GPS Data, War Exposure, and Child Health *

Transcription

1 GPS Data, War Exposure, and Child Health * Richard Akresh University of Illinois at Urbana-Champaign, NBER, BREAD, and IZA German Daniel Caruso University of Illinois at Urbana-Champaign Harsha Thirumurthy University of North Carolina at Chapel Hill and BREAD October 15, 2014 Abstract This is the first paper to use household survey data from multiple countries (Ethiopia and Eritrea) involved in an international war to measure the impact of conflict on children s health. The paper uniquely incorporates GPS information on the distance between survey villages and conflict sites to accurately measure a child s war exposure; results indicate the negative impacts are 35-65% larger than they would be if exposure is measured at the imprecise regional level. The identification strategy uses event data to exploit exogenous variation in the conflict s geographic extent and timing as well as the exposure of different birth cohorts while in utero or early childhood. Results indicate that war-exposed children have lower height-for-age Z-scores, with children in the war-instigating and losing country (Eritrea) suffering more than the winning nation (Ethiopia). Results are robust to including region-specific time trends, alternative conflict exposure measures, and addressing potential bias due to selective migration. Keywords: Child health; Conflict; Economic shocks; Africa JEL classification: I12, J13, O12 * Contact Information: Richard Akresh, University of Illinois at Urbana-Champaign, Department of Economics, 1407 West Gregory Drive, David Kinley Hall, Room 214, Urbana, IL akresh@illinois.edu; German Daniel Caruso, University of Illinois at Urbana-Champaign, Department of Economics, 1407 West Gregory Drive, David Kinley Hall, Room 214, Urbana, IL gdcarus2@illinois.edu; Harsha Thirumurthy, University of North Carolina at Chapel Hill, Department of Health Policy and Management, CB 7411, Chapel Hill, NC harsha@unc.edu.

2 1. Introduction Environmental and economic shocks are important determinants of current health and well-being in low-income countries where many subsistence-level households have limited risk-coping mechanisms. The effects of these shocks, which can range from sudden illnesses to weatherrelated disasters to armed conflicts, and households responses to them have been a central focus of research on health and economic outcomes in developing countries (Maccini and Yang, 2009; Maluccio et al., 2009, Akresh et al., 2012). There is extensive research on the importance of shocks experienced in utero and early childhood (for a comprehensive review see Almond and Currie, 2011). The occurrence of these shocks has been considered to be especially harmful not only because they may affect health outcomes in the short-term but also because they may influence health and economic outcomes in adulthood. Identifying the causal effects of early life shocks on subsequent health outcomes in a convincing manner has proven to be challenging, however, due to limitations in available data and a number of confounding factors. Several studies that have used longitudinal data coupled with information on early-life or in utero exposure to shocks and shown persistent and long-term effects on health, education, and economic outcomes, but our understanding of how shocks such as armed conflicts affect future health outcomes remains limited due to a number of limitations in available data. The possibility that early-life growth disturbances might affect future health outcomes of children is particularly relevant in sub-saharan Africa, where armed conflict has occurred with greater frequency than in many other regions of the world. Nearly 70 percent of all countries in sub-saharan Africa are home to an active armed conflict or a recently ended one (Raleigh et al., 2010). In many instances, particularly in developing countries, the conflicts are started or are 2

3 exacerbated by territorial disputes. 1 Despite the casualties and destruction caused by conflicts and the potential for such shocks to affect various indicators of well-being, the impacts of conflict on health have received surprisingly limited focus in the literature. 2 Limitations in available household survey data, including an inability to clearly identify the extent to which survey households were affected by conflicts, as well as a lack of well-suited identification strategies have contributed to incompleteness and weaknesses in the current evidence base. This paper studies the overall effects of the Eritrea-Ethiopia conflict on the future health status of children by using survey data from both countries that include detailed information on the geographic location of households and taking advantage of variation in the timing of the conflict with respect to when the children were born. By using data on the geographic location of children s households, the paper is able to better identify the effect of conflict exposure than many previous studies. In addition, the paper analyzes the health effects of in utero exposure compared to early childhood exposure to the conflict, thereby assessing the relative importance of growth disturbances during these two periods. The paper also estimates the effect of the conflict using several alternative methods of determining the extent to which children may have been exposed, thereby using a variety of different approaches to estimate the degree to which the conflict affected the health outcomes. Importantly, our paper also begins to examine several plausible mechanisms by which the conflict may have affected health outcomes 1 The United States Central Intelligence Agency World Factbook (2010) lists over 180 regions in the world that have existing disputes over international land or sea boundaries or have resource or resident disagreements; 41 of these disputes are in sub-saharan Africa. 2 Seminal work on conflict focuses on understanding the causes and spread of war and its role in reducing growth (Collier and Hoeffler, 1998; Miguel, Satyanath, and Sergenti, 2004; Guidolin and La Ferrara, 2007; Do and Iyer, 2010). The magnitude of conflict s long-term negative economic consequences are debated in the literature (see Davis and Weinstein (2002) for Japan; Brakman, Garretsen, and Schramm (2004) for Germany; Bellows and Miguel (2009) for Sierra Leone). There is also a growing literature examining the relationship between conflict and education outcomes (Ichino and Winter-Ebmer, 2004; Akresh and de Walque, 2008; Swee, 2009; Miguel and Roland, 2011; Shemyakina, 2011). Research focusing exclusively on soldiers finds large negative impacts on their earnings, and soldiers exposed to more violence face a harder time reintegrating into civilian society (Angrist, 1990; Imbens and van der Klaauw, 1995; Humphreys and Weinstein, 2007; Blattman and Annan, 2009). 3

