Armed Conflict, Household Victimization and Child Health in Côte d Ivoire

Transcription

1 Armed Conflict, Household Victimization and Child Health in Côte d Ivoire Camelia Minoiu International Monetary Fund* The World Bank October 16, 2012 Olga Shemyakina School of Economics Georgia Institute of Technology *The views expressed herein are those of the author and should not be attributed to the IMF, its Executive Board, or its management.

2 Motivation Large scale physical destruction arising from armed conflict and natural disasters Macro level studies suggest rapid catch-up growth in physical capital and macroeconomic outcomes (Miguel & Roland, 2011; Davis & Weinstein, 2002) The leading question is not whether wars harm human capital stocks, but rather in what way, for whom and how persistently (Blattman and Miguel, 2010: p. 64)

3 Effects of Armed Conflict on HHDs and Individuals Education Ichino & Winter-Ebmer (2004), Akbulut-Yuksel (2009), Shemyakina (2011), Valente (2011), Akresh & de Walque (2011) Political Participation: Bellows & Miguel (2009), Blattman (2009) Migration and Labor Market Participation Kondylis (2010), Menon and Rodgers (2011), Fernandez, Ibanez and Pena (2011)

4 Civil War & Child Health Civil war and crop failure in Burundi (Bundervoet, Verwimp & Akresh 2009) An extra month of war exposure decreases children s height-forage z-score by st.d. compared to non-exposed children Effects of war in Rwanda (Akresh, Verwimp & Bundervoet 2011) Decrease in the stature of affected children Eritrean-Ethiopian conflict (Akresh, Lucchetti & Thirumurthy 2012) Declines in child health

5 Nutrition and Health in Early Childhood There is no reversal of poor nutrition early in life and the damage to health is permanent (Barker, 1999). Programming process: a fetus adjusts to short-term changes in his or her environment such adaptation is beneficial in the short run, but is detrimental to long-term health (Godfrey and Barker, 2000) Shocks in early childhood have been linked: to lower education and labor market outcomes, lower stature as adults, poorer learning outcomes, poor health (Almond and Currie 2011)

6 Snapshot We use the conflict in Côte d Ivoire as a quasi experiment to analyze the impact of the conflict on children s health Measured by height-for-age z-scores (HAZ) We explore conflict impact mechanisms Common shocks: province-level Idiosyncratic: household-level victimization Migration Model Specification: OLS with fixed effects Province of residence, year and month of birth Province-specific time-trends

7 Our Contribution Establish a baseline result regarding the effect of armed conflict in CIV on child health using: Data collected before, during, and after the conflict Alternative sub-samples Explore the channels through which conflict affects child health: Rich data on households experiences during the conflict from the 2008 survey Compare the effect of the idiosyncratic shock when the common shock is also present Literature on gender bias

8 The Ivorian Conflict

9 Côte d Ivoire: Former Economic Powerhouse of West Africa

10 Children in the Northern Region of Khorogo

11 The Ivorian Conflict Prelude: Prosperity for 3 decades ( ) followed by political instability throughout the 1990s December 1999 Military coup, resulting in socio-political crisis Causes and controversies: Start: Nationality laws Voting rights Land ownership September 19, 2002 with multiple attacks in three cities (North, Center, South)

12 The Ivorian Conflict Rebel forces withdraw to northern areas and establish parallel administration, economy, treasury, judicial system, and security structure (UNSC, 2010) Widespread roadblocks to extract payments Limited delivery of formal government services (education, health, water)

13 The Ivorian Conflict During the conflict Relatively few casualties annual battle fatalities ~ 600 (2002, 2003) Large displacement of populations 2.7 million, including the internally displaced and 4 million (includes evacuees and refugees to Liberia, Sierra Leone, Mali, Guinea, Burkina Faso, Ghana) Widespread harassment of foreigners and migrant workers The end March 2007 Ouagadougou Political Accord Occasional bouts of violence throughout 2007

14 Data and Empirical Strategy

15 Cross-Sectional Household and Individual Data Before: 2002 HLSS During: 2006 MICS3 survey Children born before and after the start of the war After: 2008 HLSS Questions on conflict-induced victimization Economic losses Health impairment Displacement Other forms of victimization All surveys: Socio-economic characteristics of hhs and individuals HAZ for ~15,000 children aged 6-60 months old

