Extreme temperatures and violence

In his book Civilization and Climate, Ellsworth Huntington (1876–1947) noted1 that “Almost any American or European who has travelled or resided within the tropics will confess that he has occasionally flown into a passion, and perhaps used physical violence, under circumstances which at home would merely have made him vexed.” This begs the question — why is there no violent conflict when severe droughts, heavy floods or hot temperatures hit rich countries. One reason is that high levels of social and political stability exist in comparatively developed countries: farmers’ crops fail, but they have insurance; property is damaged, but recovery centres are available to house victims; the injured are treated in hospitals; state agencies rush to assist. When disasters strike truly destitute societies with low levels of social stability, it compounds already poor governance, economic marginalization and substantial environmental vulnerabilities. Some studies in environmental security are in danger of promulgating a modern form of environmental determinism by suggesting that climate conditions directly and dominantly influence the propensity for violence among individuals, communities and states. For example, increased temperatures have recently been shown to be correlated with more violence and decreased temperatures with less violence2–4, leading to the claim that climatic anomalies are linked to social conflict at all scales and across all major regions of the world. The implication is that poor people act violently for natural reasons. However, such de-politicized analyses remove violence from its local, social and political contexts, and reduce conflict to an immediate and unmediated function of physical, biological and physical–geographical signals. Instead, the impression is given that environmental conditions determine conflict occurrence, type and rate, in line with an environmental determinist perspective that has been widely discredited as a lens for academic research about social instability. We caution against a renewed environmental determinism in the study of a climate–conflict link. Political (for example, regime type) and economic (for example, country-level income and inequality) measures should not be discarded in conceptual and statistical models of conflict occurrence. Most studies on the relationships between climate change, degradation, resources, disasters and violence confirm that any political violence is contingent on the political and economic characteristics of societies5. The deterministic approach, in contrast, is marked by substantial disregard for the complicated social processes and historical circumstances under which contemporary conflict emerges. In our view, environmental triggers need to be analysed in the context of political, social, demographic or economic explanations of conflict. Hsiang, Meng and Cane4 argue that the latter influences are ‘bad controls’ due to the possibility of endogenous effects of weather indicators on, for example, gross domestic product. But to exclude these variables — in favour of ‘fixed effects’6 — undermines key factors that are known to influence political violence. These factors include government capacity, poverty levels, democratic transitions, population characteristics, seasonality and previous and surrounding violence patterns. Inevitably, such an approach overemphasizes climate change as a causal factor. If the social setting that engenders conflict is ignored, there is also danger that multiple types of conflict become conflated. Conflict in wealthy areas can be reduced to the level of baseball skirmishes instead of large-scale societal violence. Most conflict researchers would not accept that such individual displays of anger compare to the political instability experienced in developing countries. Conflict rates and types in similar physical environments vary greatly. Civil wars, militia activity, communal violence and rioting may occur in the same state, but are subject to different political triggers7. Even when studies have detected a climate/ weather signature — such as the significant links between precipitation variation and violence — researchers often seek to explain the results through the incentives and disincentives for conflict. For example, studies of rebel movements show decreases in violence during the rainy season, when it is difficult for both national and rebel armies to traverse terrain with poor infrastructure8. In pastoral areas, violent activity is highest in periods preceding the rainy season, when strategic efforts to gain territory, wealth (for example, through cattle raiding) and control of migration paths is likely to yield the most profitable outcome. These strategic movements are associated with climate patterns, where participants consider and interact with the environment in their decision-making, cost/benefit calculations and logistical considerations. Failing to cast explanations in such terms risks reaching conclusions that are little different from those ascribing poverty to latitudinal location or lessened individual productivity to hot climates, as was common in European and American scholarship about a century ago. A de-politicized conflict framework necessarily ignores how politics in developing states create environments where risks accumulate for poorer, marginalized and geographically peripheral communities. The implications of such an analysis are especially problematic; namely, that a stimulus (temperature)–response (violence) relationship is a valid interpretation of conflict and that the more dangerous forms of violence (intergroup, genocide, civil and international war) will occur as poor people cannot contain violence during periods of environmental change. In arguing that