Climate, armed conflict and the problem of attribution slippage

A new paper in the journal Nature explores the role of climate as a risk factor in armed conflict.  It is interesting for a number of reasons, not least since it uses the method of expert elicitation to explore differences and similarities in experts’ subjective interpretation and judgement of evidence.  But the study also suffers from a conceptual ambiguity with regards to the attribution of conflict to ‘climate’ or to climate change.

Katharine Mach of Stanford University has previously advanced general arguments in favour of using expert elicitation by the IPCC to clarify areas and degrees of consensus between experts with respect to aspects of climate science and impacts.  The method offers clear advantages over crude and simplistic ‘head counts’ that others in climate communication research and advocacy have promoted.  In this new paper, Mach leads a team which has applied the method of expert elicitation to assess the relationship between climate and armed conflict.

The claim that climate change exacerbates violent conflict is one that has been advanced by some for at least two decades, if not longer.  It lies behind some of the most inflated rhetoric of the ‘existential climate crisis’ and feeds the discourses of climate securitisation and climate refugees.

This rhetoric reached a peak in the autumn 2015 just ahead of the Paris Climate Summit, amidst claims that the Syrian civil war was a ‘climate war’.  For example, Bernie Sanders claimed back then that climate change is linked to terrorism and will cause international conflicts.  And it is not entirely far-fetched to suggest that this public fear of future climate wars, and the imagined “tides of migrants” thus unleashed, was used by some to mobilise the Syrian migrant crisis in Europe to feed the UK’s Brexit vote in June 2016.

Yet the evidence for the Syrian civil war being a “climate-caused” war, let alone a “human-caused climate change-caused” war, is thin.  Jan Selby, I and two colleagues explored this claim in detail in our 2017 paper and concluded that “policymakers, commentators and scholars alike should exercise far greater caution when drawing such linkages or when securitising climate change”.

So for this reason, and others too, this new paper from Mach et al. which carefully explores the basis of these broad-scale claims is welcome.  This is especially so since the study uses expert elicitation to expose expert judgement, a method of knowledge assessment I have long espoused and also experimented with.    

The study is careful in the way that experts are selected, and is suitably nuanced in the way that questions are framed, the elicitation process managed and, for the most part, that conclusions are drawn.  For example, it concludes that, for these experts, and relative to other drivers of armed conflict, “the role of climate is judged to be small”.

But there is one conceptual problem in the study to which I draw attention.  And it is one that subtly pervades significant numbers of so-called climate impacts studies.  Is it climate that is the risk factor under investigation, or is it human-caused climate change?

The paper in fact makes clear that experts were asked whether they believed that climate (in its broadest sense, i.e., “climate-related variability [including El Nino], hazards, trends and change”) contributed to the risk of armed conflict.  With respect to climate over the past century these experts judged that only around 5% (range 3 to 20%) of conflict risk had been influenced by climate.

The experts were then asked about the future effects of climate on their evaluation of future conflict risk, if the world as a whole warmed by either 2°C or 4°C.  (It was not clear whether this warming was supposed to be with respect to today’s baseline or to a pre-industrial climate).  Their risk evaluation of climate increased (as group means) to 13% and 26% respectively. 

What is not clear from the study, however, is the extent to which experts judged that climate’s natural variability – drought, El Nino, heatwaves, floods, etc. – contributes to conflict risk as opposed to human-caused “climate change”.  For example, given that the world has already warmed by 1°C over the past century, is the experts’ 5% estimate of historical risk related to a changing climate or to climate’s natural variations?

The study seems to be set-up as though, for the purposes of the expert evaluation, these differentiators of climate – its natural variability and human-caused change – cannot be distinguished.  This is what I mean by ‘attribution slippage’.  If this is so, then we are left little wiser about just how important a risk factor for armed conflict human-caused “climate change” is relative to climate.  Is the additional 8% [21%] risk judged by these experts to arise from a 1°C [3°C] warmer world (I’m assuming a pre-industrial baseline) the result of either human-caused climate change or risk factors that become more sensitive to climate’s natural variability?

The significance of this question returns us to the overall motivation for conducting studies like this in the first place and why claims about climate change and conflict proliferate.  The normative goal for most people when thinking about violent conflict would be: ‘How can we lessen the risk of future armed conflict?’ 

Can this best be achieved by reducing emissions of greenhouse gases and thus shaving of fractions of degrees Celsius of global warming from the future?  Or should we aiming to reduce the sensitivity of future human societies to climate events – of whatever cause – through improving human development, increasing state capability, reducing intergroup inequality, investing in peace-making, and so on—all risk factors deemed by these experts to be much more significant than climate in provoking armed conflict?

Reducing emissions of carbon dioxide might lessen the risk of future armed conflicts a little.  But satisfying human development needs, reducing inequality and investing in institutions that bring political stability would seem likely to accomplish a much greater reduction in risk.

Mike Hulme, 22 June 2019