16 Conflict Event Data Armed Conflict Location and Event Data (ACLED) Exact dates and locations of conflict events: violence against civilians, riots, protests, and armed battles. Conflict regions Provinces for which ACLED reports at least one episode of violence from September 2002 to December 2007 Within each location and for each year, incidents are manually matched to children s province-of-residence and year-of-birth

17 Côte d Ivoire and ACLED data Shaded areas represent conflict regions. Darker shades indicate a greater number of conflict events reported in the ACLED dataset. In the legend, the No data category stands for no reported incidents in the dataset and is treated as zero exposure to conflict in the analysis. The category (8, 187] includes 12 provinces, some of which had relatively lowintensity conflict (between 10 and 30 events) and some with relatively high-intensity conflict, such as Abidjan in the south (187 events), Bouaké in the center (62 events), and the province of Guiglo in the west (48 events).

18 Baseline Empirical Specification HAZ = α + δ + λ + β (Conflict Region *War Cohort ) + ε ijt j t jt 1 j t ijt Where HAZijt is the height-for-age z-score for child i in province j born in year t; α j are province-of-birth fixed effects, δ t are birthcohort fixed effects, λ jt are province-specific trends in cohort health. The coefficient estimate for β 1 captures the impact of conflict on the health of children born between September 2002 and December 2007 ( War Cohort ) and thus exposed to the conflict either in infancy or in utero. Allow for gender-specific impact through interaction with Female Controls: Child: ethnicity and religion Household head: age, education, gender Mother: age, education

19 Baseline Results

20 Impact of the Conflict on Child Health [1] [2] [3] [4] [5] [6] Conflict region*war Cohort ** ** (0.094) (0.149) Conflict region*war Cohort*Female Conflict region*exposure 0-24 months ** (0.092) (0.155) Conflict region*exposure at least 25 months ** ** (0.113) (0.159) Conflict region*exposure 0-24 months*female Conflict region*exposure at least 25 months*female Conflict region*exposure (no. of months) ** ** (0.003) (0.004) Conflict region*exposure (no. of months)*female Female 0.216*** 0.217*** 0.216*** 0.217*** 0.216*** 0.217*** (0.061) (0.060) (0.061) (0.060) (0.061) (0.060) Rural household *** *** *** *** *** *** (0.092) (0.094) (0.092) (0.094) (0.092) (0.094) Province-specific trends no yes no yes no yes Observations 15,151 15,151 15,151 15,151 15,151 15,151 R-squared Robust standard errors in parentheses, clustered at the province level. * significant at 10%; ** significant at 5%; *** significant at 1%. The dependent variable is the height-for-age z-score. All regressions include province fixed effects, month-of-birth fixed effects, and provincespecific time trends. In columns 2, 4, 6 the coefficient estimates on interactions between 'Conflict region' or 'Exposure' variables and the female dummy were jointly statistically insignificant and are not shown. All estimates are weighted by inverse sampling probability.

21 Impact of the Conflict on Child Health [7] [8] [9] [10] [11] [12] Conflict region*war Cohort *** *** (0.077) (0.134) Conflict region*war Cohort*Female (0.103) (0.107) Conflict region*exposure 0-24 months ** *** (0.124) (0.166) Conflict region*exposure at least 25 months ** ** (0.101) (0.161) Conflict region*exposure 0-24 months*female 0.283** 0.332** (0.108) (0.133) Conflict region*exposure at least 25 months*female (0.080) (0.078) Conflict region*exposure (no. of months) *** ** (0.002) (0.004) Conflict region*exposure (no. of months)*female (0.002) (0.002) Female (0.120) (0.121) (0.120) (0.121) (0.120) (0.121) Rural household *** *** *** *** *** *** (0.083) (0.085) (0.083) (0.085) (0.083) (0.085) Province-specific trends no yes no yes no yes Observations 15,151 15,151 15,151 15,151 15,151 15,151 R-squared Robust standard errors in parentheses, clustered at the province level. * significant at 10%; ** significant at 5%; *** significant at 1%. The dependent variable is the height-for-age z-score. All regressions include province fixed effects, month-of-birth fixed effects, and provincespecific time trends. In columns 2, 4, 6 the coefficient estimates on interactions between 'Conflict region' or 'Exposure' variables and the female dummy were jointly statistically insignificant and are not shown. All estimates are weighted by inverse sampling probability.

