Economic Costs of Conflict DEVELOPMENT ECONOMICS II, HECER March, 2016
Outline Introduction Macroeconomic costs - Basque County Microeconomic costs - education/health Microeconomic costs- social capital Conflict March, 2016 2 / 42
Introduction Research Questions What post-conflict programs work best? What determines recovery? What can prevent recurrence of conflict? Conflict March, 2016 3 / 42
Introduction Research Questions What post-conflict programs work best? What determines recovery? What can prevent recurrence of conflict? What are the economic costs of conflict? economic growth, physical capital, human capital, social capital Who bears the costs of conflict? Conflict March, 2016 3 / 42
Introduction Economic Costs of Conflict Conflict March, 2016 4 / 42
Introduction Economic Costs of Conflict Surveys: Blattman & Miguel (2010) Macroeconomic Costs reduce per capita GDP growth reduce FDI inflows,exports, trade flows reduce domestic investment, savings redirect public expenditure to defence reduce tourist inflows, tourism receipts reduce demand for transportation Conflict March, 2016 4 / 42
Introduction Economic Costs of Conflict Surveys: Blattman & Miguel (2010) Macroeconomic Costs reduce per capita GDP growth reduce FDI inflows,exports, trade flows reduce domestic investment, savings redirect public expenditure to defence reduce tourist inflows, tourism receipts reduce demand for transportation Micro/Household-level Costs reduce household investments (agriculture) reduce school enrollment, grade completion reduce health (height for age z-scores) reduce reported life satisfaction reduce social capital Conflict March, 2016 4 / 42
Introduction Post-conflict scenarios Recovery High rates of return to capital & labor Financial aid - foreign & domestic rapid recovery to steady state possible Poverty traps Depleted capital stock may not recover post-war uncertainties increase risks decrease investments Conflict March, 2016 5 / 42
Macroeconomic effects of conflict Conflict March, 2016 6 / 42
The Economic Costs of Conflict Abadie, A., & Gardeazabal, J. (2003). The economic costs of conflict: A case study of the Basque Country. American Economic Review, 113-132. ETA established in 1959, demands independence for Basque country from 1975 ETA becomes violent Business in Basque affects: deaths, kidnappings-for-ransom, extortion, robberies. cease-fire agreement in 1998 Conflict March, 2016 7 / 42
The Economic Costs of Conflict Abadie, A., & Gardeazabal, J. (2003). The economic costs of conflict: A case study of the Basque Country. American Economic Review, 113-132. ETA established in 1959, demands independence for Basque country from 1975 ETA becomes violent Business in Basque affects: deaths, kidnappings-for-ransom, extortion, robberies. cease-fire agreement in 1998 What is the economic cost of extremist violence? comparing Basque region to average Spain not correct- Basque performing better pre-conflict Conflict March, 2016 7 / 42
Synthetic Control Method (SCM) J + 1 regions 1 treated, J controls 1,...T 0,...T time periods. Region 1 treated from T O onwards Y it = Yit N + α it D it is the observed outcome for region i at time t α it = Yit I Y it N is the effect of the intervention on region i at time t To estimate (α 1,T0+1,..., α 1,T ) need to estimate Y1t N, for t {T 0+1,...T } Conflict March, 2016 8 / 42
Synthetic Control Method (Contd) use pre-treatment information on outcomes and predictors to weight the controls in such a way so that they resemble the treated unit creating a synthetic control counterfactual outcome for the treated unit in post-treatment periods estimated from the outcome for the synthetic control Then ˆα it = Y 1t J+1 treatment effect j=2 ω j Y jt N for t {T 0+1,...T } is the estimated Conflict March, 2016 9 / 42
Synthetic Control Method (Contd) Let W = {w j } J+1 j=2 be a set of weights s.t. w j 0 & j=2 J+1 w j = 1 The vector of weights W is chosen to min (X 1 X 0 W ) V (X 1 X 0 W ) X is matrix of pre-treatment Z and Y W is the weight for controls Note it is an extension of the standard difference-in-difference methodology relaxes the assumption of a common linear trend between the treated and the controls in the absence of treatment Conflict March, 2016 10 / 42
Economic Costs of violence in Basque Country Treatment: Violence by ETA from 1975 onwards Treated Unit: Basque country Control Units: all other 16 regions Controls are all regions not affected by ETA related violence Outcome Variable: real GDP per capita (1986 prices) Pre-treatment period: 1955-1974 Treatment Period: 1975-1997 Conflict March, 2016 11 / 42
