Economic shocks and civil conflict: Evidence from financial crises

Economic shocks and civil conflict: Evidence from financial crises Eric Lenz January 20, 2017

Abstract Recent development research relates civil conflict to externally-driven income shocks, but severe declines in income also relate to financial crisis. Can financial crisis and its external determinants influence civil conflict risk? Using direct measures of financial crisis, and controlling for countryspecific effects on the onset and incidence of conflict, I find evidence that exchange rate fluctuations drive income decline and financial crisis to increase conflict risk. This research contributes to existing development research that has shown significant conflict risk from two determinants of developing country financial crisis, but not of exchange rates and financial crisis itself.

I.

Introduction

The current role of a financial crisis in causing falling incomes is hard to ignore. Since 1986 financial crisis has been the main driver of worldwide economic recessions (Kannan and Terrones, 2013). The economic recessions, defined by falling incomes, also positively relate to civil conflict risk (Blomberg and Hess, 2002). This relationship is more recently confirmed by Hull and Imai (2013) who suggest that civil conflict risk can be determined by declining income that is driven by foreign interest rates. Another foreign component, commodity terms of trade, further predicts the onset of civil conflict (Janus and Riera-Crichton, 2015) and supports the hypothesis that external economic fluctuations have important domestic effects. I suggest that another external factor, exchange rates, also influence conflict risk through domestic income channels and financial crisis. The important causal relationship between civil conflict and economic activity is not necessarily one-way1 . Poor economic conditions bring about conflict (Grossman, 1991; Hirshleifer, 1995; Collier and Hoeffler, 2004), but a civil conflict also clearly results in declining economic activity (Polachek and Sevastianova, 2010; Hull and Imai, 2013). The relationship between financial crisis and the economy is similar - financial crisis results in poor economic and financial conditions (Valencia and Laeven, 2008; Kannan and Terrones, 2013; Reinhart and Rogoff, 2009), but a decline in economic activity may precede financial crisis in developing countries (Goldstein and Turner, 1996). I address the possible endogenous 1 For instance, development literature defines the contentious relationship between income and conflict as the povertyconflict nexus. Alexander (2005) and Djankov and Reynal-Querol (2010) discover a spurious relationship between poverty and conflict via fixed-effects estimation. Janus and Riera-Crichton (2015) and Hull and Imai (2013) find a significant, negative relationship. This paper contributes to this research and finds statistical evidence for economic and financial motivations for conflict’s onset.

1

nature of financial crisis and conflict through independence tests and instrumentation via two-stage least squares. Another shared aspect of civil conflict and financial crises is that they are difficult to predict. However, if financial crises relate to economic recession, then research into economic recession and civil conflict can provide a starting point for analysis. Blomberg and Hess (2002) measure the effects of economic recession and expansion on the probability of foreign and civil conflict. However, civil conflict is more probable in developing countries than in industrial countries if recent development research confirms the poverty-conflict nexus (Cerra and Saxena, 2008; Hull and Imai, 2013; Janus and Riera-Crichton, 2015). Past research into developing country financial crises also shows that developing countries face more pronounced economic downturns (Goldstein and Turner, 1996). Therefore, for predictive purposes and to inform policy, the analysis of factors that contribute to developing country financial crisis is worthwhile. The three external predictors of financial crises in developing countries are foreign interest rates, commodity terms of trade, and exchange rates (Goldstein and Turner, 1996). The first predictor, foreign interest rates, experiences volatility and decline prior to crisis (Sachs and Velasco, 1996; Goldstein and Turner, 1996). In a developing country prior to crisis, foreign capital flows into a country due to appreciating real exchange rates and rising domestic interest rates (Sachs and Velasco, 1996). The rising domestic interest rates provoke more foreign currency denominated commercial bank lending (Goldstein and Turner, 1996) and this less costly commercial bank lending may also occur under a fixed-exchange rate system. The fixed-exchange rate system allows a developing country to peg it’s domestic currency to stable, foreign-denominated currency (Mishkin, 1999). Therefore, external factors that affect a developing country’s borrowing and lending, such as volatility in foreign currency lending and foreign interest charged on that lending, influence the risk of financial crisis. Mishkin (1999) suggests a strong role from exchange rate fluctuations and general interest rates for the onset of financial crisis, but does not emphasize terms of trade. Sachs and Velasco (1996) echoes the importance of real exchange rates, but does not emphasize international interest rates or terms of trade. Goldstein and Turner (1996) concludes all three are important external factors for financial crisis. The recent development research into conflict’s onset and incidence suggest two external predictors: foreign interest rates and commodity terms of trade. The measure of trade competitiveness, commodity terms of trade, increases the probability of conflict by 0.5% with every 1% decline at time (t-1) (Janus and Riera-Crichton, 2015). Another predictor of civil conflict, foreign interest rates, increase the probability of conflict by 30% with every 4% decline in gross domestic product (Hull and Imai, 2013). The instrumentation of gross domestic product follows research The common external components of financial crisis and conflict in developing countries may be linked; however, a clear relationship has not yet been formally established. This paper seeks to establish such a relationship between external crisis determinants, civil conflict, and financial crisis. I investigate the relationship between financial crisis, economic shocks, and civil conflict in an international panel from 1970 to 2011. I control for country-specific effects on the onset of conflict using fixed-effects to account for unobserved, time-invariant heterogeneity (Djankov and Reynal-Querol, 2010). I find