22 Baseline with Controls [1] [2] [3] [4] [5] [6] [7] [8] [9] [10] [11] [12] Conflict region*war Cohort ** ** ** *** (0.144) (0.156) (0.135) (0.145) Conflict region*war Cohort*Female (0.106) (0.114) Conflict region*exposure 0-24 months * ** ** (0.170) (0.193) (0.190) (0.212) Conflict region*exposure at least 25 months ** ** ** ** (0.158) (0.165) (0.161) (0.167) Conflict region*exposure 0-24 months*female 0.312** 0.277* (0.145) (0.146) Conflict region*exposure at least 25 months*female (0.083) (0.090) Conflict region*exposure (no. of months) ** ** ** *** (0.004) (0.004) (0.004) (0.004) Conflict region*exposure (no. of months)*female (0.002) (0.002) Province-specific trends yes yes yes yes yes yes yes yes yes yes yes yes Child controls yes yes yes yes yes yes yes yes yes yes yes yes Household head controls yes no yes no yes no yes no yes no yes no Mother controls no yes no yes no yes no yes no yes no yes p-value F-test of zero effect of: Child ethnicity Child religion Household head's characteristics Mother's characteristics Observations 13,664 12,132 13,664 12,132 13,664 12,132 13,664 12,132 13,664 12,132 13,664 12,132 R-squared Robust standard errors in parentheses, clustered at the province level. * significant at 10%; ** significant at 5%; *** significant at 1%. The dependent variable is the height-for-age z-score. All regressions include province fixed effects, month-of-birth fixed effects, and provincespecific time trends. In columns 2, 4, 6 the coefficient estimates on interactions between 'Conflict region' or 'Exposure' variables and the female dummy were jointly statistically insignificant and are not shown. All estimates are weighted by inverse sampling probability.

23 Robustness to Alternative Baseline Group : significant socio-political crisis marked by collapse in economic activity and social spending Eliminate from control group children born after December 1999 and before September 2002, affected by end-1999 military coup

24 Robustness: Alternative Baseline Group [1] [2] [3] [4] [5] [6] [7] [8] [9] [10] [11] [12] Conflict region*war Cohort ** *** *** *** (0.323) (0.358) (0.301) (0.337) Conflict region*war Cohort*Female 0.240* (0.121) (0.136) Conflict region*exposure 0-24 months *** *** *** *** (0.305) (0.350) (0.303) (0.358) Conflict region*exposure at least 25 months ** ** ** ** (0.338) (0.373) (0.342) (0.383) Conflict region*exposure 0-24 months*female 0.557*** 0.555*** (0.119) (0.125) Conflict region*exposure at least 25 months*female (0.088) (0.111) Conflict region*exposure (no. of months) ** ** ** ** (0.008) (0.009) (0.008) (0.009) Conflict region*exposure (no. of months)*female (0.002) (0.003) Household head controls yes no yes no yes no yes no yes no yes no Mother controls no yes no yes no yes no yes no yes no yes p-value F-test of zero effect of: Child ethnicity Child religion Household head's characteristics Mother's characteristics Observations 10,128 8,977 10,128 8,977 10,128 8,977 10,128 8,977 10,128 8,977 10,128 8,977 R-squared Notes: Robust standard errors in parentheses, clustered at the province level. * significant at 10%; ** significant at 5%; *** significant at 1%. The dependent variable is the height-for-age z-score. All regressions include child controls, province fixed effects, month-of-birth fixed effects, and province-specific time trends. Controls as in Table 3. All estimates are weighted by inverse sampling probability.

25 Analysis by Sub-sample [1] [2] [3] [4] [5] [6] [7] [8] Head is Head is not Poor Non-poor Girls Boys Rural Urban educated educated Conflict region*war Cohort * * ** ** ** ** (0.268) (0.217) (0.269) (0.141) (0.238) (0.213) (0.274) (0.216) Conflict region*war Cohort*Female * (0.215) (0.178) (0.139) (0.179) (0.158) (0.224) Female 0.317** (0.129) (0.119) (0.129) (0.166) (0.069) (0.171) Rural household *** *** *** *** *** (0.093) (0.096) (0.082) (0.119) (0.083) (0.106) p-value t-test of equality of coefficients on Conflict Region*War Cohort across sub-samples Observations 6,700 8,030 7,340 7,811 8,753 6,398 6,696 8,429 R-squared Notes: Robust standard errors in parentheses, clustered at the province level. * significant at 10%; ** significant at 5%; *** significant at 1%. The dependent variable is the height-for-age z-score. All regressions include province fixed effects, month-of-birth fixed effects, and province-specific time trends. Households are classified as poor if an index of asset wealth is below average. The asset index is calculated based on seven types of assets: living in dwelling with cement walls, cement floor, metal or cement roof, electricity, phone, toilet, and access to natural gas, coal or electricity for cooking. The index is the first factor extracted using principal components analysis on the seven components, explains 47 percent of their joint variance, and has been standardized to have zero mean and unit variance. In all columns other than 3 and 4, the coefficient estimates on interactions between 'Conflict region' or 'War Cohort' and the female dummy are jointly statistically insignificant and not shown. Estimates are weighted by inverse sampling probability. Data sources: 2002 and 2008 Côte d'ivoire HLSS, 2006 Côte d'ivoire MICS3, and Raleigh et al. (2010).

