Policy Research Working Paper 9430 The Scars of Civil War The Long-Term Welfare Effects of the Salvadoran Armed Conflict Pablo Acosta Javier E. Baez Germán Caruso Carlos Carcach Poverty and Equity Global Practice October 2020 Policy Research Working Paper 9430 Abstract This paper estimates the long-term effects on human capital during the civil war saw a reduction in their probability of accumulation and subsequent labor market outcomes of being employed by 6 percentage points, and of getting a in utero and early childhood exposure to the civil war in high-skilled job by 5 percentage points, 20 to 30 years after El Salvador (1980–92), the second longest and deadliest it happened. The civil war also reduced their education by civil conflict in Central America. Identification is obtained 0.8 year, as well as their enrollment and literacy rates. Sub- from spatial and intertemporal variation in the intensity of group analysis indicates that exposed males and indigenous the conflict drawn from historical archive data comprising groups experienced the largest losses in human capital and records of human casualties, disappearances, and refugees. had weaker performance in the labor market. The results show that people born in highly violent areas This paper is a product of the Poverty and Equity Global Practice. It is part of a larger effort by the World Bank to provide open access to its research and make a contribution to development policy discussions around the world. Policy Research Working Papers are also posted on the Web at http://www.worldbank.org/prwp. The authors may be contacted at gcaruso@worldbank.org. The Policy Research Working Paper Series disseminates the findings of work in progress to encourage the exchange of ideas about development issues. An objective of the series is to get the findings out quickly, even if the presentations are less than fully polished. The papers carry the names of the authors and should be cited accordingly. The findings, interpretations, and conclusions expressed in this paper are entirely those of the authors. They do not necessarily represent the views of the International Bank for Reconstruction and Development/World Bank and its affiliated organizations, or those of the Executive Directors of the World Bank or the governments they represent. Produced by the Research Support Team The Scars of Civil War: The Long-Term Welfare Effects of the Salvadoran Armed Conflict Pablo Acosta (World Bank) Javier E. Baez (World Bank) Germán Caruso (World Bank) Carlos Carcach (Escuela Superior de Economía y Negocios, El Salvador) JEL: D31, I00, J13 Keywords: Armed conflict, Long term impacts, El Salvador 1. Introduction Internal conflicts are a constant threat to economic development worldwide. Over half of the nations have experienced an armed conflict at some point in the last 50 years (Petterson and Wallensteen 2015). After a drop at the end of the 20th century, there has been an upward trend in the number of internal conflicts since the early 2000s. The Uppsala Conflict Data Program (UPCP) globally recorded 40 armed conflicts with at least 25 battle deaths per annum for 2014, the highest number on record after 1999. Over 25% of these conflicts were civil wars that caused more than 1,000 battle deaths in a single year. The data also show that in 2014 the number of people forcibly displaced by armed conflicts worldwide reached close to 60 million, making it the highest annual increase since comparable records began in 1989 (United Nations High Commission for Refugees 2014). Civil conflicts are a source of huge devastation, ranging from loss of lives and forced displacement, destruction of human capital, physical infrastructure and private property to disruption of economic and political systems. Internal warfare has spillover effects in the form of refugees, crime and illegal trade into neighboring nations. At the macro level, countries often see growth slow down shortly after armed conflicts surge. Over time, however, most countries experience rapid post-war economic recovery, including convergence of key factors of production (population and human capital) and standards of living (Blattman and Miguel 2010; Miguel and Roland 2006; Brakman et al. 2004; Davis and Weinstein 2002). Yet, at the micro level, a growing body of empirical research has been uncovering how wars inflict a subtler but long-lasting burden on the human capital of affected populations, undermining their long-term productivity and well- being (Bundervoet et al. 2009; Akbulut- Yuksel, 2017; Akresh et al. 2011). This paper contributes to the latter literature by providing evidence of the long-run human capital and productivity costs of civil wars. In particular, the paper investigates the labor market