2

a strong relationship between financial crisis and civil conflict which is not dependent on a country’s ethnic composition. The relationship is also unrelated to the national decline in gross domestic product suggested by the research.

II.

Data

The international dataset from multiple sources spans the years 1970 to 2011 in over 100 countries. The main indicators of interest are measures of domestic conflict, systemic banking crises, and exchange rates. I also control for countries with intermediate ethnic diversity (Fearon and Laitin, 2003) following methods of Janus and Riera-Crichton (2015) with interaction of external economic factors. The primary independent variables are the external determinants of systemic banking crises and the incidence of systemic banking crises (referred to from now on as financial crises). I construct the incidence of financial crisis from Valencia and Laeven (2012) as a dummy variable taking the value of “1” for each country, i, and year, t, during the crisis and “0” otherwise2 The simple, straightforward represetation of crisis incidence is identical to civil conflict’s incidence and represents a great deal of information related to the timing of crises. Several factors determine when a financial crisis begins and ends such as non-performing loans as a percentage of total loans, gross fiscal costs and output loss as a percentage of GDP, and minimum real GDP growth (Valencia and Laeven, 2012). Valencia and Laeven (2008) also cross-check the dates of crises with the timing of deposit runs, deposit freezes, liquidity support, and bank interventions. The accuracy and precision of this dataset is also superior to the data set from Reinhart and Rogoff (2009) according to work by Chaudron and de Haan (2014). This collection with precise timing of financial crises allows for a rigorous analysis of recessions that bring about intrastate conflict3 . I focus on two external macroeconomic factors that change before and during a financial crisis: commodity terms of trade and international interest rates (Goldstein and Turner, 1996). The commodity terms of trade data, from Janus and Riera-Crichton (2015), is compiled from the shares of imports and exports for 6 commodity categories and commodity prices. The specification of annual money market interest rates follow the methodology of Hull and Imai (2013) and the data comes from the IMF’s International Financial Statistics database. The main dependent variable is the onset of civil conflict and is constructed from the Uppsala Conflict Database Project and the International Peace Research Institute of Oslo, Norway (UCDP/PRIO). The UCDP/PRIO specifications do not differ greatly from year to year by version (Hull and Imai, 2013). This paper specifies civil conflict’s onset following Bazzi and Blattman (2014); Janus and Riera-Crichton (2015) as the onset of civil conflict involving 25 deaths). This conflict variable identifies the first year in which at least 25 civilians or combatants4 . The construction of conflict follows Bazzi and Blattman 2 Reinhart and Rogoff (2009) constructs financial crisis incidence similarly to represent the incidence of banking, currency, and sovereign debt crises. 3 Valencia and Laeven (2008, 2012) also consider a larger sample of countries over 42 years in comparison to Reinhart and Rogoff (2009) who examine fewer countries over many more years. Despite this smaller sample of countries, Reinhart and Rogoff (2009) identify multiple crisis types in a centuries-long time span. 4 An alternative conflict specification is 1,000 deaths of armed combatants or armed combatants and civilians. This