Trends in GDP: Basque vs. Synthetic Control 10% loss in pcgdp over the period 1975-1997. Conflict March, 2016 12 / 42
GDP Predictor Means Only 2 control regions get positive weights: Catalonia and Madrid. Conflict March, 2016 13 / 42
Placebo Tests Placebo test: perform analysis for control as if the control was treated Since control was not treated we should not find any treatment effect Conflict March, 2016 14 / 42
Some studies show no long-run effects Miguel, E., Roland, G. (2011). The long-run impact of bombing Vietnam. Journal of Development Economics, 96(1), 1-15. Conflict March, 2016 15 / 42
No effect of bombings on long run poverty in Vietnam Non-bombed areas vs. bombed areas: Bombed areas did NOT have lower... Local poverty rates Consumption levels Infrastructure Literacy Population density Conflict March, 2016 16 / 42
No effects of war? Compares districts within Vietnam National growth rate may have been faster in the absence of war Government investment/foreign aid could have gone to other, non-bombed regions Private foreign investment may have been greater if it were not a post-conflict country Selection bias: countries that recover (e.g. Vietnam) more likely to collect good data than countries that collapse (e.g. Somalia) Conflict March, 2016 17 / 42
Effects on Human Capital Conflict March, 2016 18 / 42
Motivation Conflict March, 2016 19 / 42
Case Study: Conflict & Schooling in Tajikistan Shemyakina, O. (2011). The effect of armed conflict on accumulation of schooling: Results from Tajikistan. Journal of Development Economics, 95(2), 186-200. Tajikistan: Background Independence in 1991 after collapse of Soviet Union Civil war 1992-1998 At least 50,000 men killed; 55,000 children orphaned Over 600,000 internally displaced Conflict March, 2016 20 / 42
Tajikistan Map Conflict March, 2016 21 / 42
Enrollment Rates Conflict March, 2016 22 / 42
Estimation Strategy Difference-in-Difference: Exploit variation in geographical extent of violence & exposure of birth cohorts to violence Mandatory schooling between ages 7-15. Pre-1992 1992-98 Not Affected Region 16+ by 1992 school age but not exposed Affected Region 16+ by 1992 war cohort Conflict March, 2016 23 / 42
Estimation Strategy Regression Equation: Assumption: S ijk = α 1j + β 1k + γ 1 (P j K i ) + ɛ ijk In absence of war, same schooling trend in war-affected and peaceful regions No unobserved time-varying region FE correlated with conflict Conflict March, 2016 24 / 42
Mandatory Schooling Result: Girls affected Conflict March, 2016 25 / 42
Taking Stock: Gender Differences Hypothesis: Equal effects school infrastructure destruction, household displacement, unavailability of teachers should have equal effects Conflict March, 2016 26 / 42
Taking Stock: Gender Differences Hypothesis: Equal effects But... school infrastructure destruction, household displacement, unavailability of teachers should have equal effects Boys may be more affected more likely to enlist (or forced conscription) more likely to be sent to work to help HH cope with shock Girls may be more affected Unsafe, protection from rape reduced budget - substitution towards boys Conflict March, 2016 26 / 42
Similar effects on health childhood nutrition status long-run health conflict (shock) early life nutrition reduced long-run productivity poverty trap childhood health status has multiple dimensions height-for-age Conflict March, 2016 27 / 42
Conflict & behavior in Burundi Burundi: Background 2 ethnic groups: Hutu (85%) & Tutsi (14%) Army - Tutsi dominated; 1993 Government - Hutu dominated 1993: President Melchior Ndadaye (Hutu) assassinated by the Army 1993-2003: Civil war in Burundi over 300,000 killed; 1.2 million displaced Peace process ongoing since 2005. Conflict March, 2016 28 / 42
Burundi Bundervoet, T., Verwimp, P., & Akresh, R. (2009). Health and civil war in rural Burundi. Journal of Human Resources, 44(2), 536-563. Compared to non-affected areas, children (under 5) in affected areas had lower height-for-age score more severely affected by greater exposure This short-run health impact of the Burundi civil war could have a long-run welfare impact through an adverse effect on future schooling, adult health, and income levels. Conflict March, 2016 29 / 42
Effects on Social Capital Conflict March, 2016 30 / 42
Motivation Concerns about the linkages between post-conflict recovery, local institutions and economic development. Conflict March, 2016 31 / 42
Motivation Concerns about the linkages between post-conflict recovery, local institutions and economic development. Why do some regions make faster post-conflict recovery than others? Conflict March, 2016 31 / 42