performance of adults who were exposed in utero or early childhood to the violence generated by the civil war in El Salvador (1980-1992), one of the longest and bloodiest conflicts in Central America. At the start of the war, El Salvador was already one of the poorest and most unequal countries in Latin America. In order to address the endogeneity between the socioeconomic conditions and the civil war, we exploit exogenous intertemporal and spatial variations in the intensity of the 2 conflict as a source of identification, a plausible instrument to estimate the impact of civil war exposure on economic outcomes. As such, this paper relates to the empirical approaches employed by other important papers in the literature (Akresh et al., 2012; Blattman and Annan, 2011; Bundervoet et al., 2009). We estimate a difference-in-differences model, factoring an index variable for war exposed areas (ranking the different Salvadoran municipalities) along with a cohort-based measure of years of exposure to the war. We combine data from the 2007 national population census with event information on the timing and geographic location of the conflict from historical data archives between 1981 and 1985. The results indicate that the consequences of the conflict span far beyond the direct human and economic costs and are still being realized 20 to 30 years later as individuals affected early in life, who are still alive today, are entering the labor market. Exposure to the civil war reduced their probability of being employed by 6 percentage points and the probability of getting a high-skill job by 5 percentage points in the affected municipalities. In disentangling the mechanisms at play, we find that disruptions in human capital formation through reduced inputs (such as weaker nutrition, lower school enrollment, and destruction of facilities) played an important role. The country’s most war-torn regions saw a reduction in school attainment of around 0.85 year. These negative effects are robust to the inclusion of several control variables, time trends, different sample specifications and placebo tests. Other outcomes related to psychological effects, shorter life expectancy and overall health outcomes were not explored in the paper due to lack of data. However, traditional literature for similar conflict situations has shown that these outcomes are likely to be affected as well. The remainder of the paper is organized as follows. The next section provides an overview of the civil war in El Salvador and the country’s context. Section 3 describes the data used in the analysis. Section 4 explains the empirical strategy, and Section 5 presents the empirical findings, including a discussion of the mechanisms behind the impacts and the results of the robustness analysis. Finally, Section 6 concludes. 3 2. Civil War in El Salvador The roots of the civil conflict in El Salvador date back to the late 19th century. Land reforms initiated in the 1880s sought the abolition of communal land to support the expansion of an export- oriented coffee-based agriculture (Pérez Brignioli 2001). Land privatization was the embryo of social tensions and socioeconomic inequality, pitting a growing mass of dispossessed peasants attached to a strong culture of access to common land against a government-private sector alliance seeking the unlimited expansion of a coffee-based economy. Social discontent continued to grow through most of the first half of the 20th century, compounded by sharp drops in coffee prices, economic collapse following the Great Depression in 1929, frequent uprisings of peasants and indigenous people against the government and the landed elite, brutal repression and recurrent political instability. The following decades were marked by economic volatility, price fluctuations of major export crops such as coffee, cotton and sugarcane, and the oil crisis which led to rising food prices and decreased agricultural output. The ousting of the military President Carlos Romero by a group of reformists and moderate officers on October 15, 1979 along with the assassination of a large number of Salvadorans such as business people, union leaders, politicians and human rights activists, tipped the sporadic violence of the 1970s between left and right wings into full-scale civil war. Indirect international involvement fueled additional violence and a human rights crisis. The Salvadoran conflict followed the victory of the Sandinista revolution in Nicaragua and the final years of the Cold War. This was seen by the U.S. government as a serious threat of communist expansion in the region, prompting it to increase its military aid to the government and right-wing groups. In contrast, countries from the Eastern Bloc, in particular the Soviet