The Economic and Social Costs of Aruned Conflict in El Salvador January 2003 Number 8 El Salvador's civil war (1979-91) had devastating economic and social costs. This note estimates the ground lost in terms of growth forgone, higher poverty and worsened social indicators. Had the conflict been avoided, income per capita could be almost double in 2000, the poverty rate lower by 15 percentage points and Millenium Development Goals social indicators substantially better. El Salvador's experience shows that conflicts are not only extremely costly in economic and social terms, but that recovery takes an inordinate amount of time, even with very good post-conflict policies and reforms. Background With an area of 21,000 kM2, a population of infrastructure was estimated to have a replacement approximately 6.4 million and per capita GDP of value of US$1.6 billion, the significance of which US$4,500 (PPP) in 2000, El Salvador is the smallest becomes apparent when compared with a GDP of country of Central America and the most densely just over US$5 billion in 1990. The armed conflict, populated in the Western Hemisphere. Gross however, also had enormous indirect costs (both inequalities in wealth and income built up since the economic and social) that go beyond these early part of the 20th century, and historical estimates of direct damages. In this note we disaffection with land inequality severely attempt to provide estimates of these economic and exacerbated social tensions during the 1970s. The social costs. country's political system was ill equipped to deal with these tensions. First, one might consider the dramatic cost in terms of GDP losses. Over the period 1979-1991 (the A series of confrontations between peasant and start and end of the conflict according to the Center organized labor groups and the government led to for Defense Information (CDI) which records and increasingly repressive government responses over dates worldwide armed conflicts involving at least the decade, gradually closing all legal avenues for 1,000 casualties in a single year) real per capita peaceful opposition. In the late 1970s groups of GDP fell by an impressive 75 percent. Moreover, students, peasants and trade unionists opted out of despite a strong economic recovery in the aftermath the political and electoral system founding a number of the signing of the Peace Accords, El Salvador's of guerrilla organizations. Army and paramilitary current real per capita income is still about 20 death squads embarked on a counter-insurgency, percent lower than it was in 1978. This is in sharp which in turn led to the expansion of the guerrilla contrast with the experience of the Latin America movement and an increase in political violence. The and Caribbean (LAC) Region as a whole, which violence reached a dramatic apex in March 1980 recorded an increase of almost 30 percent in real with the murder of the archbishop of San Salvador, per capita income over 1979-91, and about 60 Oscar Romero by a paramilitary squad. Although percent for 1978-00. the guerilla movement, unified under the Frente Farabundo Marti para la Liberaci6n Nacional Second, as a result of lower economic growth, per (FMLN), failed to spark a national rebellion, it capita output is also lower, which in turn should effectively carried out a low-intensity guerilla war result in higher poverty levels and a worsening of establishing strong areas of influence in the north social indicators. To the best of our knowledge, and east of the country. there are no estimates of the social costs of the conflict, in terms of poverty levels and key social In 1989, the new administration of President indicators. This note builds on results by Wodon et Cristiani came to office determined to bring an end al.I and Lopez2 to assess the impact of El to the conflict. Several key agreements in 1991 culminated in the Peace Accords signed in Mexico on January 16, 1992. The Peace Accords brought Wodon Q., R. Castro-Fernandez, G. Lopez-Acevedo, c. Siaens, an end to a conflict that had no winner but many C. Sobrado, and J.P. Tre (2001). Poverty in Latin America: trends losers. The civil war cost some 75,000 lives, (1986-1998) and Determinants, LCR, The World Bank, displaced 1 million people (about 20 percent of the !Vashington, D C. Salvadoran population in the early 1990s), and ' Lopez. H. 'The Cost of Armed Conflict in Central America', al a ra p plto i th eal 99 s a LCR, The World Bank, Washington, D.C Salvador's conflict on a set of indicators that capture one percent increase in GDP would reduce infant the essence of the MDGs. mortality by 0.37 percent. But once the country reaches per capita GDP of US$5,000 the impact of The impact of growth on non-monetary an additional one percent increase in GDP would indicators reduce infant mortality by 0.35. The importance of growth for poverty reduction is Lost growth: intervention analysis well documented in the economic literature. GDP growth seems to have little effect on income We noted above the dramatic fall of per capita GDP distribution and therefore a one percent increase in during the conflict years, and in this regard it seems aggregate growth tends to increase proportionally reasonable to assume a significant impact of the war the mean income of any given quintile. But beyond on GDP. The question we address now refers to the the money metric approach to poverty magnitude and time profile of such impact. To measurement, there is also empirical evidence estimate the impact we rely on time series pointing toward the positive impact of growth on intervention analysis techniques. Briefly, these non-monetary indicators of well being. For example, techniques allow the decomposition of GDP growth Wodon et al. (2001) present cross country into a "regulare component that would capture the