Reassessing Sampling Bias in Climate-Conflict Research

Is research into the links between climate change and conflict biased, and does this bias undermine our ability to draw conclusions about climate-conflict links? Adams et al. (2018, henceforth AIBD) argue the literature on climate-conflict links suffers from endemic sample selection bias. Because of this, the literature overstates links between climate change and conflict. In this article, we revisit the issue of sampling bias in climate-conflict research using a broader measure of scholarly interest based on bibliometric data from Google Scholar searches of some leading journals in the climate-conflict literature. We find i) weaker evidence of sampling on the dependent variable (armed conflict), but some sampling on types of conflict that fit with posited mechanisms and conflict typologies in the climate-conflict literature, and argue that some oversampling of conflict-prone cases may be a warranted; ii) researchers are sampling on the independent variable, with countries more exposed to climate stress receiving more attention; and iii) even stronger evidence of a streetlight effect, with former British colonies and countries of more general interest to scholars and the international community — as proxied by country-specific studies indexed by the Library of Congress and UNESCO World Heritage Sites — receiving greater attention. Thus, our findings both confirm and challenge existing conclusions in the debate over sampling bias in climate-conflict research. While researchers are sampling on climate change stress, the streetlight effect is stronger than previously suggested. This streetlight effect poses significant challenges for research and policy communities: how confidently can we generalize from well-studied cases? Does the streetlight effect diminish our ability to identify the specific economic, political, and social contexts in which climate-related conflicts occur? And finally, should funding agencies take the accumulating evidence for a streetlight effect into account when making funding decisions and prioritizing particular countries and world regions?