3

(2014); Janus and Riera-Crichton (2015) a datum of “1” to indicate the onset of civil conflict, “0” to indicate periods without civil conflict, and excluded observations during the civil war to account for endogenous effects from war on the independent variables. The summary statistics in Table 1 show the number of observations, mean, standard deviation, and minimum and maximum values for all variables. The conflict and financial crisis variables are dummy variables representing the incidence of civil conflict and financial crisis. The variable list in Table 1 include the two dependent conflict variables, time-varying direct and indirect shocks, and ethnic fractionalization: Table 1: Summary Statistics

Civil conflict onset Civil conflict incidence Financial crisis incidence 4 Exchange rate 4 Commodity terms of trade FIMM PA kaopen peg Recession EF (Fearon) Real GDP growth Population (000)

Obs

Mean

Std. Dev.

Min

Max

4739 5927 5931 5260 4087 2146 5288 5481 5513 5931 5513 5624

0.050 0.182 0.068 0.120 0.000 5193.457 -0.026 0.459 0.165 0.480 3.748 37413.689

0.217 0.386 0.252 0.472 0.012 211408.780 1.527 0.498 0.372 0.265 6.756 128128.739

0.000 0.000 0.000 -0.223 -0.084 0.001 -1.895 0.000 0.000 0.002 -66.120 229.588

1.000 1.000 1.000 21.481 0.103 9695422.000 2.389 1.000 1.000 1.000 106.280 1348174.500

Note: The onset and incidence of civil conflict is current to 2011 with data from the Uppsala Conflict Data Program in 2016. Real GDP, population, and exchange rates are from the Penn World Tables 9.0 Feenstra et al. (2015). Recession incidence is a dummy variable for GDP growth < 0 defined by Blomberg and Hess (2002). Commodity terms of trade is the 3-year moving average from Janus and Riera-Crichton (2015) available to year 2009. Exchange rates are 3-year moving averages following methods identical to Janus and Riera-Crichton (2015). Financial crisis incidence is a dummy variable constructed from Valencia and Laeven (2012). Independent variables are limited to non-missing observations of ethnic fractionalization from Fearon and Laitin (2003).

The economic control variables from the Penn World Tables follow economic research into conflict from Janus and Riera-Crichton (2015) and Hull and Imai (2013) and are real gross domestic product (GDP), GDP growth, and population (in thousands). Janus and Riera-Crichton (2015) include GDP per capita; however, GDP and population typically have opposing relationships to conflict. Hull and Imai (2013) does not include GDP, but rather GDP growth. Conflict and development research suggests that low gross domestic product and its loss per annum increase the probability of conflict and large populations are more likely to experience civil war than smaller populations (Collier and Hoeffler, 2004; Fearon and Laitin, 2003). specification defines civil war in Janus and Riera-Crichton (2015)

4

III.

Independence of Civil Conflict and Financial Crisis

The two-way tabulations show the dependent relationship between civil conflict and financial crisis. The incidence of financial crisis on the left-hand side is compared to the simultaneous occurence of conflict’s onset and incidence. There are 21 observations of financial crisis occuring simultaneously with conflict’s onset, and 88 occurences of crisis during conflict. This translates to approximately 7% of financial crises occuring at the onset of conflict and 22% occuring of crises occuring simultaneous during civil conflicts5 . Table 2: Independence Tests for Financial Crisis Incidence and Civil Conflict, 1970-2011

Banking Crisis Incidence 0

1

Total

Observations Pearson sig LR sig Fisher’s exact

0 4221 95 94 283 93 6 4504 95 100

Conflict Onset (t) 1 Total 214 5 91 21 7 9 235 5 100

4435 100 94 304 100 6 4739 100 100

4739 0.106 0.124 0.131

0 4532 82 93 316 78 7 4848 82 100

Conflict Incidence (t) 1 Total 991 18 92 88 22 8 1079 18 100

5523 100 93 404 100 7 5927 100 100

5927 0.054 0.059 0.061

Note: Each cell contains three rows which are frequencies, row percentages, and column percentages.

The independence tests reject the null hypothesis of independence at the 1% significance level for conflict and crisis. The p-values for the Pearson, likelihood ratio, and Fisher’s test all suggest a dependent relationship between conflict and crises at time (t). Therefore, the simultaneous occurence of conflict and financial crisis may suggest an endogenous relationship. Hull and Imai (2013) address the endogenous relationship between gross domestic product and conflict via two-stage least squares regression and the instrumentation of GDP growth with international interest rates. I modify the linear regression model from Hull and Imai (2013) to instrument GDP growth and financial crisis with an external determinant of financial crisis: exchange rates. 5

I also estimate the onset of civil conflict with the onset of financial crisis; however, the independence tests cannot reject the null hypothesis of independence at the 5% significance level.