elasticities for a set of social indicators to growth. dynamics of GDP in the absence of extreme shocks, These social indicators would capture the essence and a second component that would capture of the MDGs and include infant mortality, under-5 aberrant behaviors caused by extraordinary events mortality, and child malnutrition in the health sector, such as an armed conflict.5 Further we allow these net primary enrolment ratio, net secondary aberrant observations or outliers to affect the enrolment ratio, gross tertiary enrolment and adult "regular" component of growth according to three illiteracy in the education sector and telephone line4 different time profiles: (i) additive, which would density in the infrastructure sector. In the capture a pulse or a one single shot to the growth computation of these elasticities, Wodon et al. make rate in a year; (ii) level shift, which would represent use of an econometric specification that takes into an outlier with permanent significant effects on the account the level of urbanization, time, and the growth rate; and (iii) transitory shift, which would country's level of development (as reflected by the capture disturbances with long-lasting but decaying level of GDP per capita in US$PPP). effects. Table 1: Elasticities of Social Indicators to Growth We perform the empirical procedure both on per capita GDP growth and on GDP growth. The results Per Capita GDP (US$PPP) using GDP growth are then converted into per capita 2,500-5,000 5,000-10,000 GDP assuming no changes to the population profile. For each dependent variable we construct the virtual Infant mortality -0.37 -0.35 GDP series under three scenarios. In the first, we do Under-5 mortality -0.47 -0.37 not introduce any control and hence the statistical Child malnutrition N.S. -1.1 problem is reduced to an outlier search and Net primary enrolment 0.02 0.04 correction procedure. The second scenario controls for the evolution of the global economy (as captured Net secondary enrolment 0.32 0.23 by the median growth rate in the LAC Region). The Illiteracy rate -0.05 -0.11 third and final scenario also controls for external Gross tertiary enrolment 0.23 0.69 factors that are country specific. We do this by Telephones 0.90 0.52 introducing, in addition to the median growth rate in NS = Not significantly different from zero. the LAC Region, the terms of trade of goods and Source Wodon et al (2001) non-factor services as explanatory variables. We therefore compute virtual GDP series under six different scenarios, something that in principle Table 1 reports the estimated elasticities for two should ease concerns regarding the sensitivity of the segments of per capita GDP (US$PPP). They results. suggest that for a country such as El Salvador (with real per capita GDP of US$4,500 at PPP prices) a The implementation of this procedure for El Salvador would identify in every case two transitory shifts 3 corresponding to 1979 and 1980 (the first and This result has been recently stressed by Dollar D and A. Kraay second years of the conflict). We estimate that in (2000), 'Growth is Good for the Poor", DRG, The World Bank, 1979, when per capita GDP declined by 6 percent, Washington, oc. C growth could have been as much as 2.5 percent. 4Since the cross-country correlation between the number of telephone lines, and electricity generated and paved roads is 0.8 and 0.7 respectively, this variable can be considered as a good proxy for infrastructure in general. See Lopez (2001) for technical details 2 Table 2: El Salvador-The Social Cost of the Armed Conflict VIRTUAL ACTUAL Per Capita GDP GDP RAW' LAC2 TOT' RAW' LAC2 TOT3 Per Capita GDP (US$PPP) 4,500 8,413 8,451 8,387 9,045 7,932 8,705 Poverty Poverty Head Count (%) 41 25 25 25 23 24 23 Health Infant Mortality 29 23 23 23 23 24 23 Under-5 Mortality 35 27 27 28 27 28 27 Child Malnutrition 12 7 6 6 6 7 6 Education Net Primary Enrolment 82 84 84 84 84 84 84 Net Secondary Enrolment 35 41 41 40 41 41 40 Illiteracy Rate 18 17 17 17 17 17 17 Gross Tertiary Enrolment 18 26 28 28 29 29 26 Infrastructure Telephone Lines 10 14 14 14 14 14 13 1. No controls; 2. LAC growth rate as control; 3. LAC growth and terms of trade (TOT) as control. For 1980, when per capita GDP declined by 13 What would have been the impact on poverty? percent, we find that GDP growth could have been These figures would suggest that in the absence of above 3 percent. This correction for 1980 (16.3 the armed conflict, the poverty rate could be lower percentage points) could be split into the effect by about 15 percentage points and hover around 25 being caused by the 1979 transitory shift (estimated percent, or close to poverty rates in Costa Rica-a at -5.9 percent) and by the new transitory shift shock neighboring country that did not experience conflict (estimated at about -10 percent). The half-life of the and that has been able to maintain remarkable transitory shifts is estimated at about two years (i.e., progress in fighting poverty and improving social every two years the impact of the shocks on the conditions. regular component of growth is halved). The virtual GDP series would now be computed by assuming B3eyond its devastating toll on per capita GDP and the corrected GDP growth rates. headcount poverty, we also find the armed conflict to have had significant negative impact on social Table 2 reports actual and virtual GDP (US$PPP) outcomes and especially on health indicators. We together with actual and virtual social indicators for estimate the armed conflict to account for about 6 El Salvador in 2000 for each of the scenarios percentage points of the observed infant mortality described above. The column under the heading rate (29 per thousand) and for about 8 percentage RAW reports results for the model without controls. points of the observed