Union, were crucial allies to the leftist factions. The full-fledged war lasted for 12 years and resulted in more than 80,000 deaths. Given the small size of the population at the time of the war (around 4.6 million), the death toll is substantial, corresponding to an average homicide rate of approximately 150 deaths per 100,000 inhabitants. Beyond this burden of death, 8,000 disappearances were recorded, and more than a million people (nearly 22% of the pre-war national population) were forcibly displaced internally (Benítez Manaut, 1988). Children were recruited as guerilla fighters, and death squads deliberately targeted civilians in rural villages and urban centers. In addition to the human costs, the conflict led to 4 widespread devastation and destruction of key infrastructure. The intensity of the conflict peaked in years 1982 and 1983. During 1980 and 1992, annual GDP per capita growth averaged -1.9% with a lower peak of -13.3% in 1980, and annual inflation averaged 18.1% with a high peak of 31.9% in 1986 (World Development Indicators, 2019). Estimates based on synthetic counterfactual methods estimate an average annual loss of GDP per capita of 21.5% during the period of the conflict (Costalli et al. 2014). The first 3 years (1981-1983) represent the most violent period of the civil war. During this period, the USIP’s Commission on the Truth received direct complaints concerning 5,375 victims of serious acts of violence (1993); there were at least 12,501 deaths recorded (Christian Legal Aid); and 164,297 displaced persons were reported in 1981 (Montes, 1986). However, in computations based on our own collected data for the years 1981 to 1985 (our sources of data include OMS reports, historic municipal records and local newspaper reports about incidents at the time of the civil war), we assessed the number of deaths due to war operations at 17,912, the number of homicides at 10,635, and the number of displaced persons at 204,805. Figure 1 shows a map with the number of deaths by municipality due to war between 1981 and 1985. By 1985, 171 of the 262 municipalities in the country, 65.2% of the total, had been (directly) affected by the war. 3. Data We use two sources of data for this paper. The first is the Salvadoran National Population and Housing Census, collected by the National Institute of Statistics and Census (DIGESTYC) between May 12 and 25, 2007. The census surveyed close to 1.7 million households, which corresponds to an official count of approximately 5.7 million inhabitants. The instrument used for this paper asked questions about the characteristics of the dwelling (i.e. its geographic location, ownership, construction materials, access to basic services, among others), household composition and structure, ownership of basic assets and individual-level variables such as demographics (including ethnicity), school attainment, health status, labor participation, occupation, type of job, hours worked, remittances, migration and mortality. Additionally, the census collected information on the date and municipality of birth for each individual counted. 5 We use a 10% random sample of the census to implement our empirical analysis. Summary statistics of the main variables used in this paper, including the dependent variables of interest, are shown in Table 1. Statistics related with human capital show evidence that 79.8% of the Salvadoran population can read and write, and most of them attend or attended an educational center (77.9%). On average, the total years of education for this population is 8.1 years. Labor market indicators show that 49.5% is employed and 25.9% is underemployed (working fewer than 40 hours per week). Internal migration has been important too, with 29.1% migrated, and 9.6% migrating during the civil war. Finally, population characteristics show that 54.9% of the population are women and 2.3% are indigenous. The second source collects data by municipality of victims during the period of civil war in El Salvador. The database contains different categories of victims between 1981 and 1985: homicide, suicide, transit accident, war victim, and displaced person. This study uses war victims and homicides as indicators of violence during the civil war. Evidence shows that during the civil war more than 76% of the victims recorded died because of war or homicide. Not all municipalities faced the same intensity of violence. This is reflected in the variation of homicide rates across municipalities. In order to identify the most affected municipalities without defining an arbitrary threshold, this work uses a cluster analysis method which allows us to identify the geographical areas with high and low intensities of violence. In particular, this paper clusters municipalities according to the number of victims. 