models

Variables                        7             8             9

Log [(real GDP per          -0.457 ***     -0.260 **     -0.282 **
capita).sub.t]                (-3.03)       (-2.03)       (-2.39)

Real GDP per capita         -3.397 ***    -2.305 ***     -1.766 **
[growth.sub.t]                (-3.30)       (-2.87)       (-2.38)

Had [conflict.sub.t]         5.606 ***     5.292 ***     5.428 ***
                              (7.54)        (9.30)        (9.56)

Log [(global
real GDP).sub.t]

Global real GDP
[growth.sub.t]

Duration of
[conflict.sub.t]

Terms of [trade.sub.t]      -4.890 ***
                              (-2.92)

Natural resource             6.847 ***
[rent.sub.t]                  (3.69)

Natural [resource            -5.235 **
rent.sup.2.sub.t]             (-1.97)

Secondary school                           -11.822 *
enrollment [rate.sub.t]                     (-1.82)

Secondary school                           76.206 **
enrollment [rate.sub.t]                     (2.26)

Ethnic fractionalization

Youth [bulge.sub.t]

Frequency of cabinet
[changes.sub.t]

Type of political
[regime.sub.t]

Type of political
[regime.sup.2.sub.t]

UN peacekeeping                                          0.931 ***
[operations.sub.t]                                        (3.55)

1960s                                     -0.877 ***    -0.824 ***
                                            (-3.31)       (-4.10)

1970s                          0.066        -0.153        -0.103
                              (0.35)        (-0.75)       (-0.62)

1980s                        0.779 ***     0.521***      0.494 ***
                              (4.52)        (3.15)        (3.26)

1990s                        0.752 ***     0.633***      0.558 ***
                              (4.76)        (4.24)        (3.99)

Constant                    -4.559 ***    -5.190 ***    -5.473 ***
                              (-3.40)       (-4.85)       (-5.11)

[[sigma].sup.2.sub.v]        1.440 ***     1.182 ***     1.210 ***
                              (7.45)        (6.58)        (6.85)

Log Likelihood                 -1098         -1387         -1463

[chi square]                   138.1         177.0         197.5

Number of observations         4,646         6,303         7,087

Number of countries             153           184           195

Note: The dependent variable is a dummy for the recurrence
of a conflict at the next period; the coefficients are
estimated with a random-effect logit panel model for the
sample period 1960-2010. z-statistics are reported in
parentheses, with *, **, and *** indicating statistical
significance at the 1%, 5%, and 10% levels, respectively.

Source: Authors' estimations

Table 2: Determinants of post-conflict real GDP
per capita growth: system-gmm dynamic panel

                                 1             2             3

Real GDP per capita          0.289 ***     0.291 ***     0.241 ***
[growth.sub.t-1]              (14.44)       (14.57)       (11.26)

Log real GDP per            -0.030 ***    -0.033 ***    -0.028 ***
[capita.sub.t]                (-6.82)       (-7.14)       (-5.87)

Risk of conflict             -0.089 **     -0.087 **     -0.093 **
[recurrence.sub.t]            (-2.33)       (-2.27)       (-2.47)

Duration of                   -0.011                     -0.012 **
[conflict.sub.t]              (-1.42)                     (-2.00)

Time since the end of                      0.004 ***
[conflict.sub.t]                            (3.02)

Depth of contraction                                      -0.056
during [conflict.sub.t]                                   (-1.03)

Terms of [trade.sub.t]

Temperature
[variability.sub.t]

[FDI.sub.t]

Landlocked

Global real GDP
[growth.sub.t]

UN peacekeeping
[operations.sub.t]

Trade [openness.sub.t]

Trade
Openness*[GDP.sub.t]

Constant                     0.253 ***     0.253 ***     0.239 ***
                              (6.91)        (7.30)        (6.11)