5

Table 3: Independence Tests for Financial Crisis Incidence and Civil Conflict, 1970-2011

Banking Crisis Incidence 0

0

Conflict Onset (t) 1 Total

3551 97 94 245 94 6 3796 96 100

1

Total

Observations Pearson sig LR sig Fisher’s exact

122 3 88 16 6 12 138 4 100

3673 100 93 261 100 7 3934 100 100

3934 0.017 0.029 0.023

0 3782 90 93 268 86 7 4050 89 100

Conflict Incidence (t) 1 Total 433 10 91 43 14 9 476 11 100

4215 100 93 311 100 7 4526 100 100

4526 0.049 0.058 0.055

Note: Each cell contains three rows which are count frequencies, row percentages, and column percentages.

IV. A.

Empirical Strategy Linear Two-Stage Least Squares

The modified model from Hull and Imai (2013) to estimate conflict with indirect economic shocks in a panel time-series framework with fixed-effects:

Crisisit = θi + θitrend t + β1 Si(t−1) + β2 Xi(t−1) + vit

(1)

ˆ it + δ2 Xi(t−1) + it Conf lictit = γi + γitrend t + δ1 Crisis

(2)

Conf lictit represents the dummy variable for the incidence of conflict in country i at time t. The main results from recent models in Janus and Riera-Crichton (2015); Hull and Imai (2013) exclude effects from traditional time-invariant control variables in favor of fixed-effects by country, θi and γi , and country-specific time trends θitrend t and γitrend t. The control variables, Xi,(t−1) , include popular economic indicators of conflict. The Valencia and Laeven (2012) financial crisis variable, Crisisit , varies by country and year, and is instrumented with indirect economic shocks, Si,(t−1) . This method to account for endogeneity of income growth and conflict in Hull and Imai (2013) is adapted for financial crisis and conflict. The methods of linear regression are currently popular in macroeconomic development literature (Bazzi and Blattman, 2014; Janus and Riera-Crichton, 2015). However, the resulting errors from linear probability regression are heteroscedastic unless corrected as error variance is conditional on the independent variables. Therefore, I control for heteroscedasticity by estimating robust standard errors

6

clustered by country (Bazzi and Blattman, 2014; Janus and Riera-Crichton, 2015). There is a strong case for including country-specific factors in conflict research as time-variant economic factors, like GDP per capita and population, differ greatly from country to country (Alexander, 2005; Djankov and Reynal-Querol, 2010; Sambanis, 2004; Elbadawi and Sambanis, 2002; Bates, 2005; Ciccone, 2011; Miguel and Satyanath, 2011). When the country-specific heterogeneity is accounted for using fixed effects, GDP per capita and population coefficients become insignificant. Djankov and Reynal-Querol (2010) and Alexander (2005) find a statistically insignificant relationship between poverty and conflict-risk which generally weakens the economic opportunity motivations for conflict in Collier and Hoeffler (2004). However, this paper and recent economic development research supports such motivations (Hull and Imai, 2013; Janus and Riera-Crichton, 2015).

V.

Results

A.

Fixed-effects estimation with commodity terms of trade and financial crisis

I confirm the relationship between commodity terms of trade and conflict in Table 4 and follow the methods in Janus and Riera-Crichton (2015) to similarly test the effects from financial crisis incidence and the 3-year moving average of exchange rates. The results in Table 4 Column 2 show the effect of the 3-year moving average of commodity terms of trade on civil conflict in intermediate ethnically diverse countries. The Wald test for the joint hypothesis that variable coefficients are equal to zero is included for all models to determine if economic shocks in intermediate ethnically diverse countries contribute to the onset of conflict. The estimation results show that the lag of 3-year moving average of commodity terms of trade in intermediate ethnically diverse countries in Table 4 Column 2 determine the onset of civil conflict6 However, the same interactions with exchange rates and financial crisis show no effect in Table 4 Column 3-6. The ethnic composition of a country seems to play no role in the relationship between crisis and conflict. However, there is evidence that financial crises do positively relate7 to civil conflict’s onset in Table 4 Column 5.