4 of children, focusing in particular on the potential role of health-seeking behaviors such as use of antenatal care. The Eritrea-Ethiopia conflict was based on a territorial border dispute between the two countries. 3 When Eritrea, formerly a province of Ethiopia, became independent in 1993 following a long guerrilla conflict, sections of the new border were never properly demarcated. Full-fledged fighting began in May 1998 over these areas, which have been described as desolate and inconsequential. More than 300,000 troops were dug in and deadlocked on both sides of the border. Most of the conflict s casualties were soldiers, since most civilians left the conflict-torn areas, leaving the armies to fight over empty villages. The availability of Demographic and Health Survey data that were collected in both countries in the years following the conflict (Eritrea in 2002; Ethiopia in 2000 and 2005) provides a unique opportunity to study the effects of this conflict on subsequent health outcomes of children. The primary outcome that we study is the height-for-age Z-score, which is likely to capture lingering effects of any growth disturbances that may have occurred earlier in life during the conflict. This paper makes four main contributions to the literature on the impacts of shocks on children s welfare and goes significantly beyond previous research that examined the effects of the Eritrea-Ethiopia conflict on child outcomes in Eritrea only (Akresh, Lucchetti, and Thirumurthy, 2012). First, to correctly identify the impact of conflict-related shocks, this paper addresses the traditional difficulty in correctly classifying children s exposure. Whereas a standard approach taken is to compare large regions that did and did not experience fighting, we use global positioning system (GPS) data on the distance between survey villages and conflict 3 In the past 30 years, border wars were fought in Africa (Djibouti and Eritrea in 2008, Mauritania and Senegal starting in 1989, Burkina Faso and Mali in 1985, Ethiopia and Somalia in 1982), Asia (Cambodia and Thailand in 2008, India and Bangladesh in 2001, Israel and Lebanon starting in 2000, India and Pakistan in 1999, Thailand and Laos starting in 1987, India and China in 1987, Pakistan and India starting in 1984, Iran and Iraq starting in 1980, Vietnam and China starting in 1979), and South America (Ecuador and Peru in 1995, Ecuador and Peru in 1981). 4

5 sites to accurately measure the likely exposure that a child had to the conflict. We show that this approach makes a difference in the estimated effects of conflict, which decrease as distance to the conflict sites increases. We also present evidence that the results are robust to various other ways of classifying children s exposure and region of residence. Second, because of the fortuitous timing of the household survey data collection, we are able to explore whether effects of the conflict differ between children who were in utero at the time of the conflict those who were in early childhood. While the possibility that growth disturbances in early childhood can affect children s growth is well accepted, there is greater debate about the later-life impacts of disturbances during the in utero period (see Almond and Currie (2011) for a review). 4 There are a number of reasons why in utero exposure may be harmful to child health. These include poorer maternal nutrition due to disruptions in food supply or income shocks, a lack of adequate antenatal care, and the possibility that the conflict reduced the number deliveries in the presence of trained providers. Our paper examines the relative importance of exposure during these two important periods, something that has not been done in previous research on the effects of armed conflicts. A third contribution of this paper is that it measures the welfare impacts for the two sides involved in a conflict, thereby providing a more comprehensive and robust understanding of how such conflicts affect children s well-being. At the conclusion of many armed conflicts that have been studied, there is usually a country or a region that it is less deprived. However, due to data limitations, studies have not analyzed the differential impact on children s health of being on the 4 Research on how environmental factors during the in utero period shape future outcomes range from birth weight to diseases later in life as well education and employment outcomes. Much of this research on the fetal origins hypothesis follows seminal work by Barker (1990). The specific in utero environmental factors that have been considered in research include maternal sickness, maternal stress, and maternal nutrition all of which can be affected to some degree by wars and conflicts. To clearly establish the causal pathway, several studies have compared cohorts that were affected in utero to cohorts that were already born or those that were about to be conceived at the time of the shock. Research by Almond (2006), for example, has shown that in utero exposure to the 1918 Influenza Pandemic resulted in large long-term effects on health, education, and wages. 5

6 winning side of the conflict. Our approach examines whether conflict-affected individuals from Ethiopia, which won the conflict, suffered suffer smaller health consequences than individuals from Eritrea. Finally, this paper also controls for factors such as migration out of conflictaffected regions when estimating the overall impact of the conflict. Since the survey data include information on each household s residential location during the war, we are able to compare the magnitude of possible misclassification errors when an individual s residence at the time of the survey is different than during the war. Controlling for selective migration out of conflictaffected regions during or after a conflict turns out to influence the size of the estimates that are obtained. Our results indicate that the conflict resulted in a significant reduction in the height-forage Z-scores of boys and girls who resided in conflict-affected regions at the time of the conflict. The effect on children s height-for-age Z-scores increased with the number of months of conflict exposure. In Eritrea, each month of exposure (in utero and early childhood) results in a significant reduction in the Z-score of standard deviations while the corresponding effect in Ethiopia is standard deviations. Based on the average number of months that children in the two countries were exposed to the conflict in utero or early childhood, these estimates translate into overall reductions in the Z-score of nearly 0.5. Further evidence of a causal effect of conflict exposure on child health comes from our results based on classifying conflict exposure on the basis of distance from conflict areas. In Ethiopia, a considerably larger country than Eritrea, we find evidence of a declining impact of the conflict as distance from the conflict areas increases. Our results also point to larger health impacts in regions where there was a greater proportion of the population that was internally displaced due to the conflict. Finally, our results suggest that in general, each month of exposure during early childhood has a larger, more 6

7 negative impact on height-for-age Z-scores than each month of exposure in utero. In Eritrea, however, in utero exposure also has a sizable effect on children s height-for-age. The remainder of the paper is organized as follows. Section 2 provides an overview of the history of the Eritrea-Ethiopia conflict and sketches the spatial and temporal event data for the most recent war. Section 3 describes the survey data used in the analysis and explains the key variables. Section 4 describes the empirical identification strategy and Section 5 presents the main results as well as robustness tests. Section 6 concludes. 2. Background on the Eritrean-Ethiopian conflict The conflict between Eritrea and Ethiopia lasted two years beginning in 1998 and stemmed from a border dispute. The two countries had a long history of conflict with each other even before The post-world War II period saw the former Italian colony of Eritrea become a region of Ethiopia, but growing dissatisfaction with the Ethiopian occupation led to a prolonged period of armed struggle by the Eritrean People s Liberation Front (EPLF) against the Ethiopian Marxist government. The conflict against Ethiopia ended in 1991 and coincided with the end of the Ethiopian civil war in which a coalition of rebel groups the Ethiopian People's Revolutionary Democratic Front (EPRDF) overthrew the government and came to power under the leadership of Meles Zenawi. Following a referendum in Eritrea in May 1993, the sovereign nation of Eritrea was formed with the EPLF leader Isaias Afwerki as President (EPLF was later renamed the People's Front for Democracy and Justice). The immediate period following Eritrean independence saw generally friendly relations between Eritrea and Ethiopia, in part because the governments had fought together against the previous Marxist government that formerly controlled Ethiopia. 7