Number of observations         1,916         1,916         1,563

Number of countries             89            89            83

Sargan Test                    1943          1943          1605

[chi square]                   333.0         343.3         206.5

                                 4             5             6

Real GDP per capita          0.168 ***     0.286 ***     0.150 ***
[growth.sub.t-1]              (7.82)        (13.51)       (6.27)

Log real GDP per            -0.021 ***    -0.022 ***    -0.025 ***
[capita.sub.t]                (-5.38)       (-5.69)       (-5.89)

Risk of conflict            -0.121 ***    -0.103 ***     -0.079 **
[recurrence.sub.t]            (-3.19)       (-2.59)       (-1.97)

Duration of                   -0.002         0.012         0.002
[conflict.sub.t]              (-0.11)       (-1.25)       (-0.02)

Time since the end of
[conflict.sub.t]

Depth of contraction
during [conflict.sub.t]

Terms of [trade.sub.t]         0.043
                              (0.92)

Temperature                                  0.001
[variability.sub.t]                         (0.85)

[FDI.sub.t]                                              0.007 ***
                                                          (5.31)

Landlocked

Global real GDP
[growth.sub.t]

UN peacekeeping
[operations.sub.t]

Trade [openness.sub.t]

Trade
Openness*[GDP.sub.t]

Constant                     0.182 ***     0.178 ***     0.216 ***
                              (5.40)        (5.34)        (6.25)

Number of observations         1,618         1,617         1,166

Number of countries             83            78            81

Sargan Test                    1696          1667          1247

[chi square]                   119.9         272.7         108.3

                                 7             8

Real GDP per capita          0.296 ***     0.256 ***
[growth.sub.t-1]              (14.65)       (12.31)

Log real GDP per            -0.037 ***    -0.035 ***
[capita.sub.t]                (-7.67)       (-7.12)

Risk of conflict             -0.083 **    -0.112 ***
[recurrence.sub.t]            (-2.14)       (-2.73)

Duration of                   -0.003        -0.002
[conflict.sub.t]              (-0.91)       (-0.71)

Time since the end of
[conflict.sub.t]

Depth of contraction
during [conflict.sub.t]

Terms of [trade.sub.t]

Temperature
[variability.sub.t]

[FDI.sub.t]

Landlocked                  -0.093 ***
                              (-4.17)

Global real GDP                            0.481 ***
[growth.sub.t]                              (5.83)

UN peacekeeping
[operations.sub.t]

Trade [openness.sub.t]

Trade
Openness*[GDP.sub.t]

Constant                     0.311 ***     0.276 ***
                              (7.79)        (6.88)

Number of observations         1,856         1,748

Number of countries             84            89

Sargan Test                    1865          1744

[chi square]                   370.4         331.6

                                 9            10

Real GDP per capita          0.289 ***     0.206 ***
[growth.sub.t-1]              (14.44)       (8.47)

Log real GDP per            -0.030 ***    -0.058 ***
[capita.sub.t]                (-6.74)       (-8.31)

Risk of conflict             -0.096 **     -0.115 **
[recurrence.sub.t]            (-2.48)       (-2.29)

Duration of                   -0.011      -0.020 ***
[conflict.sub.t]              (-1.44)       (-3.25)

Time since the end of
[conflict.sub.t]

Depth of contraction
during [conflict.sub.t]

Terms of [trade.sub.t]

Temperature
[variability.sub.t]

[FDI.sub.t]

Landlocked

Global real GDP
[growth.sub.t]

UN peacekeeping                0.014
[operations.sub.t]            (1.04)

Trade [openness.sub.t]                     -0.120 *
                                            (-1.74)

Trade                                      0.020 **
Openness*[GDP.sub.t]                        (2.10)

Constant                     0.252 ***     0.470 ***
                              (6.87)        (8.27)

Number of observations         1,916         1,261

Number of countries             89            73

Sargan Test                    1942          1279

[chi square]                   334.2         205.1

                                11            12            13

Real GDP per capita          0.226 ***     0.277 ***     0.225 ***
[growth.sub.t-1]              (8.51)        (14.02)       (11.16)