under-5 mortality rate (35 Under heading LAC it is possible to find the results percent). A more significant effect is found on child for the model that controls for the evolution of the malnutrition where about half of the malnutrition rate global economy. Finally, under the heading TOT we can be explained by the conflict. We estimate that in report the results that also include the terms of trade the absence of the conflict the malnutrition rate as explanatory variable. could be around 6 percent. Table 2 indicates that depending on the statistical Regarding the education sector, our estimates assumptions virtual per capita GDP would vary from suggest that in the absence of the conflict all the a minimum of US$7,932 to a maximum of US$9,045. indicators would show an improvement, but this That is, our lowest estimate for the virtual per capita would be especially the case for secondary and GDP would suggest an income level that is about 75 tertiary enrolment rates. This may be explained by percent larger than the actual value in 2000. Our the different role that economic considerations take highest estimate would suggest income levels that for dropping out of school at different schooling double those observed in 2000. Thus, El Salvador's levels (less important for primary education). income per capita could have been between 75 to 100 percent higher in 2000 had the conflict not Finally, on infrastructure we also find that the conflict occurred. may have had a significant negative effect. Our estimates suggest that in the absence of the armed 3 conflict telecommunications coverage could be opportunities forgone - opportunities to increase about 40 percent higher than actual. As noted in output, to reduce the number of households living in footnote 2, we could expect similar results for other poverty, and to improve social welfare. Even with infrastructure indicators such as electricity and very good policies it takes an inordinately long time paved roads. to recover lost ground. This underscores the importance of making every effort to prevent armed Conclusions conflicts from erupting in the first place if the Bank and the international community are going to This note has presented estimates of the social succeed in the fight against poverty. costs of El Salvador's armed conflict (1979-1991). To this end we have constructed a virtual per capita GDP series that simulates the level of income that El Salvador might have enjoyed should armed conflict not have taken place. GDP growth losses are CPR Unit translated into social losses by means of the cross country elasticities of a set of social indicators to This Dissemination Note was prepared by Humberto Lopez, growth. The social indicators have been chosen so World Bank country Economist for El Salvador, and edited by the that they capture the essence of the MDGs. Conflict Prevention and Reconstruction (CPR) Unit This note series is intended to disseminate good practice and key findings on conflict prevention and reconstruction. The series is edited by All in all, the results suggest significant social costs. the CPR Unit in the Social Development Department of the Our lowest estimate of virtual GDP indicates that Environmentally and Socially Sustainable Development Network should armed conflict not have taken place, El of the World Bank The views expressed in these notes are those Salvador's per capita GDP in 2000 could be at least of the authors and do not necessarily reflect the views of the World Bank Group, its Executive Directors, or the countries they 75 percent higher than its actual value; that the represent. CPR Dissemination Notes are distributed widely to poverty headcount could have been lower by 15 Bank staff and are also available on the CPR website percentage points, child malnutrition halved to about (http://www.worldbank.ora/conflict). 6 percent, and infant mortality cut by one-fourth from its current levels. We also argue that in the absence of the conflict the country's education indicators would be significantly better, especially in secondary CPR Dissemination Notes and tertiary education where enrolments (35 and 18 percent respectively) could be 6 and 10 percentage No. 1 (March 2002) Rebuilding the Civil Service in a points higher, and that public infrastructure coverage Post-Conflict Setting: Key Lessons of Experience could be higher by about 40 percent. No. 2 (April 2002): Aid, Policy and Growth in Post-Conflict Countries. El Salvador is considered a relative success story in post-conflict reconstruction. There has been no No 3 (May 2002). Child Soldiers Prevention, renewal of conflict, it has undergone a smooth Demobilization and Reintegration. transition to democracy, growth has revived, especially in the immediate post-conflict period, and No. 4 (June 2002) The Structure of Rebel Organizations: over the past decade the Government has Implications for Post-Conflict Reconstruction. maintained prudent macroeconomic policies and No. 5 (October 2002): The Conflict Analysis Framework implemented major structural, social and institutional (CAF): Identifying Conflict-related Obstacles to reforms. Despite this impressive track record, a Development. decade after the fighting stopped El Salvador's economy remains considerably behind where it was No. 6 (November 2002): Colombia: Development and in terms of output level when the conflict started and Peace in the Magdalena Medio Region has lagged substantially in terms of poverty and social indicators. No. 7 (December 2002): Conflict and Labor Markets in Manufacturing The Case of Eritrea El Salvador's experience shows that armed conflicts No. 8 (January 2003): The Social and Economic Costs of are extremely costly not only in terms of lives lost Armed Conflict in El Salvador and physical damage incurred, but also in terms of 4