1 The analysis is robust to different clustering techniques, outliers, and misclassification. 2 1 For splitting the sample, we have proceeded to cluster the municipalities according to their intensity of violence. Instead of relying on an aggregate indicator or focusing on one of them individually, our multivariate procedure allows us to exploit the multi-factor nature of civil war processes. Consequently, we build the clusters using k-means, the L1- norm, and the index of Caliński and Harabasz in order to determine the optimal number of clusters, which in this case happens to be two. We run this procedure 100 times, assigning a different random seed each time. This procedure provides us with 6 potential groupings.While there is a set of municipalities that is always assigned to the same group (35 to the high intensity of violence group and 28 to the other), there are 10 municipalities that switch groups. We also group the municipalities using the composite index (univariate procedure). The clusters according to the composite index are stable and never change. 2 Method for Cluster analysis (Caliński and Harabasz 1974). 6 4. Identification Strategy We employ a difference-in-differences strategy that exploits plausibly exogenous within- municipality cross-cohort variation in the intensity during the first five years of the war (1981- 1985), the most violent period of the whole conflict. Identification is derived from a comparison of the outcomes of similarly aged individuals in more and less affected municipalities. Basically, the empirical design relies on two sources of variation: i) across space, i.e., comparison of identical birth cohorts (1981–1985) from “high” and “low” intensity geographic areas; and ii) across time, i.e., comparison in “high” intensity areas of treated cohorts (born between 1981 and 1985) and cohorts old enough to have completed their education at the time of the shock (born right before 1961). The underlying identifying assumption is that in absence of the war, the observed differences across different cohorts in each outcome would be similar across more and less affected municipalities. We present evidence below supporting the presence of parallel trends in labor market and schooling outcomes. The main estimation equation is specified as follows: = + � ℎ � + + + ′ + (1) where is the outcome of interest for individual i, from cohort t and born in municipality j; are cohort fixed effects; are municipality fixed effects; is a set of regional- and individual-level control variables, including regional trends and gender fixed effects; and is a random, idiosyncratic error term. As the name suggests, the variable “Affected District” is an indicator variable that takes value equal to one if the municipality of birth of individual i was among the deadliest during the war and zero otherwise. This variable captures geographical variation in the level of violence caused by the conflict as measured by the homicides and war victims between 1981 and 1985 in municipality j. The variable “Exposed Cohort” is an indicator variable that takes value equal to one if the individuals were born during the deadliest period of the war (1981-1985) and zero otherwise. This variable captures all the time variation of the level of violence at the time of birth. The average treatment effect β measures the impact of the war on the outcome of interest Y. To determine the affected municipalities, a cluster analysis methodology was used by municipality. Instead of defining an arbitrary threshold, we use a multivariate procedure which exploits clusters using k-means to identify the most violent group of municipalities, clustering 7 them into two groups based on the distribution of the municipalities’ rates of homicides and war victims as a proxy for level of violence. The clusters obtained are robust to multiple variations of the methodology. This model identifies the average impact of the war under the assumption of “common trends” between groups in the absence of conflict. That is, the estimations in this paper assume that the trajectory of the individuals born in less affected municipalities is a valid counterfactual of the trajectory of the individuals in the more affected municipalities. This paper finds that trends for cohorts born before 1981 are indeed parallel. Figure 2 shows an example of these trends for years of education. The average difference in years of education between individuals born in 1950 in the more affected and less affected municipalities is 2.5 years of education. However, this difference reduces to 1.5 years for individuals born in 1985, benefiting those born in less violent municipalities, which supports the validity of our assumption. 