26 Additional Robustness Checks Women who had children during the conflict Placebo tests using pre-war data Selective mortality

27 Women who had children during the conflict [1] [2] [3] 1=Educated 1=Married Age Conflict Region (0.070) (0.026) (0.687) Had Child During Conflict ** 0.094*** *** (0.025) (0.023) (0.366) Conflict Region*Had Child During Conflict (0.035) (0.023) (0.599) Constant 0.400*** 0.764*** *** (0.031) (0.020) (0.599) Observations 15,689 15,700 15,700 Notes: Robust standard errors in parentheses, clustered at the province level. * significant at 10%, ** significant at 5%, and *** significant at 1%. The sample contains all women surveyed in 2006 and 2008 that were of fertile age during the conflict (i.e., years old). Data sources: 2008 Côte d'ivoire HLSS, 2006 Côte d'ivoire MICS3 and Raleigh et al. (2010).

28 Placebo Tests: 1994 and 1998/1999 DHS [1] [2] [3] [4] [5] [6] Conflict Region*War Cohort (0.159) (0.158) (0.158) (0.263) (0.245) (0.260) Conflict Region*War Cohort*Female * (0.418) (0.413) (0.418) Conflict Region*Female 0.168** 0.126* 0.175** (0.073) (0.071) (0.068) War Cohort*Female (0.285) (0.284) (0.278) Female 0.116* 0.128** 0.121* (0.058) (0.057) (0.059) (0.061) (0.059) (0.061) Rural *** *** *** *** *** *** (0.067) (0.060) (0.064) (0.067) (0.057) (0.065) Mother controls no yes no no yes no Household head controls no no yes no no yes p-value F-test of zero effect of: Mother's characteristics Household head's characteristics Observations 4,076 4,066 4,042 4,076 4,066 4,042 R-squared Notes: Robust standard errors in parentheses, clustered at the province level. * significant at 10%%, ** significant at 5%, *** significant at 1%. The dependent variable is the height-for-age z-score for children aged 6 60 months. "War Cohort" is an indicator variable for children born during January 1997-December Mother controls include age, education (a dummy variable for literacy), ethnicity (Akan (reference category), Northern Mande, Southern Mande, Krou, Voltaique/Gur, and other), and religion (Christian, Muslim, and (reference category)). Household head controls include age, gender, and education (a dummy variable for literacy). All regressions include province fixed effects, month-of-birth fixed effects, and province-specific time trends. Estimates are weighted by inverse sampling probability. Data sources: 1994 and 1998/99 Côte d'ivoire DHS, and Raleigh et al. (2010).

29 Sex Ratios by Year of Birth

30 Conflict Impact Mechanisms

31 2008 (Post-Conflict) Survey Section Household victimization experiences due to the conflict We compiled a simple victimization index as an average of all questions To explore specific mechanisms, we also grouped all the war-experience questions into four categories

32 [1] [2] [3] Conflict Region Non-Conflict Region Difference in Means [1]-[2] Response Tabulations for Victimization Indices Economic losses Because of the conflict Were your assets/properties damaged? *** Did your revenues decrease? *** Did you lose your job? *** Did you lose your farm? ** Did you lose livestock? * Did you lose any other productive assets? *** Overall, has the conflict affected your life? *** Health impairment Do you experience conflict-related nightmares? Do you experience conflict-related ailments (anxiety, stress)? Did you experience any conflict-related illness? *** Have you consulted a psychologist? Displacement Have you been displaced by the conflict? *** Are you currently displaced by the conflict? Did your household move because of the conflict? *** Did you have to hide because of the conflict? *** Victim of violence Have you been a victim of conflict-related violence? *** Did you witness conflict-related deaths in the household? *** Have you been forced into begging or prostitution? *** Notes: * significant at 10%; ** significant at 5%; *** significant at 1%. The question "Overall, has the conflict affected your life?" refers to losses of economic activity, difficulties in caring for oneself or finding shelter, loss of employment, dropping out of school, losing parents, losing assets or goods, or experiencing complete destruction of assets or goods. The question "Have you been a victim of conflict-related violence?" refers to theft, rape, other sexual violence, being wounded, or experiencing other troubles. Significance levels (column 3) refer to one-sided t-tests of the null that variables means are higher inside than outside the conflict region. Estimates are weighted by inverse sampling probability.

