Friday, August 2, 2013
Climate Change and Conflicts: A Statistical Study
I came across an article on BBC [Rise in violence 'linked to climate change'] which talks about research by a team of US scientists who establish the relation between climate change and conflicts. After this, I looked for this publication to see which statistical tools they were using to establish causality. The article is published in science magazine [link]. The model they use is a time-series model given by
conflict_variableit = β × climate_variableit + μi + θt + ϵit
, where locations are indexed by i, observational periods are indexed by t, β is the parameter of interest and ϵ is the error. If different locations in a sample exhibit different average levels of conflict - perhaps because of cultural, historical, political, economic, geographic or institutional differences between the locations - this will be accounted for by the location-specific constants μ (“fixed effects”). Time specific constants θ (a dummy for each time period) flexibly account for other time-trending variables such as economic growth or gradual demographic changes that could be correlated with both climate and conflict.
This is a very simple model, still very powerful in determining correlation between conflict and climate change. What needs to be done is collect data (most difficult part) and then a regression to check if 'β', μi and θt are significant, and the researchers successfully prove that. However, the paper ends with the important part of 'causation', where the field is still in infancy to establish a clear cause which relates climate change and conflicts. There are many competing plausible theories, and there is still a lot of scope and research left to unravel that mystery. Why I find this interesting is because correlation and causation are confused very easily, and this study is an example where the way BBC reports the findings is very different from the way authors report in publication. Authors (Hsiang et al.) are reporting the quantitative influence of climate change on conflicts, not a causal link between the two variables, whereas BBC reports it as a causal link with rise in violence linked to climate change. One can watch this video to see what it means to differ between correlation and causation.