5. Empirical Analysis 5.1 Results i) Labor Market Outcomes Table 2 reports the results of estimating Equation 1 using as dependent variable several labor market outcome indicators. The model specification for each outcome includes controls for district fixed effects and regional trends. We initially investigate the impacts of the civil conflict on the labor engagement of individuals in the sample. The results show that two decades after the war finished, the employment rate of those born in a violent area at the time of the war was reduced by 6 percentage points due to the conflict (column 1). These impacts are large in a sample where the average labor force participation is 50%, and they are statistically significant at 99% of confidence. In columns 2, 3, and 4, we analyze the differential impact on other labor market outcomes. We find that being born in a highly violent area during the war also reduced the probability of later getting a high-skill job 3 by 5 percentage points. Our estimates indicate that in violent 3 High-skill jobs include managers, professionals, technicians and associate professionals, and armed forces occupations. On average, high-skilled workers earn $4 per day while low-skilled workers earn $1.38. 8 municipalities the war also increased people’s future probability of underemployment by 2.6 percentage points. Finally, the war also affected the probability of migrating internationally. Exposure at birth to the civil war increased the international migration rate by almost 2 percentage points. ii) Human capital Table 3 presents the results of our main specification for the effect of war exposure on human capital accumulation. We find that being born in a violent municipality at the time of the war reduced the years of education by 0.8 year (column 1). These impacts represent a reduction of 11% in a sample with a mean of eight years of education. In columns 2 and 3, we analyze the differential impact on other human capital outcomes. We find that exposure to the war also reduces the literacy rate by about 8 percentage points. In addition, our estimates indicate that being born in a municipality with high intensity of civil war violence decreased by 8 percentage points people’s probability of ever being enrolled at school. Finally, our results find that the war also increased child mortality by 8 percentage points. 5.2 Differential impacts by gender and for indigenous population Table 4 and Table 5 show the results of our main specification for the effect of war exposure by sex and ethnicity respectively, on both the probability of being employed and years of education. Our results show that the negative effects of the war are stronger for males and for the indigenous population. For instance, Table 4 shows that having been born in a violent municipality reduces the probability of being employed by about 8 percentage points for males and 4 percentage points for females. Moreover, having been born in a violent municipality reduces the years of education by 1 year for males compared to 0.7 year of education for females. Furthermore, in Table 5 we find that exposure to the war reduced the probability of being employed for indigenous people (11 percentage points) more than for non-indigenous (6 percentage points). Being exposed to the war had an impact on human capital accumulation, but the results for non-indigenous people are not significant and thus inconclusive. 9 5.3 Robustness checks We conducted several falsification tests and robustness checks on our identification strategy. First, we estimated our main specification for cohorts whose labor and educational outcomes were not at all affected by the war. In particular, we focused on those who completed their investments in human capital accumulation by the time the war started. Therefore, we defined as “non-exposed” those born before 1950. Table 6 shows no effect on human capital accumulation, except for child mortality for this non-exposed cohort. But results using the same control variables (Table 3, column 4) indicate that the effect on the exposed cohort was much greater. We also ran the same specification for outcomes such as age, which cannot be affected by the impact of the war. In Table 7 we show that the specification does not generate spurious significant coefficients. Finally, we simulate false locations and false times for the armed conflict. To this end, we ran our same regression 100 times with randomly generated armed conflicts (either different departments or times from the control group). In 96 of the trials our results were not significant at the 10% level. 