Motivation Concerns about the linkages between post-conflict recovery, local institutions and economic development. Why do some regions make faster post-conflict recovery than others? Local cooperative norms growth & development. Preferences (consumption, saving, investment behavior) growth & development. Conflict March, 2016 31 / 42
Motivation Concerns about the linkages between post-conflict recovery, local institutions and economic development. Why do some regions make faster post-conflict recovery than others? Local cooperative norms growth & development. Preferences (consumption, saving, investment behavior) growth & development. Conflict can shift social norms affect expectations, preferences, choices Conflict March, 2016 31 / 42
Motivation Concerns about the linkages between post-conflict recovery, local institutions and economic development. Why do some regions make faster post-conflict recovery than others? Local cooperative norms growth & development. Preferences (consumption, saving, investment behavior) growth & development. Conflict can shift social norms affect expectations, preferences, choices violence = preferences = post conflict recovery Conflict March, 2016 31 / 42
Conflict & behavior in Burundi Voors, M. J., Nillesen, E. E., Verwimp, P., Bulte, E. H., Lensink, R., & Van Soest, D. P. (2012). Violent conflict and behavior: a field experiment in Burundi. American Economic Review, 102(2), 941-964. 1993-2003: Civil war in Burundi Survey 300 household heads from 35 communities in 2009 24 communities experienced violence during 1993-2003 11 did not explanatory variables collected in a previous 2007 survey main measure of violence is % of community killed during 1993-2003 attacks Conflict March, 2016 32 / 42
Altruism: Experiment Index: 0 (purely selfish) - 100 (totally altruistic) Conflict March, 2016 33 / 42
Altruism: Results Conflict March, 2016 34 / 42
Risk Aversion: Experiment choose between certain gain/loss and a lottery Conflict March, 2016 35 / 42
Risk Aversion: Results Conflict March, 2016 36 / 42
Time Preference: Experiment earlier you switch from A to B, more patient Conflict March, 2016 37 / 42
Time Preference: Results Conflict March, 2016 38 / 42
Results Summary Greater exposure to violence is associated with more altruistic behavior more risk-seeking behavior more impatience Conflict March, 2016 39 / 42
Results Summary Greater exposure to violence is associated with more altruistic behavior more risk-seeking behavior more impatience Note: channel not clear in-group cooperation & out-group antagonism are the shifts permanent? effect on future investments ambiguous Conflict March, 2016 39 / 42
Can post-conflict aid build social capital? Fearon et al. (2009). Can development aid contribute to social cohesion after civil war? Evidence from a field experiment in post-conflict Liberia. American Economic Review, 99(2), 287-291. Can community-driven reconstruction (CDR) programs promote social reconciliation & increase social cooperation? Or are social norms and institutions too slow moving? RCT in post-conflict Liberia Treatment villages (42): community development committee (CDC) setup via elections. CDCs then manage & implement dev projects. Control villages (41) - no CDCs. Local cooperation measured via public goods game. Exposure to CDR program increases public goods contribution by 8%. Conflict March, 2016 40 / 42
Limitations nonrandom exposure to violence: who are victimized? rich vs. poor gender ethnicity, political participation Conflict March, 2016 41 / 42
Limitations nonrandom exposure to violence: who are victimized? rich vs. poor gender ethnicity, political participation nonrandom attrition: who survive? who migrates? Conflict March, 2016 41 / 42
Taking stock Long-run: some evidence that capital stocks recover but recovery contingent on political stability & institutions (e.g. Germany, Japan, Vietnam) Conflict March, 2016 42 / 42
Taking stock Long-run: some evidence that capital stocks recover but recovery contingent on political stability & institutions (e.g. Germany, Japan, Vietnam) Substantial negative effects on human capital & labor disproportionate burden on the poor, elderly, children & women Conflict March, 2016 42 / 42
Taking stock Long-run: some evidence that capital stocks recover but recovery contingent on political stability & institutions (e.g. Germany, Japan, Vietnam) Substantial negative effects on human capital & labor disproportionate burden on the poor, elderly, children & women Effects on social capital little understood. Conflict March, 2016 42 / 42