33 Household Victimization Map Notes: Shaded areas represent regions where conflict-induced victimization was reported. Darker shades indicate a greater share of households reporting at least one level of victimization (one 'yes' answer to the questions underlying each index). The location of the cities on the map is approximate. Data source: Based on the 2008 Côte d'ivoire HLSS.

34 Maps by Type of Victimization Notes: Shaded areas represent regions where conflict-induced victimization was reported. Darker shades indicate a greater share of households reporting at least one level of victimization (one 'yes' answer to the questions underlying each index). In the legend, the No data category refers to the southern province Sassandra for which anthropometric information is missing for all observations in the 2008 survey.

35 Correlation Matrix for Number of Conflict Events and Share of Households Reporting At Least One Type of Victimization Conflict Region Victimized Victimized: Economic losses Victimized: Health Victimized: Displacement Victimized 0.249* Victimized: Economic losses 0.249* 0.992* Victimized: Health impairment 0.233* 0.661* 0.653* Victimized: Displacement * 0.526* 0.772* Victimized: Victim of violence 0.331* 0.554* 0.581* 0.832* 0.943* * indices statistical significance at the 5 percent level.

36 Impact of Victimization on Child Health: 2008 Survey [1] [2] [3] [4] [5] [6] Panel A. Full sample: Victimized * * ** * * (0.345) (0.343) (0.357) (0.361) (0.393) (0.394) Observations 2,026 2,026 1,975 1,975 1,821 1,821 R-squared Panel B. Non-migrants: Victimized ** ** ** ** ** ** (0.317) (0.315) (0.304) (0.314) (0.343) (0.358) Observations 1,686 1,686 1,642 1,642 1,509 1,509 R-squared p-value t-test that "Victimized" coefficients for migrant and non-migrant households are equal Province specific trends no yes no yes no yes Child controls no no yes yes yes yes Household head controls no no yes yes no no Mother controls no no no no yes yes Notes: Robust standard errors in parentheses, clustered at the province level. * significant at 10%; ** significant at 5%; *** significant at 1%. The top two rows report coefficients from the regressions on the full sample, while the next two rows refer to the sub-sample of non-migrant households (i.e., households that have lived in their current location since before the start of the war). All regressions include province fixed effects, month-of-birth fixed effects, and province-specific time trends. HAZ = α + δ + λ + β (Victimization ) + ε ihjt j t jt 2 h ijt

37 Impact of Victimization on Child Health (cont d) [7] [8] [9] [10] [11] [12] Panel A. Full sample: Victimized * * ** * ** * (0.443) (0.450) (0.482) (0.495) (0.443) (0.453) Victimized*Female (0.761) (0.785) (0.934) (0.945) (0.866) (0.885) Observations 2,026 2,026 1,975 1,975 1,821 1,821 R-squared Panel B. Non-migrants: Victimized * * * * * (0.573) (0.598) (0.640) (0.664) (0.611) (0.645) Victimized*Female (0.966) (0.987) (1.132) (1.148) (1.019) (1.045) Observations 1,686 1,686 1,642 1,642 1,509 1,509 R-squared p-value t-test that "Victimized" coefficients for migrant and non-migrant households are equal Province specific trends no yes no yes no yes Child controls no no yes yes yes yes Household head controls no no yes yes no no Mother controls no no no no yes yes Notes: Robust standard errors in parentheses, clustered at the province level. * significant at 10%; ** significant at 5%; *** significant at 1%. The top two rows report coefficients from the regressions on the full sample, while the next two rows refer to the sub-sample of non-migrant households (i.e., households that have lived in their current location since before the start of the war). All regressions include province fixed effects, month-of-birth fixed effects, and province-specific time trends.

38 Victimization as a Conflict- Impact Mechanism: Does the type of victimization matter? Does its impact vary with the place of residence (the covariate shock)?