6. Conclusions This paper has shown the long-lasting effects of El Salvador’s civil conflict in the 1980s on employment and human capital outcomes, particularly on education and child mortality. Our findings are consistent with those reported in the existent literature. Cohorts that were more exposed to conflict, visà-vis those non-exposed or less-exposed cohorts, are more likely to remain trapped in low-income situations and have fewer labor-market opportunities due to deteriorated human capital. The war provoked a welfare reduction for the people exposed to it as well as long-term effects for future generations. These effects have not been uniformly distributed across the society. Thus, the stronger the exposure was, the more difficult it is to make these scars disappear. In particular, these impacts would require stronger state presence and increased governance in the form of development policies aimed at reducing regional gaps in territories that were greatly affected by the civil war. 10 These findings call for policy actions to help victims of conflict and their descendants improve their standard of living. While many countries have instituted war reparation benefits to victims (including El Salvador), interventions targeted to actively improve education and labor market policies are recommended since they can better help to overcome the main barriers for sustained income generation. However, more research is needed to identify the best policy outcomes to improve the life quality of groups specifically affected by conflict versus others affected by recurrent economic shocks that can have different long-lasting impacts on individuals. 11 References Akbulut-Yuksel, M. (2017). War during childhood: The long run effects of warfare on health. Journal of health economics, 53, 117-130. Akbulut-Yuksel, M. (2009). Children of war: The long-run effects of large-scale physical destruction and warfare on children. IZA Discussion Paper 4407. Akresh, R., Bhalotra, S., Leone, M., & Osili, U. O. (2012). War and stature: growing up during the Nigerian Civil War. The American Economic Review, 102(3), 273-277. Akresh, R., Verwimp, P., & Bundervoet, T. (2011). Civil war, crop failure, and child stunting in Rwanda. Economic Development and Cultural Change, 59(4), 777-810. Akresh, R., Caruso, G. D., & Thirumurthy, H. (2014). Medium-Term Health Impacts of Shocks Experienced In Utero and After Birth: Evidence from Detailed Geographic Information on War Exposure (No. w20763). National Bureau of Economic Research. Almond, D. (2006). Is the 1918 Influenza Pandemic Over? Long-Term Effects of In Utero Influenza Exposure in the Post-1940 U.S. Population. Journal of Political Economy, 114, 672-712. Annan, J., Blattman, C., Mazurana, D., & Carlson, K. (2011). Civil war, reintegration, and gender in Northern Uganda. Journal of conflict resolution, 55(6), 877-908. Blattman, C., & Miguel, E. (2010). Civil war. Journal of Economic literature, 48(1), 3-57. Benítez Manaut, R. 1988, El Salvador 1984-1988: guerra civil, economía y política. Realidad, Año 1, Número 6, p. 527-540. Bundervoet, T., Verwimp, P., & Akresh, R. (2009). Health and civil war in rural Burundi. Journal of human Resources, 44(2), 536-563. Caliński, T., & Harabasz, J. (1974). A dendrite method for cluster analysis. Communications in Statistics-theory and Methods, 3(1), 1-27. Carcach, C. A. (2008). El Salvador. Mapa de violencia y su referencia histórica. Centro de monitoreo y evaluación de la violencia desde la perspectiva ciudadana, San Salvador. Commission on the Truth for El Salvador. (1993). From madness to hope: the 12-year war in El Salvador. Report of the Commission on the Truth for El Salvador. UN Security Council, Annex S/25500, 1993, 5-8. Cuando la guerra bajó de los cerros. 2015. (October 14). La Prensa Gráfica. Retrieved October 15, 2015, from http://www.laprensagrafica.com/revistas/septimo-sentido/72963-cuando-la-guerra- bajo-de-los-cerros 12 Davis, D. R., & Weinstein, D. E. (2002). Bones, bombs, and break points: the geography of economic activity. American Economic Review, 92(5), 1269-1289. de Walque, D., & Verwimp, P. (2009). The demographic and socio-economic distribution of excess mortality during the 1994 genocide in Rwanda (No. 4850). The World Bank. Torres Rivas, E. 1991, Crisis and conflict, 1930 to present. In Bethell, L. (ed.), Central America since Independence, p. 69-118. Cambridge: Cambridge University Press. Larrain, B., F, 2003, El Salvador: ¿Cómo volver a crecer?, Banco Interamericano de Desarrollo, Serie de Estudios Económicos y Sectoriales, RE2-03-006. Menjívar, R., 1980, Acumulación Originaria y Desarrollo del Capitalismo en El Salvador. San José, Costa Rica: EDUCA: Miguel, E., & Roland, G. (2006). The long run impact of bombing Vietnam (No. w11954). National Bureau of Economic Research. Montes, S., 1986, El Salvador: En Busca de Soluciones para los Desplazados. Instituto de Investigaciones. Instituto de Derechos Humanos. San Salvador: Universidad Centroamericana José Simeón Cañas. Pérez Brignoli, Héctor (2001) La rebelión campesina de 1932 en El Salvador. En El Salvador, 1932. Los sucesos políticos. Anderson, Thomas R., pp. 17-54, Dirección de Publicaciones e Impresos, CONCULTURA, San Salvador Ungo, G.M., 1984, Causas y perspectivas de la guerra civil en El Salvador. Revista Mexicana de Sociología, vol. 46, No. 3, p. 143-154. United Nations High Commission for Refugees and United Nations Development Programme. 