8 At the time of Eritrean independence, both countries claimed sovereignty over three border areas: Badme, Tsorona-Zalambessa, and Bure (see Figure 1 for a regional map of Eritrea and Ethiopia highlighting these three areas). Continued disputes in these three border areas combined with larger conflicts over trade and other economic issues proved to be a major obstacle to maintaining peace between the two countries. 5 In May 1998, fighting broke out between Eritrean and Ethiopian soldiers and security police in the Badme area, which was under Ethiopian control. 6 Within a week, the Ethiopian Parliament declared war on Eritrea. Both countries devoted substantial resources to growing their armies, augmenting their military equipment, and fortifying their borders, which included digging extensive trenches. After the initial period of intense conflict, heavy fighting resumed in February 1999 as Ethiopia succeeded, despite high casualties, in retaking the border town of Badme, but the battles around Tsorona-Zalambessa were not conclusive. Both sides initially rejected efforts by regional groups to mediate an end to the conflict, but eventually a Cessation of Hostilities agreement was brokered on June 18, 2000 and a 25-kilometer-wide demilitarized Temporary Security Zone was established along the 1,000 kilometer Eritrea-Ethiopia border and patrolled by United Nations peacekeeping forces. A final comprehensive peace agreement was signed December 12, The exact timing and location of the conflict plays a key role in our empirical strategy for identifying the effects of the conflict on child health. The conflict intensity varied across regions within Ethiopia and Eritrea. Distance from the conflict areas was an important determinant of conflict intensity. Regions far from the border zones experienced little or no fighting; whereas 5 Eritrea s independence in 1993 meant Ethiopia became a landlocked country, with implications for its trade and economic organization. 6 The Eritrea-Ethiopia Claims Commission (2005) states, The areas initially invaded by Eritrean forces were all either within undisputed Ethiopian territory or within territory that was peacefully administered by Ethiopia and that later would be on the Ethiopian side of the line to which Ethiopian armed forces were obligated to withdraw in 2000 under the Cease-Fire Agreement of June 18, The empirical analysis in this paper treats this as the date the war ended, but our results are consistent if we treat June 2000, the date when the Cessation of Hostilities agreement was brokered, as the time when the war ended. 8

9 the most intense clashes took place in the border regions near Badme, Tsorona-Zalambessa, and Bure. While exact figures of the number of casualties due to the conflict are difficult to ascertain, most estimates of the total number of fatalities, mainly among soldiers, range from 70, ,000 (Human Rights Watch, 2003). Even though most casualties were soldiers, thousands of civilians were displaced and this represented a central mechanism through which conflict may have affected child health. Displaced households suffered large reductions in food production, asset losses, and had limited access to clean water or health infrastructure. By late 1998, estimates suggest approximately 250,000 Eritreans had been internally displaced and another 45,000 Ethiopian citizens of Eritrean origin were deported from Ethiopia (Global IDP Project, 2004a). The Eritrean government and other observers estimate that during the conflict nearly 1.1 million Eritreans were internally displaced, although this number declined substantially by the conflict s end (Global IDP Project, 2004a). The Ethiopian government estimates that by December 1998, 315,000 Ethiopians were internally displaced, with the two regions that border Eritrea (Tigray and Afar) having the greatest number of internally displaced people (IDPs). The United Nations Ethiopia Country Team estimates that by May 2000 the number of IDPs in Ethiopia had risen to 360,000 (Global IDP Project, 2004b). 8 By most accounts, households directly affected by the conflict and those that were internally displaced tended to be located closest to the areas of the clashes. 3. Data 3.1 Demographic and Health Surveys in Eritrea (2002) and Ethiopia (2000 and 2005) 8 This level of conflict-induced displacement is typical, as currently 27.1 million individuals worldwide are IDPs due to conflict. For example, during the last decade in Africa, the number of IDPs due to conflict reached 3.5 million in Angola, 633,000 in Burundi, 200,000 in Central African Republic, 180,000 in Chad, 150,000 in Congo- Brazzaville, 750,000 in Côte d Ivoire, 3 million in Democratic Republic of Congo, 359,000 in Guinea, 600,000 in Kenya, 450,000 in Liberia, 550,000 in Nigeria, 600,000 in Rwanda, 70,000 in Senegal, 1.3 million in Sierra Leone, 1.5 million in Somalia, 6.1 million in Sudan, 1.7 million in Uganda, and 1 million in Zimbabwe (IDMC, 2010). 9

10 Our analyses make extensive use of three different waves of data from the Demographic and Health Surveys (DHS) conducted in Eritrea and Ethiopia. The DHS are nationally representative cross-sectional surveys that gather information on demographic topics such as fertility, child mortality, health service utilization, and nutritional status of mothers and young children. The 2002 Eritrea DHS collected detailed information on the date of birth and height of 5,341 children under five born before, during, or after the conflict with Ethiopia. The 2000 Ethiopia DHS collects similar information for 8,590 children under five, all of whom were either born before or during the conflict with Eritrea. Our analyses rely on DHS data for information on health outcomes (described below) as well as other individual and household characteristics of children including geographical information on their residence at the time of the survey and in some cases during the time of the conflict. To have a control group of children in the conflict regions of Ethiopia who were not exposed to conflict, we use the 2005 Ethiopia DHS that has information for 3,875 children under five. We exclude from the baseline analysis the nine percent of these children born before the conflict ended and use the remaining sample of 3,505 children under 54 months old in To maintain a consistent age range, we also exclude children who were 54 months or older in the 2000 Ethiopia DHS, yielding a final sample of 11,342 Ethiopian children (7,837 from the 2000 DHS and 3,505 from the 2005 DHS) Health outcomes Since children s height (conditional on age and gender) can be sensitive to past growth failures due to chronic malnutrition and illnesses, it is generally accepted as a good way to capture the longer-term health consequences of previous growth disturbances (World Health Organization, 1995). Using the anthropometric information contained in the DHS for children 0-60 months of age, we compute Z-scores for each child s height-for-age, where the Z-score is defined as the 9 Regression results are consistent if all Ethiopian children are included in the subsequent analysis. 10