39 [1] [2] [3] [4] [5] [6] 2008: Victimization Indices Panel A. Full sample: Victimized: Economic losses *** *** *** *** * * (0.345) (0.356) (0.355) (0.376) (0.408) (0.417) Victimized: Health impairment (0.305) (0.312) (0.308) (0.320) (0.290) (0.295) Victimized: Displacement * 0.587* (0.360) (0.359) (0.309) (0.306) (0.382) (0.381) Victimized: Victim of violence (0.371) (0.377) (0.387) (0.399) (0.390) (0.402) Observations 2,026 2,026 1,975 1,975 1,821 1,821 R-squared Panel B. Non-migrants: Victimized: Economic losses *** *** *** *** * * (0.353) (0.369) (0.352) (0.374) (0.405) (0.408) Victimized: Health impairment (0.439) (0.444) (0.435) (0.443) (0.423) (0.422) Victimized: Displacement (0.425) (0.445) (0.386) (0.409) (0.463) (0.488) Victimized: Victim of violence (0.407) (0.419) (0.415) (0.432) (0.418) (0.433) Observations 1,686 1,686 1,642 1,642 1,509 1,509 R-squared Province specific trends no yes no yes no yes Child controls no no yes yes yes yes Household head controls no no yes yes no no Mother controls no no no no yes yes

40 Joint impact of Conflict and Household Victimization on Child Health (Pre- and Post-Conflict Surveys). [1] [2] [3] [4] [5] [6] [7] [8] [9] [10] [11] [12] Panel A. Full sample: Conflict region*victimized * *** ** *** ** *** *** *** *** *** *** *** (0.573) (0.643) (0.537) (0.592) (0.507) (0.715) (0.513) (0.638) (0.465) (0.553) (0.508) (0.716) Victimized (0.442) (0.516) (0.383) (0.488) (0.386) (0.556) (0.328) (0.383) (0.335) (0.403) (0.388) (0.520) Conflict region*victimized*female (0.865) (0.943) (0.830) (0.863) (0.883) (0.929) Observations 7,673 7,673 7,594 7,594 6,835 6,835 7,673 7,673 7,594 7,594 6,835 6,835 R-squared Panel B. Non-migrants: Conflict region*victimized *** *** * *** ** *** ** *** ** *** (0.570) (0.655) (0.592) (0.663) (0.599) (0.746) (0.646) (0.812) (0.660) (0.805) (0.592) (0.825) Victimized (0.516) (0.660) (0.478) (0.619) (0.495) (0.615) (0.553) (0.670) (0.602) (0.708) (0.557) (0.639) Conflict region*victimized*female (1.215) (1.274) (1.176) (1.224) (1.208) (1.306) Observations 5,816 5,816 5,757 5,757 5,144 5,144 5,816 5,816 5,757 5,757 5,144 5,144 R-squared Province specific trends no yes no yes no yes no yes no yes no yes Child controls no no yes yes yes yes no no yes yes yes yes Household head controls no no yes yes no no no no yes yes no no Mother controls no no no no yes yes no no no no yes yes Notes: Robust standard errors in parentheses, clustered at the province level. * significant at 10%; ** significant at 5%; *** significant at 1%. The top two rows report coefficients from the regressions on the full sample, while the next two rows refer to the sub-sample of non-migrant households (i.e., households that have lived in their current location since before the start of the war). All regressions include province fixed effects, month-of-birth fixed effects.,

41 Conclusions

42 Main Results Confirm results from earlier studies: Children in conflict regions experienced significant health setbacks compared to children from less affected regions: s.d. It is remarkable how consistent the negative effect of conflict on HAZ is across countries and conflicts Provide direct evidence on the mechanisms at work: household war-related victimization along several dimensions

43 Main Results (cont d) Mechanisms: Household victimization has a negative impact on child health Index de-composition: War-induced economic losses have a particularly strong effect on child health in conflict-affected regions Displacement : There are no statistically significant differences in the effects of victimization for migrant vs. non-migrant households

44 Policy Implications Loss in child health is cumulative, children stagnate and follow a different growth path Studies show that Shorter individuals have worse education and employment outcomes, and lower wages (especially males) Early, targeted intervention is key Cash transfers, programs to improve nutrition, create employment and restore household revenues at the same time would be the key instruments

45 Promising Directions for Research Few studies on the effect of development aid or other interventions in post-conflict settings Arcand and Wouabe (2009) Community assistance programs are not less effective in regions with higher conflict intensity Relatively small investments in a post-conflict environment can have a large impact on well-being