By Dominic Rohner, Faculty of Business and Economics (HEC Lausanne), University of Lausanne and Center for Economic Policy Research (CEPR), and Alessandro Saia, Faculty of Business and Economics (HEC Lausanne), University of Lausanne
Armed conflict is a major obstacle to human happiness and prosperity. The most visible consequence of warfare is, of course, the human death toll, leaving millions of families shattered. But below this surface, the grim consequences of fighting go further. The economic cost is very considerable, with the average war leading to a total loss of about 15% of GDP, human capital accumulation is slowed down, inter-group trust is threatened, and psychological suffering and trauma become widespread.1
While academic research on conflict has boomed in recent years, the lion’s share of contributions has focused on factors that are well-suited for statistical analysis but that are difficult to modify by policymakers. In particular, while we know that ethnic diversity, adverse weather shocks and natural resource discoveries play a role in the occurrence of conflict, there are not many obvious policies that can modulate these parameters.
In contrast, key policy choices of governments related to the biggest areas of public activity have received only very limited attention. This is hardly surprising. Any quantitative appraisal of major government policies, such as, education, health or welfare state policies, faces important statistical challenges. In particular, given that these policies are endogenously chosen, one can think of many confounders that jointly determine, say, education spending and peace-building. For example, a benevolent and capable government may boost both education perspective and the scope for peace. Hence, if Singapore and Finland benefit from peace and good education outcomes, while, say, Libya and Somalia have a more dismal schooling performance and greater levels of political unrest, this may not reflect any causal impact of education on peace but could be entirely driven by the general quality of governance. Put differently, any positive correlation on the country-level between good education outcomes and peace may be spurious and not reflect any causal impact of education on peace.
To overcome such methodological challenges, one option is to focus on some form of natural experiment, where some exogenous variation in education supply is exploited for studying its impact on the risk of conflict. This is exactly the goal of our recent study “Education and Conflict: Evidence from a Policy Experiment in Indonesia“2, where we carry out an empirical analysis of how school construction affects the risk of armed conflict. In particular, we focus on a quasi-natural experiment in Indonesia: the so-called INPRES programme that represents one of the biggest-scale and most rapid schooling expansion programmes in the world so far. The reason we focus on the INPRES programme is its large scale and the arguably exogenous planning of school construction.
Under the auspices of this policy programme, over 61,000 new primary schools were constructed between 1974 and 1978, representing a more than doubling of the stock of schools. The variation introduced by the INPRES programme has been investigated by various academic articles3 but has, so far, never been explored for the study of armed fighting. While this gap in the literature may seem surprising, one reason may be the lack of existing conflict data covering the years around the programme’s implementation. To overcome this challenge, we have put together our own novel and very extensive dataset of conflict events at the local district level, covering all of Indonesia during the period 1955-1994. Making use of web crawling, scraping and text recognition techniques, we have drawn on information from over 820,000 newspaper pages to code variables of conflict events taking place in a given district and year. Our panel dataset contains 289 districts over 40 years, amounting to 11,560 observations.
Our main statistical analysis relies on a so-called “difference-in-difference” approach, where we exploit the impact of sharp changes in education provision, resulting in sharp changes in conflict measures. We find that INPRES school construction resulted in a significant, quantitatively important decrease in the conflict risk. One more school per 1 000 children of primary-school age resulted in a drop in three-quarters of the baseline conflict risk at the end of the sample period. It is further shown that the effects are stronger in areas with high returns to education and with a latent risk of inter-religious tensions, pointing to the presence of both economic and social channels of transmission. The latter finding is also supported by the additional result that individual-level inter-religious trust gets boosted by more intense INPRES school construction.
These novel findings suggest that – arguably well-designed — primary school expansion programmes such as INPRES in Indonesia can have a high potential for curbing conflict. Still, we need additional evidence on other types of programmes in other contexts, such as in a very different country or for another school level (say, secondary school). However, these first results imply that – at least in some contexts — primary school extension may yield the double dividend of not only boosting productivity but also curbing conflict.
1. For a literature review of the costs of conflict, see Rohner, Dominic (2018) “Natural Resources and Conflict”, in: Palgrave Macmillan (eds) The New Palgrave Dictionary of Economics. Palgrave Macmillan, London; and Collier, Paul, 2007, The bottom billion: Why the poorest countries are failing and what can be done about it. Oxford: Oxford University Press.
2. See Rohner, Dominic, and Alessandro Saia (2019), “Education and Conflict: Evidence from a Policy Experiment in Indonesia”, CEPR Discussion Papers 13509.
3. See e.g. Duflo, Esther. “Schooling and labor market consequences of school construction in Indonesia: Evidence from an unusual policy experiment.” American Economic Review 91, no. 4 (2001): 795-813.