It is widely believed that education is a crucial factor in curbing political violence, but establishing causal evidence of this notoriously difficult. This column uses a large-scale school construction programme in Indonesia and newspaper reports of violence to tackle this problem. The results show that the construction of primary schools led to statistically significant reductions in conflict that grew larger over time.
Politicians, journalists, and NGOs tend to believe that education can be the solution that stops armed fighting. Kofi Annan, the former Secretary-General of the UN for example stressed that “[e]ducation is, quite simply, peace-building by another name. It is the most effective form of defence spending there is.” And Nelson Mandela put forward that “[n]o one is born hating another person because of the colour of his skin, or his background, or his religion. People must learn to hate, and if they can learn to hate, they can be taught to love (...)”. Or in the words of the Philippine’s Chief Negotiator Miriam Coronel-Ferrer, “[e]ducation is the lasting solution to Mindanao war."
But what do the data say? It turns out that there is also statistical evidence in economics that documents a series of suggestive empirical regularities between economic perspectives and political violence. In particular, some of the strongest predictors of conflicts are poverty and dismal economic opportunities (Miguel et al. 2004), and human capital accumulation could be a prominent way of improving economic chances. In turns, perspectives of well paid jobs may result in a larger opportunity cost of leaving paid work for participating in armed violence. Other subtle arguments linking education to political stability are that wealth embodied in human capital can be less easily appropriated than physical capital and that education may transmit values of tolerance (De la Brière et al. 2017). At the same time, of course, one could think of scenarios where high education in the presence of poorly working labour markets could lead to an explosive blend of discontent fuelling tensions.
Directly studying the nexus between education and political violence is notoriously difficult, as both variables are endogenous, and causality could go both ways and be influenced by a series of confounders. For example, cross-country correlations of education being positively associated with peace could be spurious and entirely driven by, for example, good governance or high GDP. In a recent paper, we aim to carry out an empirical investigation of the effects of education on the risk of conflict, drawing on a quasi-natural experiment with the help of which we strive to address a series of statistical biases pertaining to the study of the impact of education (Rohner and Saia 2019).
In particular, we assess the impact of the INPRES school construction programme in Indonesia, which took place between 1974 and 1978 and is one of the largest and fastest school construction programmes ever implemented. Under its aegis, 61,000 new primary schools were constructed, which represents more than a doubling of the stock of available schools. This programme has been extensively studied by economists (most prominently by Duflo 2001), finding, for example, that it improved education levels and labour market outcomes. To the best of our knowledge, our paper is the first to link this school construction programme to armed conflict (Rohner and Saia 2019). In order to be able to carry out this analysis we first had to create, using newspaper data and applying techniques of web crawling, web scraping, and text recognition, a novel dataset on conflict in Indonesia at a fine-grained (district) level over a 40-year period.
Our main identification strategy relies on a difference-in-differences approach, where we exploit the impact of sharp changes in education provision, leading to sharp changes in fighting outcomes. The baseline specification controls for district fixed effects (filtering out time-invariant factors), province-year fixed effects (filtering out shocks to the 26 provinces), and district-specific time trends (accounting for long-run evolution and catching-up). We also control for potential drivers of school construction and carries out a vast array of sensitivity checks, controlling for other government programs, migration, weather, and natural resource shocks, among others. We investigate robustness using alternative conflict data and other econometric specifications, and restricting the analysis to a series of subsamples.
We find that INPRES school construction has indeed led to a statistically significant decrease in conflict. This impact of INPRES school construction is found to become bigger over time, as one would expect (i.e. in the years after the programme, the number of cohorts leaving the new INPRES schools has mechanically increased, and with greater age treated cohorts have approached the critical period of the life cycle when child soldiering peaks, and where education typically can make the largest difference). The effect of INPRES is quantitatively sizable, with a one standard deviation increase in schools built resulting, at the end of the sample period, in a drop of three-quarters from the baseline conflict risk.
In a following step, we study the potential mechanisms at work. It turns out that both the religious composition of the local population as well as the local economy affect the size of the impact of school construction. As expected, we find stronger effects in places where the economic returns to education are high (where opportunity cost effects are plausible), and in areas that are religiously polarised (where open-mindedness and tolerance may be particularly critical). Using individual-level data, we also document that exposure to INPRES school construction increases inter-religious trust and community involvement. This is in line with the result that while increased schooling reduces the potential for violent rebellion, it does not curb the scope for peaceful protests. Thus, schooling can lead to a shift from violence to voice.
To sum up, our results suggest that while the lack of education may not be a unique determinant of political violence, an increase in schooling could well be a key public policy for reducing armed civil conflict intensity. These findings are based on a large-scale primary school construction programme in Indonesia, and a natural question that remains is to what extent these findings also apply to other contexts. In particular, we encourage further research studying school expansion outside Indonesia and beyond primary schools.