Log real GDP per            -0.071 ***    -0.031 ***    -0.036 ***
[capita.sub.t]                (-8.64)       (-7.15)       (-7.88)

Risk of conflict             -0.103 *     -0.098 ***    -0.146 ***
[recurrence.sub.t]            (-1.70)       (-2.59)       (-3.89)

Duration of                 -0.020 ***      -0.010       -0.012 **
[conflict.sub.t]              (-2.99)       (-1.44)       (-1.97)

Time since the end of
[conflict.sub.t]

Depth of contraction
during [conflict.sub.t]

Temperature
[variability.sub.t]

Landlocked

Global real GDP
[growth.sub.t]

UN peacekeeping
[operations.sub.t]

Trade [openness.sub.t]       -0.119 *
                              (-1.66)

Trade                         0.018 *
0penness*[GDP.sub.t]          (1.89)

Trade Openness*Global        0.622 **
[growth.sub.t]                (2.27)

Foreign aid per                             -0.001
[capita.sub.t]                              (-0.39)

Secondary school                                         0.239 ***
enrollment [rate.sub.t]                                   (2.90)

Ethnic
[fractionalization.sub.t]

Cabinet [Changes.sub.t]

Natural resource
[rent.sub.t]

IMF [Program.sub.t]

Constant                     0.562 ***     0.262 ***     0.293 ***
                              (8.32)        (7.28)        (7.91)

Number of observations         1,103         1,908         1,761

Number of countries             73            89            89

Sargan Test                    1125          1930          1738

[chi square]                   234.6         328.3         265.9

                                14            15            16

Real GDP per capita          0.289 ***     0.240 ***     0.230 ***
[growth.sub.t-1]              (14.40)       (11.47)       (10.94)

Log real GDP per            -0.032 ***    -0.033 ***    -0.038 ***
[capita.sub.t]                (-6.87)       (-7.14)       (-8.05)

Risk of conflict             -0.091 **    -0.134 ***    -0.168 ***
[recurrence.sub.t]            (-2.38)       (-3.43)       (-4.24)

Duration of                   -0.011       -0.012 **       0.002
[conflict.sub.t]              (-1.07)       (-2.11)       (-0.62)

Time since the end of
[conflict.sub.t]

Depth of contraction
during [conflict.sub.t]

Temperature
[variability.sub.t]

Landlocked

Global real GDP
[growth.sub.t]

UN peacekeeping
[operations.sub.t]

Trade [openness.sub.t]

Trade
0penness*[GDP.sub.t]

Trade Openness*Global
[growth.sub.t]

Foreign aid per
[capita.sub.t]

Secondary school
enrollment [rate.sub.t]

Ethnic                        -0.036
[fractionalization.sub.t]
                              (-1.33)

Cabinet [Changes.sub.t]                   -0.009 ***
                                            (-3.99)

Natural resource                                         -0.048 **
[rent.sub.t]                                              (-2.40)

IMF [Program.sub.t]

Constant                     0.278 ***     0.290 ***     0.317 ***
                              (6.67)        (7.56)        (8.23)

Number of observations         1,908         1,691         1,767

Number of countries             88            89            89

Sargan Test                    1940          1669          1757

[chi square]                   334.2         274.2         300.0

                                17            18

Real GDP per capita          0.289 ***     0.208 ***
[growth.sub.t-1]              (14.43)       (6.47)

Log real GDP per            -0.029 ***    -0.088 ***
[capita.sub.t]                (-6.61)       (-8.20)

Risk of conflict             -0.087 **    -0.241 ***
[recurrence.sub.t]            (-2.26)       (-3.24)

Duration of                                  0.010
[conflict.sub.t]                            (0.68)

Time since the end of                      0.013 ***
[conflict.sub.t]                            (2.62)

Depth of contraction                      -0.495 ***
during [conflict.sub.t]                     (-3.64)

Temperature                                 -0.002
[variability.sub.t]                         (-0.64)

