Will the economic costs of climate change lead to more violent political conflicts in the future? While giving a definite answer to that question is an exercise in futility, I believe we can improve on previous research by combining and improving on three aspects on current climate-conflict research: Theory-near empirical testing, improved causal inference and dynamic forecasting. From previous research, I come with the following observations:
- To date, conflict research has not been able to establish a causal effect between economic growth as influenced by climatic variations and armed conflict (see e.g. Ciccone (2011)’s critique of Miguel et.al. (2004)). One possible reason for this zero-finding is that the relevant signal in rainfall is removed when averaging over countries. Moving to sub-national units of observation, however, leads to other methodological issues that research to date has not dealt with properly.
- Empirical tests in this field are often made many steps away in the causal chain from the theories that we are interested in (e.g. correlating rainfall with violent events in some spatially defined area, arguing that the mechanism is about some individuals being affected economically by this rainfall, which again leads them to choose violent behavior).
- Very few studies are actually looking at climate change (rather most studies look at climate variability), and those that do must convince readers how many hundred year old data is relevant to explain the world today. There is a conceptual, theoretical and interpretative gap between climate variability studies and effects of climate change in the future. Moreover, even if that gap can be closed, any statement about conflict-propensities in the future needs to take into account the development of other relevant factors between now and the relevant future.
This thesis aims first to improve on the causal inference methodology in climate-conflict research and get closer to the causal mechanisms that we are interested in. This is done in three articles, all of which builds on the idea that precipitation can, when treated carefully, be seen as a randomized treatment. The articles study different types of violent political outcomes and attitudes to such use of violence. They all try to get closer to the causal mechanisms, while discussing methodological issues of the identification of causal effects. Second, through using dynamic regression models and socio-economic scenarios that have climate change mitigation