6 7

This result is comparable to the main results in Janus and Riera-Crichton (2015). I test the robustness of this relationship in the Appendix.

7

Table 4: Conflict’s onset and direct economic shocks, 1970-2009 (1) -0.0515 [0.196]

CTOT movavg lag

(2) 0.631 [0.413]

(3)

(4)

0.00333 [0.00857]

-0.00531 [0.0158]

(5)

(6)

0.0260∗ [0.0147]

0.0228 [0.0217]

-1.137∗∗ [0.568]

CTOT EF

XR movavg lag

XR EF

0.0143 [0.0181]

FinDummy4

FinCrisis EF t

0.00372 [0.0309] 0.0276∗∗∗ [0.00874] 3917 151 138 0.666 xtreg

Constant Observations Countries Conflicts WaldTestPval Regression

0.0329∗∗∗ [0.00974] 3487 139 133 0.243 xtreg

0.0273∗∗∗ [0.00922] 3691 143 124 0.765 xtreg

Standard errors in brackets ∗

p < 0.10,

∗∗

p < 0.05,

∗∗∗

p < 0.01

8

0.0292∗∗∗ [0.0104] 3328 132 122 0.721 xtreg

0.0263∗∗∗ [0.00874] 3917 151 138 0.143 xtreg

0.0308∗∗∗ [0.00988] 3487 139 133 0.409 xtreg

The results in Table 8 Column (3) show insignificant relationship between the 3-year average change in exchange rates and civil conflict for an identical sample from Janus and Riera-Crichton (2015). This result is robust to a specification for intermediate ethnically diverse countries in Column (4). The results show little evidence of a relationship between finnacial crisis and conflict in ethnically diverse countries (6) and in the entire sample (5). However, this relationship does improve for the entire sample from 1970-2011 for all countries.

B.

Indirect economic shocks and civil conflict

I address my main hypothesis that financial crisis strongly relates to civil conflict through exchange rate shocks in the following section. The instrumentation of gross domestic product growth from Hull and Imai (2013) provides the starting point in Table 8. I instrument GDP growth with international interest rates via methods from Hull and Imai (2013) in Table 8 Columns (1-2) and instrument GDP growth with the 3-year moving average of exchange rates for identical years and countries in Columns (3-4). The second-stage results are shown for efficiency in Columns (2) and (4) and the same regressions without instrumentation in Columns (1) and (3). The main hypothesis is that an endogenous relationship between crisis and conflict can be addressed through exchange rate movements as exchange rate movements cause systemic banking crises in developing countries and crises relate to civil conflict. Therefore, I first instrument financial crisis with international interest rates via Hull and Imai (2013) in Table 8 and then with a lagged 3-year moving average specification from Janus and Riera-Crichton (2015) in Table 10 for exchange rates. The results in Table 8 show that financial crisis The movements of international interest rates influence GDP growth and decline such that the risk of civil conflict falls or rises (Hull and Imai, 2013). I test the hypothesis that international interest rates influence the relationship between financial crisis and conflict in Table 10.

VI.

Conclusion

Economic shocks from financial crises lower the opportunity cost of engaging in civil war through declines in exchange rates and financial instability. The estimated odds ratios from the pooled and conditional fixed-effects logit models show the effect of economic recession, financial crisis, and crisis determinants on the risk of civil conflict. The relationship between poverty and conflict has been described as spurious when accounting for country-specific effects (Djankov and Reynal-Querol, 2010; Alexander, 2005), but the present research shows an increase in conflict risk during and at the onset of financial crisis. The odds of a civil conflict erupting with a financial crisis in the previous year are approximately the same as when a financial crisis is absent. The combination of financial crisis and a pegged exchange rate approximately doubles the odds of civil conflict. Future research into the financial aspects of conflict can relate to the size and severity of systemic banking crises as suggested by methods in Chaudron and de Haan (2014). Caballero (2016) addresses

9

Table 5: Conflict’s incidence and average exchange rate growth instrumentation on GDP, 1971-2004

t

PennGDPgrowth

(1) -0.000452 [0.000876]

(2) -0.0000701 [0.000941]

(3) 0.000134 [0.00102]

(4) 0.00128 [0.00124]