Landlocked                                  -0.059
                                            (-1.60)

Global real GDP                              0.014
[growth.sub.t]                              (0.04)

UN peacekeeping                             0.063 *
[operations.sub.t]                          (1.76)

Trade [openness.sub.t]                      -0.028
                                            (-0.37)

Trade                                        0.008
0penness*[GDP.sub.t]                        (0.76)

Trade Openness*Global                        0.011
[growth.sub.t]                              (0.02)

Foreign aid per                           -0.013 ***
[capita.sub.t]                              (-3.28)

Secondary school
enrollment [rate.sub.t]

Ethnic                                     -0.098 **
[fractionalization.sub.t]
                                            (-2.29)

Cabinet [Changes.sub.t]                   -0.012 ***
                                            (-3.71)

Natural resource                             0.010
[rent.sub.t]                                (0.34)

IMF [Program.sub.t]           -0.005        -0.006
                              (-1.27)       (-1.21)

Constant                     0.231 ***     0.701 ***
                              (6.85)        (7.69)

Number of observations         1,916          773

Number of countries             89            57

Sargan Test                    1942          797.8

[chi square]                   332.4         215.0

Notes: The dependent variable is the growth rate
of real GDP per capita at time t, starting with the end of conflict.
The model is estimated with the system-GMM dynamic panel approach
for the sample period 1960-2010.

The sample includes only post-conflict observations.

z-statistics are reported in parentheses, with *, **,
and *** indicating statistical significance at the 1%, 5%,
and 10% levels, respectively.

Source: Authors' estimations

Table 3: Determinants of post-conflict real GDP per capita growth
(with the alternative measure of conflict risk):
System-GMM dynamic panel

                                            Models

Variables                        1             2             3

Real GDP per capita          0.272 ***     0.285 ***     0.151 ***
[growths.sub.t-1]             (12.13)       (11.51)       (5.93)

Log (real GDP per           -0.059 ***    -0.052 ***    -0.036 ***
[capita.sub.t])               (-9.16)       (-7.35)       (-6.43)

Risk of conflict            -0.092 ***     -0.074 **    -0.132 ***
[recurrence.sub.t]            (-2.67)       (-2.13)       (-4.03)

Duration of                 -0.022 ***    -0.030 ***      -0.000
[conflict.sub.t]              (-2.93)       (-2.99)       (-0.53)

Depth of contraction                       -0.167 **
during [conflict.sub.t]                     (-2.55)

Terms of [trade.sub.t]                                     0.015
                                                          (0.25)

Temperature
[variability.sub.t]

[FDI.sub.t]

Landlocked

Global real GDP
[growth.sub.t]

UN peacekeeping
[operations.sub.t]

Trade [openness.sub.t]

Trade openness*
[GDP.sub.t]

Trade openness*global
[growth.sub.t]

Constant                     0.486 ***     0.432 ***     0.303 ***
                              (9.22)        (7.39)        (6.27)

Number of observations         1,176          976          1,004

Number of countries             79            72            72

Sargan Test                    1119          963.0         1006

[chi square]                   242.0         182.1         81.15

                                            Models

Variables                        4             5             6

Real GDP per capita          0.257 ***     0.138 ***     0.278 ***
[growths.sub.t-1]             (10.72)       (4.83)        (12.31)

Log (real GDP per           -0.043 ***    -0.026 ***    -0.060 ***
[capita.sub.t])               (-7.94)       (-4.88)       (-9.07)

Risk of conflict            -0.123 ***    -0.106 ***     -0.084 **
[recurrence.sub.t]            (-3.53)       (-2.79)       (-2.41)

Duration of                    0.010         0.010      -0.023 ***
[conflict.sub.t]              (1.45)        (1.43)        (-2.74)

Depth of contraction
during [conflict.sub.t]

Terms of [trade.sub.t]

Temperature                    0.000
[variability.sub.t]           (0.08)

[FDI.sub.t]                                0.006 ***
                                            (2.94)