2015. Regional Refugee and Resilience Plan 2015-2016, in response to the Syria crisis. Annual Report. Walter, K., 2008, Estados Unidos y El Salvador: La década de 1980. Estudios Centroamericanos, vol. 63, No. 713-714, p.197-200. 13 Figure 1: Average Number of Deaths Due to War by Municipality 1981-1985 Source: Own elaboration 14 Table 1: Summary Statistics, 2007 Census Standard Variables Mean Deviation Individual Characteristics Male 45.1% 0.4976 Female 54.9% 0.4976 Indigenous 2.3% 0.0476 Years of Education 8.12 4.8378 Literate 79.8% 0.4019 Attend or Attended an educational center 77.9% 0.4151 Employed 49.5% 0.5000 High-Skill Job (military, directors and managers, professionals, technicians) 16.6% 0.3725 Underemployed (<40 hours) 25.9% 0.4379 Underemployed (<30 hours) 11.6% 0.3201 Internal migration 29.1% 0.4544 Migrate During the Civil War 9.6% 0.2948 Household Characteristics Someone in the household migrated to another country 12.5% 0.3302 Last son is alive 96.3% 0.1892 15 Figure 2: Parallel Trends Before 1981 Between Violent and Non-violent Municipalities Source: Own elaboration 16 Table 2: Effects of the Civil War on Labor Market Outcomes (1) (2) (3) (4) VARIABLES Employed High-Skill Job Underemployment International Migration Exposed Cohort*Affected District -0.0622*** -0.0602*** -0.0395*** -0.0497*** 0.0286** 0.0259** 0.0175*** 0.0173* (0.0124) (0.0147) (0.0124) (0.0148) (0.0108) (0.0094) (0.0035) (0.0090) Gender Yes Yes Yes Yes Yes Yes Yes Yes Cohort FE Yes Yes Yes Yes Yes Yes Yes Yes District FE No Yes No Yes No Yes No Yes Regional Trends No Yes No Yes No Yes No Yes Observations 1,269,242 1,269,242 682,531 682,531 561,957 561,957 1,263,096 1,263,096 Robust standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1 17 Table 3: Effects of the Civil War on Human Capital (1) (2) (3) (4) VARIABLES Years of education Literate Attended to school Child Mortality Exposed Cohort*Affected District -0.6358** -0.8493*** -0.0882*** -0.0804*** -0.0882*** -0.0831*** 0.0078** 0.0831*** (0.2836) (0.2767) (0.0166) (0.0098) (0.0166) (0.0099) (0.0033) (0.0099) Gender Yes Yes Yes Yes Yes Yes Yes Yes Cohort FE Yes Yes Yes Yes Yes Yes Yes Yes District FE No Yes No Yes No Yes No Yes Regional Trends No Yes No Yes No Yes No Yes Observations 990,964 990,965 1,269,242 1,269,242 1,269,242 1,269,243 541,019 541,020 Robust standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1 18 Table 4: Effects of the Civil War by Gender Male Female (1) (2) (3) (4) Years of Years of Employed Employed VARIABLES education education Exposed Cohort*Affected District -0.0797*** -1.0064*** -0.0403** -0.7127** (0.0144) (0.2713) (0.0141) (0.2871) Cohort FE Yes Yes Yes Yes District FE Yes Yes Yes Yes Regional Trends Yes Yes Yes Yes Observations 571,757 464,556 697,485 526,408 Robust standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1 19 Table 5: Effects of the Civil War for Indigenous Populations Indigenous Non-indigenous (1) (2) (3) (4) Years of Years of Employed Employed VARIABLES education education Exposed Cohort*Affected District -0.1110** -1.5950** -0.0602*** -0.8489*** (0.0380) (0.9736) (0.0147) (0.2776) Gender Yes Yes Yes Yes Cohort FE Yes Yes Yes Yes District FE Yes Yes Yes Yes Regional Trends Yes Yes Yes Yes Observations 2,766 1,780 1,266,476 989,184 Robust standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1 20 Table 6: Effects of the Civil War on Human Capital for Non-Exposed Cohorts (1) (2) (3) (4) Years of VARIABLES education Literate Attended to school Child Mortality Born Before War*Affected District -0.2651 0.0305 0.0279 0.0071** (0.1620) (0.0244) (0.0262) (0.0032) Gender Yes Yes Yes Yes Cohort FE Yes Yes Yes Yes District FE Yes Yes Yes Yes Regional Trends Yes Yes Yes Yes Observations 663,519 1,066,175 1,066,175 533,552 Robust standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1 21 Table 7: Effects of the Armed Conflict on Unrelated Outcomes (1) (2) (3) Age Married Number of VARIABLES Children Exposed Cohort*Affected District 0.0147 0.0149 -0.0510 (0.0088) (0.0180) (0.0526) Gender Yes Yes Yes Cohort FE Yes Yes Yes District FE Yes Yes Yes Regional Trends Yes Yes Yes Observations 1,269,242 1,257,459 1,269,242 Robust standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1 22