(5) 0.00127 [0.00120]

(6) 0.000288 [0.000505]

-0.00283∗∗∗ [0.000734]

-0.0257∗∗ [0.0116]

-0.0168∗∗ [0.00849]

-0.0298 [0.0222]

-0.0287∗ [0.0149]

-0.00838 [0.00780]

-0.0251 [0.0211]

-0.0301 [0.0239]

-0.0299 [0.0236]

-0.00384 [0.0120]

-0.0192∗ [0.0104]

-0.0194∗∗ [0.00967]

-0.00845∗∗ [0.00391]

L.peg

L.kaopen

0.594∗∗∗ [0.0397]

L.conflict

Constant Observations Countries Conflicts Log lik. WaldTestPval Regression

0.173∗∗∗ [0.0215] 6365 182 963 19.58 0.000753 xtreg

0.254∗∗∗ [0.0460] 6365 182 963

0.237∗∗∗ [0.0339] 5662 162 928

0.263∗∗∗ [0.0825] 5225 162 865

0.259∗∗∗ [0.0592] 5225 162 865

0.0955∗∗∗ [0.0321] 5223 162 865

1.50e-27 xtivreg

3.55e-17 xtivreg

1.16e-41 xtivreg

1.30e-54 xtivreg

0 xtivreg

Standard errors in brackets ∗

p < 0.10,

∗∗

p < 0.05,

∗∗∗

p < 0.01

10

Table 6: Conflict’s incidence and average exchange rate growth instrumentation on financial crisis, 1971-2004

t

FinDummy4

(1) -0.000555 [0.000877]

(2) -0.000881 [0.000896]

(3) -0.000695 [0.000982]

(4) 0.000554 [0.00108]

(5) 0.000521 [0.00108]

(6) 0.0000972 [0.000458]

0.0598∗∗ [0.0274]

0.409∗∗ [0.170]

0.437∗∗ [0.199]

0.275 [0.208]

0.302 [0.194]

0.0623 [0.102]

-0.0108 [0.0222]

-0.0161 [0.0230]

-0.0153 [0.0231]

-0.000204 [0.0112]

-0.0198∗∗ [0.00999]

-0.0193∗∗ [0.00984]

-0.00875∗∗ [0.00403]

L.peg

L.kaopen

0.598∗∗∗ [0.0389]

L.Conflict

Constant Observations Countries Conflicts Log lik. WaldTestPval Regression

0.161∗∗∗ [0.0213] 6365 182 963 12.52 0.0792 xtreg

0.147∗∗∗ [0.0229] 6365 182 963

0.156∗∗∗ [0.0280] 5662 162 928

0.140∗∗∗ [0.0313] 5225 162 865

0.139∗∗∗ [0.0309] 5225 162 865

0.0613∗∗∗ [0.0138] 5223 162 865

3.25e-39 xtivreg

9.11e-18 xtivreg

8.38e-51 xtivreg

6.33e-54 xtivreg

0 xtivreg

Standard errors in brackets ∗

p < 0.10,

∗∗

p < 0.05,

∗∗∗

p < 0.01

11

financial crisis onset via net capital bonanzas and lending booms; these factors may also determine the onset of civil conflict.