Landlocked                                               -0.069 **
                                                          (-2.00)

Global real GDP
[growth.sub.t]

UN peacekeeping
[operations.sub.t]

Trade [openness.sub.t]

Trade openness*
[GDP.sub.t]

Trade openness*global
[growth.sub.t]

Constant                     0.329 ***     0.217 ***     0.503 ***
                              (7.18)        (4.94)        (9.22)

Number of observations         1,056          730          1,150

Number of countries             70            68            76

Sargan Test                    1009          766.5         1085

[chi square]                   205.9         55.74         264.0

                              Models

Variables                        7             8

Real GDP per capita          0.273 ***     0.270 ***
[growths.sub.t-1]             (12.20)       (11.83)

Log (real GDP per           -0.058 ***    -0.059 ***
[capita.sub.t])               (-9.01)       (-8.94)

Risk of conflict             -0.078 **    -0.092 ***
[recurrence.sub.t]            (-2.26)       (-2.69)

Duration of                 -0.023 ***    -0.022 ***
[conflict.sub.t]              (-2.95)       (-2.94)

Depth of contraction
during [conflict.sub.t]

Terms of [trade.sub.t]

Temperature
[variability.sub.t]

[FDI.sub.t]

Landlocked

Global real GDP              0.400 ***
[growth.sub.t]                (3.02)

UN peacekeeping                             -0.007
[operations.sub.t]                          (-0.35)

Trade [openness.sub.t]

Trade openness*
[GDP.sub.t]

Trade openness*global
[growth.sub.t]

Constant                     0.470 ***     0.491 ***
                              (8.90)        (9.00)

Number of observations         1,176         1,176

Number of countries             79            79

Sargan Test                    1118          1119

[chi square]                   251.0         242.2

                              Models

Variables                        9            10

Real GDP per capita          0.258 ***     0.258 ***
[growths.sub.t-1]             (11.30)       (11.38)

Log (real GDP per           -0.072 ***    -0.071 ***
[capita.sub.t])               (-9.44)       (-9.37)

Risk of conflict             -0.090 **     -0.077 **
[recurrence.sub.t]            (-2.53)       (-2.14)

Duration of                 -0.020 ***    -0.020 ***
[conflict.sub.t]              (-2.68)       (-2.75)

Depth of contraction
during [conflict.sub.t]

Terms of [trade.sub.t]

Temperature
[variability.sub.t]

[FDI.sub.t]

Landlocked

Global real GDP
[growth.sub.t]

UN peacekeeping
[operations.sub.t]

Trade [openness.sub.t]       -0.164 **     -0.167 **
                              (-2.30)       (-2.36)

Trade openness*              0.026 ***     0.025 ***
[GDP.sub.t]                   (2.69)        (2.58)

Trade openness*global                      0.763 ***
[growth.sub.t]                              (2.87)

Constant                     0.570 ***     0.562 ***
                              (9.35)        (9.25)

Number of observations         1,096         1,096

Number of countries             73            73

Sargan Test                    1078          1080

[chi square]                   233.6         242.1

                                            Models

Variables                       11            12            13

Real GDP per capita          0.266 ***     0.271 ***     0.272 ***
[growth.sub.t-1]              (11.81)       (12.10)       (12.13)

Log (real GDP per           -0.060 ***    -0.059 ***    -0.060 ***
[capita.sub.t])               (-9.34)       (-9.17)       (-8.84)

Risk of conflict            -0.093 ***    -0.096 ***    -0.092 ***
[recurrence.sub.t]            (-2.71)       (-2.78)       (-2.69)

Duration of                 -0.022 ***    -0.020 ***    -0.020 ***
[conflict.sub.t]              (-3.12)       (-2.85)       (-2.87)

Depth of contraction
during [conflict.sub.t]

Temperature
[variability.sub.t]

[FDI.sub.t]

Landlocked

Global real GDP
[growth.sub.t]

UN peacekeeping
[operations.sub.t]

Trade [openness.sub.t]