11 difference between the child s height and the mean height of the same-aged international reference population, divided by the standard deviation of the reference population. The heightfor-age Z-score of children is our primary health outcome of interest. Our analysis also examines other information on health behaviors and health outcomes to better understand mechanisms by which conflicts may influence height-for-age Z-scores. In particular we also examine DHS data on the occurrence of facility deliveries, the use of prenatal care, and self-reported birth size. 3.3 Measures of conflict exposure We construct three measures of a child s exposure to the Eritrea-Ethiopia conflict. The first measure is a continuous measure of the number of months of conflict exposure and is defined at the region-birth cohort level. This allows us to exploit variation across two dimensions: spatial (variation across regions in exposure to the conflict) and temporal (within a given region, the timing of whether a child was born before or during the conflict period). Specifically, we use information on a child s region of residence and date of birth to calculate the number of months the child was exposed to the conflict in utero (defined as 9 months prior to the reported date of birth) and the number of months a child was exposed to the conflict in early childhood. The duration measure is set to zero months if the child resided in a region that was not affected by the conflict. As we discussed in Section 2.2, the fighting was centered on the border regions near the three towns of Badme, Tsorona-Zalambessa, and Bure, so in Eritrea, the conflict regions are defined to include Gash Barka, Debub, and Debubawi Keyih Bahri, while in Ethiopia they are Tigray and Afar. Since conflict-induced displacement was an important mechanism through which the conflict impacted child health, we also incorporate direct measures of the number of internally displaced people in each region to proxy for the conflict s intensity. The IDP data come from the 11

12 United Nations Office for the Coordination of Humanitarian Affairs (UN OCHA) in Eritrea and Ethiopia. All of the IDPs are clustered in the three conflict regions in Eritrea and the two conflict regions in Ethiopia mentioned above (Global IDP Project 2004a, b). Specifically we use the number of IDPs in each region as a proportion of the region s pre-war population as a measure of conflict intensity. Finally, to address potential measurement error that would wrongly misclassify children as conflict-exposed because they live in a region that experienced fighting even if their village was far from the conflict sites, we construct a third measure of conflict exposure using GPS information on the residence of children at the time of the survey. Specifically, we calculate the distance from each survey village to the nearest conflict sites and classify exposure based on different distance bands of kilometers, kilometers, kilometers, and greater than 300 kilometers. We use the distance to the nearest conflict site (even if it crosses region boundaries) when creating these measures. The distance bands are also interacted with the number of months of exposure to create the third measure of conflict exposure. 4. Empirical Identification Strategy Our approach to determining the effect of the Eritrean-Ethiopian conflict on child health relies on an examination of how height-for-age Z-scores of children vary as a function of different durations of exposure to the conflict and variation in the degree to which their regions of residence were affected by the conflict. We begin by estimating the following regression that includes region and birth cohort fixed effects regression: (1) where HAZ ijt is the height-for-age Z-score for child i in region j who was born in period t, j are region fixed effects, t are year of birth cohort fixed effects, Months War Exposure t is a variable 12

13 for the total number of months that the child was exposed to the conflict either in utero or after birth (which is equal to zero for children in regions that were not affected by the conflict), and ijt is a random, idiosyncratic error term. The regression also includes household and individual level controls (X ij ) such as child gender and household head schooling. In addition, to address the potential for differential time trends in height-for-age Z-scores across regions, the regressions include region-specific time trends ( ). The coefficient 1 measures the effect of one month of exposure to the conflict on children s height-for-age Z-scores. In this regression model, it is important to emphasize that identification of the effect of the conflict comes from variation in the duration of exposure (conditional on adjustments for age cohort effects) rather than a simple comparison of conflict and non-conflict regions. We also consider whether the effect of the conflict differs according to how long a child was exposed to it while in utero and in early childhood. We estimate the following regression that builds on equation (1): (2) where Months War Exposure In Utero t represents the number of months a child was exposed to the conflict in utero and Months War Exposure After Birth t represents the number of months exposed after birth. Since residence in a conflict region alone may not imply that a child was actually exposed to conflict, we also consider an alternative definition of conflict exposure. Our alternative measure of conflict intensity (Conflict Intensity) indicates for each region the number of IDPs (divided by 10) as a proportion of the region s pre-war population. This allows us to 13

14 better identify the conflict s impact, as we compare regions with many IDPs to regions with few IDPs. We estimate the following regression incorporating the IDP data: (3) In our analyses, we also estimate modified versions of equation (3) in which the effect of the conflict varies based on the number of months of exposure in utero and in early childhood. In defining conflict exposure based on living in one of the three regions in Eritrea or two regions in Ethiopia where fighting took place or even on the basis of the number of IDPs in each region there is a potential for wrongly classifying villages far from the conflict sites as conflict exposed (and vice versa in the case of villages in non-conflict regions that are nonetheless near to conflict sites). This can result in biased estimates as some regions extend many kilometers from the conflict sites (see Afar in Ethiopia and Debubawi Keyih Bahri in Eritrea). Likewise, we might be excluding households close to conflict sites that may have been affected by conflict but were actually in a non-conflict defined region (see Figure 1, Semenawi Keyih Bahri in Eritrea). To more accurately measure a child s conflict exposure, our empirical strategy takes advantage of information on the distance of each survey village to the three main conflict sites. We test for the effect of each additional month of exposure to the conflict in three different distance categories (with a fourth category, greater than 300 kilometers, serving as the reference group): 14