12

References Alexander, M. (2005). Is poverty to blame for civil war? new evidence from nonlinear fixed effects estimation. Annual Meeting of the American Political Science Association, Washington, DC. Bates, R. H. (2005). Political Insecurity and State Failure in Contemporary Africa. (115). Bazzi, S. and Blattman, C. (2014). Economic Shocks and Conflict: Evidence from Commodity Prices. American Economic Journal: Macroeconomics, 6(4):1–38. Blomberg, S. B. and Hess, G. D. (2002). The temporal links between conflict and economic activity. Journal of Conflict Resolution, 46(1):74–90. Caballero, J. A. (2016). Do surges in international capital inflows influence the likelihood of banking crises? The Economic Journal, 126(591):281–316. Cerra, V. and Saxena, S. C. (2008). Growth Dynamics: The Myth of Economic Recovery. American Economic Review, 98(1):439–57. Chaudron, R. and de Haan, J. (2014). Identifying and dating systemic banking crises using incidence and size of bank failures. Dnb working papers, Netherlands Central Bank, Research Department. Ciccone, A. (2011). Economic Shocks and Civil Conflict: A Comment. American Economic Journal: Applied Economics, 3(4):215–27. Collier, P. and Hoeffler, A. (2004). Greed and Grievance in Civil War. Development and Comp Systems 0409007, EconWPA. Djankov, S. and Reynal-Querol, M. (2010). Poverty and Civil War: Revisiting the Evidence. The Review of Economics and Statistics, 92(4):1035–1041. Elbadawi, I. and Sambanis, N. (2002). How much war will we see?: Explaining the prevalence of civil war. Journal of Conflict Resolution, 46(3):307–334. Fearon, J. D. and Laitin, D. D. (2003). Ethnicity, insurgency, and civil war. American Political Science Review, null:75–90. Feenstra, R. C., Inklaar, R., and Timmer, M. P. (2015). The next generation of the penn world table. American Economic Review, 105(10):3150–82. Goldstein, M. and Turner, P. (1996). Banking crises in emerging economies: origins and policy options. Number no. 46 in BIS economic papers. Bank for International Settlements, Monetary and Economic Department. Grossman, H. (1991). A general equilibrium model of insurrections. American Economic Review, 81(4):912–21.

13

Hirshleifer, J. (1995). Anarchy and its breakdown. Journal of Political Economy, 103(1):26–52. Hull, P. and Imai, M. (2013). Economic shocks and civil conflict: Evidence from foreign interest rate movements. Journal of Development Economics, 103:77 – 89. Janus, T. and Riera-Crichton, D. (2015). Economic shocks, civil war and ethnicity. Journal of Development Economics, 115(C):32–44. Kannan, P., S. A. and Terrones, M. (2013). From recession to recovery: How soon and how strong. financial crises, consequences, and policy responses. Miguel, E. and Satyanath, S. (2011). Re-examining Economic Shocks and Civil Conflict. American Economic Journal: Applied Economics, 3(4):228–32. Mishkin, F. S. (1999). Lessons from the Tequila Crisis. Journal of Banking & Finance, 23(10):1521– 1533. Polachek, S. and Sevastianova, D. (2010). Does Conflict Disrupt Growth? Evidence of the Relationship between Political Instability and National Economic Performance. IZA Discussion Papers 4762, Institute for the Study of Labor (IZA). Reinhart, C. M. and Rogoff, K. S. (2009). This Time Is Different: Eight Centuries of Financial Folly. Princeton University Press, Princeton, New Jersey. Sachs, Jeffrey D., T. A. and Velasco, A. (1996). Financial crises in emerging markets: The lessons from 1995. Brookings Papers on Economic Activity, 27(1):147–216. Sambanis, N. (2004). What is civil war? conceptual and empirical complexities of an operational definition. The Journal of Conflict Resolution, 48(6):pp. 814–858. Valencia, F. and Laeven, L. (2008). Systemic Banking Crises; A New Database. IMF Working Papers 08/224, International Monetary Fund. Valencia, F. and Laeven, L. (2012). Systemic Banking Crises Database; An Update. Technical report.

14

The estimation results in Table 7 show a significant relationship between financial crisis and civil conflict given terms of trade movements in an intermediate ethnically-composed country. The effect of a recession; however, has no significant effect on civil conflict. One reason for this may be that recession’s are more strongly associated with the onset of a larger civil war (1,000 deaths) rather than a smaller civil conflict (25 deaths)8 . The decline in GDP in the previous period has no significant effect on the onset of civil conflict. Also, the effect of commodity terms of trade on conflict’s onset in Table 8 column (4) is still highly significant in the presence of financial crisis. Therefore, what can explain the strong relationship between civil conflict’s onset and financial crisis? Table 7: Robustness to control variables (1) 0.631 [0.413]

(2) 0.897 [0.582]

(3) 0.750 [0.458]

(4) 1.194∗ [0.624]

(5) 0.619 [0.410]

(6) 1.051∗ [0.609]

-1.137∗∗ [0.568]

-1.411∗∗ [0.704]

-1.176∗ [0.597]

-1.722∗∗ [0.738]

-1.131∗∗ [0.567]

-1.505∗∗ [0.717]

0.000208 [0.000382]

0.0000292 [0.000393]

0.000366 [0.000507]

0.000318 [0.000398]

0.000199 [0.000381]

