Armed conflicts pose significant challenges for the international community. Conflicts cause deaths, destroy capital and the environment, deter investment, and hinder economic activity. However, little is known about the economic costs of armed conflict and even less about the impact of conflict risk. The reason policymakers are in the dark about these relevant estimates involves two key issues. First, in order to study the impact on economic activity, we require timely and reliable measures of economic activity. Second, in order to study the impact of conflict risk, which is unobserved. The aim of this project is to shed light on these missing parts. In order to do so, we will rely on a database of more than 5 million newspaper articles and granular conflict and geographic data in order to predict conflict risk and economic activity and study the relation between the two.

The contribution of this project is two-fold. First, we will develop a forecasting framework which is able to track the entire conflict cycle, from forecasting new outbreaks, escalation of conflict, de-escalations out of conflict to the re-emergence in the post-conflict phase in which countries and regions are particularly fragile. We will do this through using cutting-edge machine learning which integrates a text-based forecast of conflict outbreaks with geo-spatial forecast of conflict dynamics during conflict. The estimated risks will be made accessible at conflictforecast.org. The website currently already features predictions based on the methodologies developed in Mueller and Rauh (2018, 2021). We will extend the current approach by moving from the country to the grid-cell level, and by predicting risk and intensity across the entire conflict cycle, rather than only the risk of an outbreak in a peaceful country.

Second, it relates to the literature on measurement of growth and GDP, in particular in developing countries. For instance, Deaton (2005) argues that we indeed do not know much about why statistics from different sources provide a different picture. There are a range of papers relying on light intensity measured at night by satellites. The rationale for using light intensity is that it tends to be generated by human activity, such as from buildings and cars in cities or on roads and has been shown to be correlated with economic activity (Chen and Nordhaus 2011, Henderson et al 2012). Pinkovskiy and Sala-i-Martin (2016) use a latent factor model to combine the information provided by government statistics, luminosity, and consumer surveys. We will add newspaper text as a source of economic activity in order to improve the accuracy and frequency of growth measures, in particular in developing countries.

The interdisciplinary work will be drawing from Economics, Political Science, International Relations, and Computer Science. The goal of the project is to be able to provide policymakers with a better understanding of the benefits of preventing conflict through early diplomatic action thereby decreasing inefficiencies in the allocation of public funds. Keynes himself emphasized the economic value of peace in his book “The Economic Consequences of Peace” (1919). To do this day, we struggle to put a number on the value of peace and this project wants to make progress on this ground by not only focusing on the dichotomy between peace and conflict, but by studying the impact of conflict risk.