15 (4) 0, , ,300 0, , ,300 Our empirical strategy includes several extensions to the regressions above as well as robustness checks. First, since the main regression results are generally based on a child s residence at the time of the survey, there is a possibility of bias due to children having resided elsewhere during the time of the conflict. We take two approaches to addressing this issue. First, all children whose households reported having lived elsewhere prior to the survey are classified as residing in a war region. In addition, in the case of Eritrea, we use actual information contained in the survey on the region of residence at the time of the conflict and examine what happens to the results when this re-classification is done. The empirical analysis includes an assessment of the effects of the conflict on boys and girls separately. Importantly we also pool the datasets from Eritrea and Ethiopia and estimate regression models in which we test whether the effects are significant. 5. Empirical Results 5.1 Difference-in-Differences Estimation (War region and Geospatial location) Table 1 summarizes several of the variables used in the analyses including the height-for-age Z- scores of children and the key measures of conflict exposure. In both Ethiopia and Eritrea, children residing near to conflict sites (<100 kms) have lower Z-scores than those residing farther away (>300 kms), although the difference is statistically significant in Ethiopia only. Children residing near the conflict sites are also worse of in other ways, as indicated by the lower 15

16 schooling attainment of household heads. In general, the regions that were affected by the conflict tended to be worse off even prior to the conflict, which points to the importance of not simply comparing the height-for-age Z-scores of children in affected and unaffected regions when seeking to determine the impact of the conflict. Panel C of Table 1 shows that on average, children between 0-60 months of age who resided less than 100 kms from conflict sites in Ethiopia were exposed to a total of 15.5 months (1.8 months in utero and 13.7 months after birth). In Eritrea, due to the timing of the DHS, the durations are somewhat longer (an average of 18.8 months of total exposure in areas near conflict sites). The results from our first attempt to examine the impact of the conflict after including region fixed effects and controlling properly for age, individual and household characteristics are reported in Table 2. We present the results of estimating equations (1) and (2) for Ethiopia and Eritrea. In columns 1 and 2, each child's residence is classified according to the region of residence at the time of the DHS. In columns 3 and 4, the child's residence is classified as the potential region of residence at the time of the conflict, whereby any children who moved during the conflict are assigned to the conflict regions. In columns 5 and 6, the child's residence is classified as their known region of residence at the start of the conflict to accurately capture conflict exposure. Column 1 of Table 2 shows that for children exposed to the conflict either in utero or in early childhood each additional month of exposure results in a significant reduction of height-for-age Z-scores (by in Ethiopia and in Eritrea). When children s exposure is classified on the basis of their potential residence at the time of the conflict (column 3), the effects on height-for-age Z-scores remain similar ( in Ethiopia and in Eritrea) and continue to be statistically significant. This is the first suggestion that our results are robust to the adjustment for factors such as migration since the time of the conflict. In Eritrea, the results 16

17 based on classifying residence as the region in which the child was reported to have resided at the time of the conflict (column 5) show that height-for-age Z-scores of children exposed to the conflict are reduced by for each month of exposure. This too is a similar effect size to what is obtained when residence is classified as the region of residence at the time of the survey. The other important result in Table 2 pertains to the distinct effect of exposure to the conflict in utero as opposed to early childhood. In both Ethiopia and Eritrea, we find that the effects of exposure after birth are larger (a reduction in the Z-score of for each month of exposure in Ethiopia and for each month of exposure in Eritrea). In Ethiopia (column 2 of Panel A), there is no significant effect on height-for-age Z-scores due to in utero exposure to the conflict. In Eritrea, however, in utero exposure does result in a significant reduction of heightfor-age Z-scores, by for each month of exposure. In summary, Table 2 provides evidence that the exposure to the conflict in early childhood is associated with larger effects on height-forage Z scores in subsequent years. It is only in Eritrea that in utero exposure also reduces heightfor-age Z-scores significantly. To interpret these results more easily, it is instructive to consider as an example the effects of children exposed to the conflict for 9 months in utero and 9 months in early childhood: a child exposed to the conflict in Eritrea for 9 months in utero experienced a reduction of in the height-for-age Z-score whereas a child exposed to the conflict for 9 months in early childhood experienced a reduction of the Z-score. In Table 3 we report the results from classifying conflict exposure as a function of the intensity of the conflict in the child s region of residence (equation 3). The results are largely consistent with those reported in Table 2. It is in the regions most affected by the conflict as measured by the proportion of IDPs that children experience the largest setbacks to their growth. Each month of exposure reduces the height-for-age-z-scores of children, and this impact 17

18 increases as a function of conflict intensity in the region. The main results, based on residence at the time of the survey, indicate that an increase of 1 percent in the IDP proportion of a region s population reduces the height-for-age-z-scores by and for each month of total exposure (in utero and in early childhood) in Ethiopia and Eritrea, respectively. These effects are similar when residence is based on classifying all children who migrated as having been in conflict regions. When the effects of the conflict are allowed to differ as a function of exposure in utero and in early childhood (column 2), we find significant effects of exposure in early childhood in both countries ( per month of exposure in early childhood in Ethiopia and in Eritrea). On the other hand, we find significant effects of exposure in utero in Eritrea only (-0.013). Tables 4a and 4b contain the results of regressions that use geospatial data on the proximity of each household to the conflict sites (equation 4). The combined effect of exposure to the conflict in utero and in early childhood can be seen in Table 4a, while the separate effects of exposure can be seen in Table 4b. Each month of combined exposure significantly reduces the height-for-age-z-scores of children near to conflict sites by in Ethiopia and by in Eritrea (Table 4a, columns 1 & 2). The effects based on classifying residence differently for children who migrated since the conflict are similar (columns 3 & 4). Importantly, Table 4a shows a modest decreasing gradient in the effect of conflict as a function of distance from the main conflict sites. The gradient is most noticeable in Ethiopia, a much larger country than Eritrea and one in which distance from the conflict sites is likely to have been more relevant for whether or not economic and nutritional conditions and consequently children s health and nutritional status would be affected. As the p-values at the bottom of Table 4a show, in Ethiopia the effect of being near to conflict sites is significantly larger than the effect of being an 18