Trade openness*
[GDP.sub.t]

Trade openness*global
[growth.sub.t]

Foreign aid per              -0.005 **
[capita.sub.t]                (-2.12)

Secondary school                             0.101
enrollment [rate.sub.t]                     (0.77)

Ethnic fractionalization                                  -0.015
                                                          (-0.41)

Number of cabinet
[changes.sub.t]

Natural resource
[rent.sub.t]

Constant                     0.500 ***     0.485 ***     0.498 ***
                              (9.42)        (9.19)        (8.27)

Number of observations         1,176         1,175         1,176

Number of countries             79            79            79

Sargan Test                    1117          1118          1117

[chi square]                   246.1         241.1         241.8

                                            Models

Variables                       14            15            16

Real GDP per capita          0.269 ***     0.270 ***     0.259 ***
[growth.sub.t-1]              (12.08)       (11.61)       (8.94)

Log (real GDP per           -0.060 ***    -0.053 ***    -0.078 ***
[capita.sub.t])               (-9.40)       (-8.32)       (-7.91)

Risk of conflict            -0.100 ***    -0.088 ***     -0.100 **
[recurrence.sub.t]            (-2.93)       (-2.59)       (-2.54)

Duration of                 -0.022 ***    -0.022 ***       0.000
[conflict.sub.t]              (-3.02)       (-2.35)        -0.37

Depth of contraction                                    -0.586 ***
during [conflict.sub.t]                                   (-4.24)

Temperature                                               -0.002
[variability.sub.t]                                       (-0.91)

[FDI.sub.t]

Landlocked                                                -0.029
                                                          (-0.73)

Global real GDP                                            0.147
[growth.sub.t]                                            (0.45)

UN peacekeeping                                            0.058
[operations.sub.t]                                        (1.60)

Trade [openness.sub.t]                                    -0.017
                                                          (-0.22)

Trade openness*                                            0.006
[GDP.sub.t]                                               (0.58)

Trade openness*global                                      0.096
[growth.sub.t]                                            (0.15)

Foreign aid per                                         -0.012 ***
[capita.sub.t]                                            (-3.02)

Secondary school
enrollment [rate.sub.t]

Ethnic fractionalization                                  -0.058
                                                          (-1.40)

Number of cabinet           -0.013 ***                  -0.012 ***
[changes.sub.t]               (-4.51)                     (-3.71)

Natural resource                           -0.062 **       0.036
[rent.sub.t]                                (-2.32)       (1.22)

Constant                     0.504 ***     0.442 ***     0.593 ***
                              (9.59)        (8.48)        (7.41)

Number of observations         1,176         1,172          757

Number of countries             79            79            57

Sargan Test                    1112          1098          785.0

[chi square]                   262.7         254.5         209.5

Notes: The dependent variable is the growth rate of real GDP
per capita at time t, starting with the end of conflict.

The model is estimated with the system-GMM dynamic panel
approach for the sample period 1960-2010.

The sample includes only post-conflict observations.

z-statistics are reported in parentheses, with *, **, and
*** indicating statistical significance at the 1%, 5%,
and 10% levels, respectively.

Source: Authors' estimations

Table 4: Estimation of different time regimes: System-GMM dynamic
panel

                                              1960-1989

Variables                                Baseline    Alternative

Real GDP per capita [growth.sub.t-1]      0.021         0.041
                                          (0.58)       (1.16)
Log real GDP per [capita.sub.t]         -0.065 ***   -0.085 ***
                                         (-8.45)      (-10.20)
Risk of conflict [recurrence.sub.t]     -0.247 **    -0.315 ***
                                         (-2.32)       (-6.50)
Constant                                0.503 ***     0.679 ***
                                          (9.13)       (10.72)
Number of observations                     570           544
Number of countries                         49           49

                                               1990-2010

Variables                                Baseline    Alternative

Real GDP per capita [growth.sub.t-1]    0.303 ***     0.337 ***
                                          (6.76)       (11.00)
Log real GDP per [capita.sub.t]         -0.028 ***   -0.053 ***
                                         (-3.05)       (-5.15)
Risk of conflict [recurrence.sub.t]       -0.003      -0.197 **
                                         (-0.02)       (-2.16)
Constant                                0.210 ***     0.409 ***
                                          (2.71)       (5.15)
Number of observations                    1,172          597
Number of countries                         86           75

Notes: The dependent variable is the growth rate of real GDP
per capita at time t, starting with the end of conflict.