-0.0000262 [0.000528]

CTOT movavg lag

CTOT EF

t

0.000000472∗∗ [0.000000220]

L5.POP Penn

L5.kaopen

0.00000120∗∗ [0.000000544] -0.00389 [0.00352]

L5.Recession

-0.00338 [0.00372] 0.00690 [0.00968]

FinDummy4 0.0329∗∗∗ [0.00974] 3487 139 133 1293.2 0.243 xtreg

Constant Observations Countries Conflicts Log lik. WaldTestPval Regression

0.0220∗ [0.0111] 3327 132 122 1295.6 0.0842 xtreg

0.0299∗∗ [0.0133] 3099 136 122 1104.1 0.229 xtreg

0.0271∗∗ [0.0104] 3315 132 120 1315.1 0.213 xtreg

0.00563 [0.0112] 0.0247 [0.0155]

0.0312∗ [0.0171]

0.0313∗∗∗ [0.00979] 3487 139 133 1295.5 0.168 xtreg

0.00206 [0.0175] 2994 130 115 1118.0 0.0533 xtreg

Standard errors in brackets ∗

8

p < 0.10,

∗∗

p < 0.05,

∗∗∗

p < 0.01

Further analysis seems to bear this out with two-way tabulations and estimations in the Appendix.

15

Table 8: Conflict’s incidence and interest rate instrumentation, 1971-2004

t

PennGDPgrowth

(1) -0.00453∗ [0.00243]

(2) -0.00453∗ [0.00243]

-0.00115 [0.00128]

-0.000969∗ [0.000524]

FinDummy4 0.225∗∗∗ [0.0578] 1724 104 200 363.1 0.175 xtreg

Constant Observations Countries Conflicts Log lik. WaldTestPval Regression

0.224∗∗∗ [0.0579] 1724 104 200 0.00000114 xtivreg

(3) -0.00455∗ [0.00242]

(4) -0.00450∗ [0.00242]

0.0469 [0.0425]

-0.0108∗ [0.00587]

0.218∗∗∗ [0.0553] 1724 104 200 365.2 0.132 xtreg

0.220∗∗∗ [0.0561] 1724 104 200 0.00000184 xtivreg

Standard errors in brackets ∗

p < 0.10,

∗∗

p < 0.05,

∗∗∗

p < 0.01

Table 9: Conflict’s incidence and exchange rate instrumentation, 1971-2004

t

PennGDPgrowth

(1) -0.00453∗ [0.00243]

(2) -0.00498∗∗ [0.00244]

-0.00115 [0.00128]

-0.00703 [0.0228]

FinDummy4 0.225∗∗∗ [0.0578] 1724 104 200 363.1 0.175 xtreg

Constant Observations Countries Conflicts Log lik. WaldTestPval Regression

0.259∗∗∗ [0.0909] 1664 103 191 3.35e-12 xtivreg

Standard errors in brackets ∗

p < 0.10,

∗∗

p < 0.05,

∗∗∗

p < 0.01

16

(3) -0.00455∗ [0.00242]

(4) -0.00497∗∗ [0.00243]

0.0469 [0.0425]

0.0268 [0.0898]

0.218∗∗∗ [0.0553] 1724 104 200 365.2 0.132 xtreg

0.231∗∗∗ [0.0592] 1664 103 191 2.86e-12 xtivreg

Table 10: Conflict’s incidence and commodity terms of trade instrumentation, 1971-2004

t

PennGDPgrowth

(1) -0.00453∗ [0.00243]

(2) -0.00433 [0.00269]

-0.00115 [0.00128]

0.0486 [0.0849]

FinDummy4 0.225∗∗∗ [0.0578] 1724 104 200 363.1 0.175 xtreg

Constant Observations Countries Conflicts Log lik. WaldTestPval Regression

0.00438 [0.307] 1262 87 93 0.0103 xtivreg

Standard errors in brackets ∗

p < 0.10,

∗∗

p < 0.05,

∗∗∗

p < 0.01

17

(3) -0.00455∗ [0.00242]

(4) -0.00470∗∗ [0.00231]

0.0469 [0.0425]

0.476 [0.524]

0.218∗∗∗ [0.0553] 1724 104 200 365.2 0.132 xtreg

0.152∗∗ [0.0657] 1262 87 93 3.64e-10 xtivreg