19 intermediate distance or far away from the conflict sites. This is not the case in Eritrea, as the effect of conflict does not appear to decline as much with distance from conflict sites. These results showcase the importance of geographical data for further illuminating the effects of conflicts and even accurately measuring the size of these effects. The results in Table 4b indicate that in the areas closest to conflict sites, exposure to the conflict in utero and in early childhood both result in negative impacts on height-for-age Z- scores. This is true in both Ethiopia and Eritrea. In areas farther from conflict sites, however, in utero exposure has a negative impact in Eritrea but not in Ethiopia. While the evidence so far suggests that there are large and significant reductions in height-for-age Z-scores as a result of conflict exposure (particularly in early childhood), an important question from the standpoint of public policy and intra-household resource allocation is whether boys and girls are affected in different ways. With this in mind, the results in Table 5 are noteworthy because they show remarkably similar negative impacts of conflict exposure in both countries for boys as well as girls. The magnitude of the impact is slightly larger for girls in Ethiopia and slightly larger for boys in Eritrea but in neither case are the differences statistically significant. As in Table 4a, the distance gradient is noticeable in Ethiopia but not in Eritrea. Tables 6 and 7 pool the data from Eritrea and Ethiopia and compare the effects of the conflict in the two countries. To test whether the impact of the conflict in Eritrea differed from the impact in Ethiopia, we include an interaction term of the main variables with an indicator variable for Eritrea. In column 1 of Table 6, the results indicate that children from conflict regions in Eritrea have significantly lower height-for-age Z-scores than children from conflict regions in Ethiopia (an additional impact of for each additional month of combined 19

20 exposure in utero and in early childhood). This result holds even when we classify residency based on potential residency during the conflict (column 3 of Table 6). When we examine the separate effect of exposure in utero vs. in early childhood (column 2 of Table 6), we again find that children in Eritrea are more affected by the conflict. They have significantly lower heightfor-age Z-scores as a result of in utero exposure (0.043 lower for each additional month of conflict exposure) and due to early childhood exposure (0.028 lower for each additional month of conflict exposure). Table 7a shows that for each category of distance from conflict sites, children in Eritrea have more negative effects on height-for-age Z-scores than children in Ethiopia. In particular, when we classify residency based on residency at the time of the survey (column 1) the effect of each month of conflict exposure (in utero or in early childhood) on the height-for-age Z-scores of children closest to the conflict sites in Eritrea is larger than in Ethiopia. Among children at intermediate distances from conflict sites or farther away from them, the effects on children in Eritrea are also significantly more negative by and 0.041, respectively. In results not reported, we find that the effects of in utero exposure are significantly more negative in Eritrea than in Ethiopia, but the effects of early childhood exposure are similar in the two countries. Finally, in Table 7b, we split the results obtained in Table 7a by in utero exposure and in early childhood exposure. Again, we find that children from Eritrea have lower height-for-age Z- scores than children from Ethiopia for each of the proximity categories. In particular, in column 1, coding geographical location based on location at the time of the survey, the results suggest that children near to conflict sites in Eritrea have lower height-for-age Z-scores than children close to the conflict sites in Ethiopia by for each month of exposure in utero and by for each month of exposure in early childhood. In column 2 the results are similar when we code 20

21 the geographical location based on potential location at the time of the conflict. Children near to conflict sites in Eritrea have lower height-for-age Z-scores than children near to conflict sites in Ethiopia by for each month of exposure in utero and by for each month of exposure in early childhood. 5.2 Robustness Checks To determine whether the effects of the conflict are robust to a range of possibilities that would generate biased results, in Appendix Tables 1-4 we present results based on alternative analyses that were performed. In Appendix Table 1, we report the results of including mother fixed effects in the regression model and thereby having better controls for household characteristics than is possible with other models that include only region fixed effects. In the smaller sample of households that have multiple children under five years of age, the results indicate significant negative effects for each month of conflict exposure in war regions for both Ethiopia and Eritrea. In contrast to the results in Table 2, the magnitude of the effects are larger in Ethiopia than in Eritrea, although it is worth noting that the sample sizes with mother fixed decrease much more in Eritrea than in Ethiopia. In Appendix Table 2, we define households proximity to the conflict sites on the basis of the relative distance within the sample rather than the absolute distance. The results indicate that relative proximity or closeness does result in larger reductions in the height-for-age Z-score. This effect of relative proximity is found for in utero exposure as well as early childhood exposure. Another concern regarding the validity of the main estimates relates to the possibility that fertility and mortality patterns may be different in regions that were exposed to the conflict with varying severity. We examine fertility patterns by comparing characteristics of women based on whether they had children during the conflict and based on their region of residence. We find 21

22 relatively little evidence that characteristics of women who had children were influenced by the conflict. In Ethiopia, there is some evidence that taller women were more likely to have children in war regions during the conflict than at other times (an effect that would only bias our main results downward rather than upward). In Eritrea, there is some evidence of an effect on the age of women who had children, but this effect is likely to be too minor to influence the main results. Lastly, our analysis likely underestimates the shock s true health impact for two reasons. First, a child s age could be mismeasured, and if this occurred, it would likely mean our estimates are lower bounds of the true impact, as parents would probably underreport the age of short children making their malnutrition seem less severe than it is. The chance of this is reduced since the household roster collects the exact birth date of all the household s children under five and misreporting on one child would be more difficult as it would influence the birth dates of the household s other children. Second, child mortality might be higher in war-exposed households. Unfortunately, we do not have health data on children who died prior to the survey, but these deceased children were likely the weakest and smallest, which means we are underestimating the total war impact. Therefore, the reported effects should be interpreted as the war s impact on child health, conditional on the child surviving to be recorded in the survey 5.3 Conflict Impact Mechanisms Understanding the specific reasons for the negative effects of the conflict on child health is essential for interpreting the main results of this paper and developing better policies to prevent such negative effects in the future. There are several reasons why child health may have been affected by the conflict. While the available data lack detailed information on some of these reasons for example, it is not possible with the DHS data to determine what effect the conflict had on agricultural production, incomes, and various environmental factors in Appendix Table 22