The model is estimated with the system-GMM dynamic
panel approach for the sample period 1960-2010.

z-statistics are reported in parenthesis, with *, **,
and *** indicating statistical significance at the 1%,
5%, and 10% levels, respectively.

The sample includes only post-conflict data.

Source: Authors' estimations

Table 5: Estimation of different geographic specifications:
System-GMM dynamic panel

Variables                          Benchmark Model

                                      excluding

                         Europe      Middle East      Asia

Real GDP per capita     0.197 ***     0.235 ***     0.238 ***
[growth.sub.t-1]         (8.21)        (10.17)       (10.27)

Log real GDP per       -0.039 ***    -0.037 ***    -0.041 ***
[capita.sub.t]           (-7.39)       (-7.41)       (-7.75)

Risk of conflict       -0.126 ***    -0.178 ***    -0.200 ***
[recurrence.sub.t]       (-2.72)       (-3.79)       (-3.85)

Constant                0.306 ***     0.298 ***     0.335 ***
                         (7.35)        (7.55)        (7.80)

Number of                 1,582         1,620         1,539
observations

Number of                  78            82            76
countries

Variables                   Benchmark Model

                               excluding

                         Africa       Americas

Real GDP per capita     0.258 ***     0.224 ***
[growth.sub.t-1]         (9.41)        (8.89)

Log real GDP per       -0.025 ***    -0.045 ***
[capita.sub.t]           (-5.79)       (-7.73)

Risk of conflict       -0.144 ***     -0.119 **
[recurrence.sub.t]       (-4.27)       (-2.28)

Constant                0.222 ***     0.341 ***
                         (6.46)        (7.47)

Number of                 1,141         1,291
observations

Number of                  54            72
countries

Variables                         Alternative Model

                                      excluding

                         Europe      Middle East      Asia

Real GDP per capita     0.224 ***     0.291 ***     0.270 ***
[growth.sub.t-1]         (9.14)        (12.39)       (11.23)

Log real GDP per       -0.056 ***    -0.043 ***    -0.054 ***
[capita.sub.t]           (-8.43)       (-7.02)       (-8.49)

Risk of conflict       -0.157 ***     -0.064 *     -0.205 ***
[recurrence.sub.t]       (-4.48)       (-1.94)       (-4.75)

Constant                0.433 ***     0.324 ***     0.437 ***
                         (8.63)        (7.07)        (8.76)

Number of                 1,065         1,071         1,013
observations

Number of                  69            71            66
countries

Variables                  Alternative Model

                               excluding

                         Africa       Americas

Real GDP per capita     0.269 ***     0.252 ***
[growth.sub.t-1]         (10.29)       (9.57)

Log real GDP per       -0.028 ***    -0.068 ***
[capita.sub.t]           (-5.00)       (-9.04)

Risk of conflict       -0.090 ***    -0.140 ***
[recurrence.sub.t]       (-3.07)       (-3.83)

Constant                0.246 ***     0.513 ***
                         (5.46)        (9.08)

Number of                  768           834
observations

Number of                  51            61
countries

Notes: The dependent variable is the growth rate of real GDP
per capita at time t, starting with the end of conflict.

The model is estimated with the system-GMM dynamic panel
approach for the sample period 1960-2010.

z-statistics are reported in parentheses, with *, **,
and *** indicating statistical significance at the 1%,
5%, and 10% levels, respectively.

The sample includes only post-conflict data.

Source: Authors' estimations