23 5, we assess the impacts of the conflict on relevant behaviors and outcomes that are measured in the DHS. The possibility that access to health services may have been reduced during the conflict may help explain the negative impacts on children s height-for-age Z-scores. We examined the effects of the conflict on the likelihood that children were born at a formal health institution where the risk of infections may be lower and the availability of personnel capable of managing complications is higher. The results in column 1 indicate that the conflict had no effect on the likelihood of delivery taking place in a health facility in Ethiopia. However, the results show that the conflict reduced the probability of delivery in a health facility by 13.9% in Eritrea. The size of the effect may explain the difference between the impacts on health in each country. A second possible way in which the conflict may have affected health outcomes is by reducing the likelihood that pregnant women sought adequate antenatal care. Appendix Table 5 presents some evidence that the likelihood of antenatal visits was significantly reduced in Ethiopia for children who were more exposed to the conflict, the effect tends to be larger. In Eritrea, however, we find no evidence of an effect on antenatal care seeking behavior. This is puzzling in light of the fact that the estimated impacts of the conflict were generally larger in Eritrea. When we examine a health indicator that could have been affected by in utero exposure to the conflict birth weight -- we find no significant effect in either country. This is a subjective measure based on women s perceptions (at the time of the survey) about the birth size of the child in comparison with the birth size of other children, and hence it has natural limitations. We also use the DHS data to examine whether the conflict had a negative impact. A third possible mechanism is the theft of assets, including livestock. Since it takes time to reverse such a loss, poverty generated by asset theft could affect all children in the household, regardless of 23

24 whether they are born during or after the conflict. Our results suggest this mechanism might not have been salient, as there is no evidence of impact on wealth for the affected children (column 3). As mentioned in Akresh et al (2012), while displacement of households within regions might be a key mechanism by which the conflict impacted children s health status, the absence of detailed survey information on the movement of households limits our ability to examine this issue thoroughly. Many families were internally displaced in both countries due to the conflict. Because of this, households could have been directly worse off through loss of harvests and assets, disruption of businesses, and reduced access to medicines and clean water. Consistent with our results, a child exposed to displacement would be worse-off compared to a non-exposed child, and the impact should be larger the longer the child is exposed to these events. There were also effects on children who were not displaced by the conflict but resided in areas with many displaced individuals (Baez, 2011; Maystadt and Verwimp, 2009). The similarity between our main results and the results when we calculate the bounds for each effect taking into account the migration status of each kid, suggests that displacement due to the conflict may indeed have affected both displaced and non-displaced children. 6. Conclusion This paper quantifies the health consequences of the Eritrea-Ethiopia conflict using a variety of different methods and conflict exposure measures. The findings indicate relatively large negative effects of the conflict on the subsequent height-for-age Z-scores of children exposed to the conflict. Since the height-for-age Z score is considered an important health indicator of children and is associated with health and employment outcomes later in life, the results in this paper offer some insight on what effect the conflict may have in the long-term. More generally, since 24

25 children and households in a number of other resource-limited settings have been exposed to similar conflicts that feature internal displacement and disruptions of food supply, this paper can has relevance for the larger literature on the short- and long-term effects of armed conflicts. Identifying the causal effects of conflict exposure on subsequent health outcomes can be challenging because the exact nature of how each child was affected by the conflict is difficult to measure, and those exposed the most may have been worse off at the outset of the conflict. The results in this paper are based on attempts to overcome these problems by comparing children exposed to the conflict for varying durations and at different stages (in utero and after of birth), and by classifying children s conflict exposure on the basis of the geographical coordinates of their residence as well as characteristics of their region. Using these various methods, we generally find that the Eritrea-Ethiopia conflict had similar negative effects on children in both countries who were exposed after birth. In Eritrea, on the other hand, we also find significantly reduced height-for-age Z-scores among children who were exposed in utero. The sizes of the impacts are noteworthy, as the typical amount of exposure for children in conflict-affected regions translates into a reduction of about 0.5 in the height-for-age Z-score following the conflict. However, this approach includes a misclassification bias due to the inclusion of children that were in the affected regions but far from the conflict sites. Including them in our exposed group produces an underestimation of the true effect of the war. In particular, defining the exposure areas in terms of distance instead of using the politically defined subnational units, we find for a child exposed to the war for the average duration, this exposure reduces a child s height by approximately 1 in Ethiopia and 2 inches in Eritrea. Understanding the detailed mechanisms by which conflicts affect the health of children remains a challenge that future research will need to address. By examining several health 25

26 behaviors during the in utero period and early childhood months, we were able to rule out the possibility that disruptions in access to antenatal care drove the effects that we found in Eritrea and Ethiopia. While birth weight is poorly measured in the data that we use and is not a comprehensive measure of fetal health, it is notable that this indicator was not affected significantly. The possibility remains that the impacts we find are therefore driven by deteriorations in food supply or access to other types of primary healthcare services, as well as outbreaks of disease in camps for internally displaced persons. Future research that includes richer information will be needed to identify such mechanisms. From the standpoint of policy, these results suggest that households may not have adequate coping mechanisms for conflicts that disrupt economic conditions and lead to displacement, even if the number of civilian casualties remains limited. The results also reinforce the importance of considering medium- and longer-term health consequences when deploying post-conflict services. 26

27 References Akbulut-Yuksel, Mevlude Children of War: The Long-Run Effects of Large-Scale Physical Destruction and Warfare on Children. IZA Discussion Paper Akresh, Richard Flexibility of Household Structure: Child Fostering Decisions in Burkina Faso. Journal of Human Resources, 44(4): Akresh, Richard, and Damien de Walque Armed Conflict and Schooling: Evidence from the 1994 Rwandan Genocide. IZA Discussion Paper Akresh, Richard, Leonardo Lucchetti, and Harsha Thirumurthy Wars and Child Health: Evidence from the Eritrean-Ethiopian Conflict. Journal of Development Economics, 99(2): Akresh, Richard, Philip Verwimp, and Tom Bundervoet Civil War, Crop Failure, and Child Stunting in Rwanda. Economic Development and Cultural Change, 59(4): Akresh, Richard, Sonia Bhalotra, Marinella Leone, and Una Osili War and Stature: Growing Up During the Nigerian Civil War. American Economic Review, 102(3): Alderman, Harold, John Hoddinott, and Bill Kinsey Long Term Consequences of Early Childhood Malnutrition. Oxford Economic Papers, 58(3): Almond, Douglas and Janet Currie Killing Me Softly: The Fetal Origins Hypothesis. Journal of Economic Literature, 25(3): Angrist, Joshua Lifetime Earnings and the Vietnam Era Draft Lottery: Evidence from Social Security Administrative Records. American Economic Review, 80(3): Barker, David J.P Mothers, Babies, and Health in Later Life. Edinburgh, United Kingdom: Churchill Livingstone. Baum, Christopher, Mark Schaffer, and Steven Stillman Enhanced Routines for Instrumental Variables/GMM Estimation and Testing. Boston College Department of Economics, Working Paper No 667. Bellows, John, and Miguel, Edward War and Local Collective Action in Sierra Leone. Journal of Public Economics, 93(11-12): Blattman, Christopher, and Jeannie Annan The Consequences of Child Soldiering. Review of Economics and Statistics, 92(4):

28 Brakman, Steven, Harry Garretsen, and Marc Schramm The Strategic Bombing of German Cities During World War II and Its Impact on City Growth. Journal of Economic Geography, 4(2): Bundervoet, Tom, Philip Verwimp, and Richard Akresh Health and Civil War in Rural Burundi. Journal of Human Resources, 44(2): Caruso, German Long-run Impacts and Intergenerational Transmission of Shocks in Early Life: Evidence from Natural Disasters in Latin America. University of Illinois, working paper. Collier, Paul, and Anke Hoeffler On the Economic Causes of Civil War. Oxford Economic Papers, 50(4): Davis, Donald, and David Weinstein Bones, Bombs, and Break Points: The Geography of Economic Activity. The American Economic Review, 92(5): Dercon, Stephan, and Pramilla Krishnan In Sickness and in Health: Risk Sharing Within Households in Rural Ethiopia. Journal of Political Economy, 108(4): Do, Quy-Toan, and Lakshmi Iyer Geography, Poverty and Conflict in Nepal. Journal of Peace Research, 47(6): Eritrea Ethiopia Claims Commission Partial Award Jus Ad Bellum Ethiopia s Claims 1-8 between The Federal Democratic Republic of Ethiopia and The State of Eritrea. The Hague, December 19, Global IDP Project. 2004a. Profile of Internal Displacement: Eritrea, Compilation of the Information Available in the Global IDP Database of the Norwegian Refugee Council. Global IDP Project. 2004b. Profile of Internal Displacement: Ethiopia, Compilation of the Information Available in the Global IDP Database of the Norwegian Refugee Council. GlobalSecurity.org (accessed December 1, 2010). Guidolin, Massimo, and Eliana La Ferrara Diamonds are Forever, Wars Are Not: Is Conflict Bad for Private Firms? American Economic Review, 97(5): Humphreys, Macartan, and Jeremy Weinstein Demobilization and Reintegration. Journal of Conflict Resolution, 51(4): Human Rights Watch The Horn of Africa War: Mass Expulsions and the Nationality Issue. Human Rights Watch, 15(3A). 28

29 Ichino, Andrea, and Rudolf Winter-Ebmer The Long-Run Education Cost of World War II. Journal of Labor Economics, 22(1): Imbens, Guido, and Wilbert van der Klaauw Evaluating the Cost of Conscription in The Netherlands. Journal of Business and Economic Statistics 13(2): Internal Displacement Monitoring Center (IDMC) Internal Displacement: Global Overview of Trends and Developments in Kleibergen, Frank, and Richard Paap Generalized Reduced Rank Tests Using the Singular Value Decomposition. Journal of Econometrics, 127(1): Krishnan, Pramila, Tesfaye Selassie, and Stefan Dercon The Urban Labor Market During Structural Adjustment: Ethiopia Report No. WPS Center for the Study of African Economies, Oxford University. Maccini, Sharon, and Dean Yang Under the Weather: Health, Schooling, and Socioeconomic Consequences of Early-Life Rainfall. American Economic Review, 99(3): Maluccio, John, John Hoddinott, Jere Behrman, Reynaldo Martorell, Agnes Quisumbing, and Aryeh Stein The Impact of Improving Nutrition During Early Childhood on Education among Guatemalan Adults. Economic Journal, 119(537): Martorell, Reynaldo, and Jean-Pierre Habicht Growth in Early Childhood in Developing Countries. In Human Growth: A Comprehensive Treatise. F. Falkner and J. Tanner, Editors. Plenum Press. New York. Miguel, Edward, Shanker Satyanath, and Ernest Sergenti Economic Shocks and Civil Conflict: An Instrumental Variables Approach. Journal of Political Economy, 114(4): Miguel, Edward, and Gerard Roland The Long Run Impact of Bombing Vietnam. Journal of Development Economics, 96(1): Moulton, Brent Random Group Effects and the Precision of Regression Estimates. Journal of Econometrics, 32(3): Raleigh, Clionadh, Andrew Linke, Havard Hegre, and Joakim Karlsen Introducing ACLED: An Armed Conflict Location and Event Dataset, Journal of Peace Research, 47(5): Rose, Elaina Consumption Smoothing and Excess Female Mortality in Rural India. Review of Economics and Statistics, 81(1):

30 Shemyakina, Olga The Effect of Armed Conflict on Accumulation of Schooling: Results from Tajikistan. Journal of Development Economics, 95(2): Stein, Zena, Mervyn Susser, Gerhart Saenger, and Francis Marolla Famine and Human Development: The Dutch Hunger Winter of New York: Oxford Press. Strauss, John, and Duncan Thomas Health Over the Life Course. In Handbook of Development Economics, vol. 4, ed. Paul Schultz and John Strauss, Amsterdam: North-Holland. Swee, Eik On War and Schooling Attainment: The Case of Bosnia and Herzegovina. Households in Conflict Network Working Paper 57. Thomas, Duncan, Victor Lavy, and John Strauss Public Policy and Anthropometric Outcomes in the Cote D Ivoire. Journal of Public Economics, 61(2): United States Central Intelligence Agency CIA World Factbook Downloaded from on November 27, World Health Organization Physical Status: The Use and Interpretation of Anthropometry. Report of a WHO Expert Committee. World Health Organization Technical Report Series 854:

31 Figure 1: Eritrea and Ethiopia Regional Map Indicating Conflict Sites Notes: The main fighting between Eritrea and Ethiopia occurred around the areas of Badme, Tsorona-Zalambessa, and Bure, which are noted on the map. Map source: Constructed by Rafael Garduño-Rivera in ArcGIS. 31