Searching for climate–conflict links

Environmental scarcity caused by climate change has been implicated as a driver of violent conflict. Now, research shows significant bias in the regions analysed for climate–conflict links. This may limit understanding of the socioeconomic and political conditions in which such conflict occurs, and how these conflicts could be prevented.

Climate change research has a streetlight problem: the tendency for researchers to focus on particular regions and countries for reasons of convenience or data availability rather than broader relevance, policy importance, or strong theoretical justification1, 2. The streetlight effect takes its name from a common anecdote used to illustrate it: a confused person looks for their lost keys underneath a streetlight, even though they lost them halfway up the street. When queried why, they respond, “This is where the light is”. Writing in Nature Climate Change, Courtland Adams and colleagues show strong evidence for the streetlight effect in research on climate–conflict links: the potential for climate change to cause armed conflict3.

Credit: Kypros/Moment/Getty

Interest in climate–conflict links spiked in the mid-2000s when United Nations Secretary General at the time, Ban Ki-Moon, asserted that the war and humanitarian crisis in Darfur were fuelled by climate-change-related desertification and conflict between farmers and pastoral communities4. More recently, Syria has become the go-to example of a conflict caused by climate change, with then US President Obama saying that climate-change-related drought and rapid urbanization “helped fuel the early unrest in Syria, which descended into civil war”5. Even if climate links to these specific conflicts are contested — in each case, it is difficult to argue that climate change was the cause or even a principal cause of the conflict6, 7 — they have nevertheless driven popular and scholarly interest.

Adams and colleagues systematically reviewed the now large body of peer-reviewed research on climate–conflict links and found it to be skewed in three major ways. First, scholars have gone looking for climate–conflict links in countries and regions that have been conflict-affected, rather than those most affected by climate change. This is known as sampling on the dependent variable, or looking mainly at cases and examples that are most likely to confirm the hypothesized relationships: in this case, that climate-change-related environmental scarcity led to conflict. By sampling on conflict — for example, cases such as Syria where violent conflict has occurred following an environmental shock — scholars ignore those places where environmental shocks were not followed by armed conflict. The drought that affected Syria also affected neighbouring Jordan, Lebanon and Cyprus, yet widespread violence did not occur there.

Second, scholars have not focused more attention on those cases most likely to be affected by climate change. Using several different measures of exposure and vulnerability to climate change, Adams and colleagues find no robust relationships between exposure or vulnerability and scholarly interest. That is, scholarly interest is not driven by the degree of climate impacts. Comparatively climate-change-exposed countries such as Bangladesh and Haiti, for instance, have received very scant attention from researchers investigating climate–conflict impacts.

Third, Adams and colleagues find strong evidence for a streetlight effect: scholars have been focusing effort on cases where it is more convenient to conduct research. Controlling for other factors such as conflict prevalence and climate change exposure, they find that former British colonies receive more attention. Conducting research in former British colonies such as India and Kenya is easier for a variety of reasons, ranging from data availability — the British Navy, for instance, collected high-quality depth data for Africa’s Lake Victoria over 100 years ago — to language (English, the lingua franca of twentieth- and twenty-first-century scientific discourse8), and the volume of existing literature on the country or region. In addition, practical, travel-related considerations may matter as well. Kenya, which ties as the single most-often-mentioned country in the climate–conflict literature, is neither particularly conflict-prone nor climate vulnerable. It is, however, a former British colony served by 18 different international airlines. Burundi is both conflict-affected and climate-vulnerable, but is Francophone and served by only three international airlines. Perhaps unsurprisingly, Burundi fails to crack the top ten countries in terms of scholarly interest.

These findings have two powerful implications for our ability to understand the socioeconomic and political conditions in which climate-related conflict is likely to emerge and for informing policy interventions designed to mitigate climate–conflict risk. First, sampling on conflict prevalence, rather than physical exposure, means researchers can say less than we might like about the specific social, economic, and political contexts in which climate change may lead to violence. Tanzania, for instance, shares many historical legacies and environmental conditions with neighbouring Kenya and Uganda, but has experienced virtually no armed conflict. These ‘dogs that don’t bark’ must be investigated more closely.

Second, if climate–conflict links are being studied primarily in atypical contexts — easy-to-access former British colonies — it limits our ability to infer what the research means for climate–conflict links more generally. Many are familiar with the ‘college sophomore problem’ in social psychology, where findings from well-designed studies conducted on a non-representative population sampled for convenience — college sophomores — may not generalize to the broader population. Research on climate–conflict links may be beset by similar problems.

The study by Courtland Adams and colleagues3 indicates that researchers have been ‘looking where the light is’ for climate–conflict links. Great care should be taken in designing future studies so as to minimize these sources of bias, and create a sounder scientific basis for informing policy interventions designed to decouple violence from climatic and environmental conditions.


  1. 1.

    Hendrix, C. S. Global Environ. Change 43, 137–147 (2017).

    Article  Google Scholar 

  2. 2.

    N at. Clim. Change 7, 229 (2017).

  3. 3.

    Adams, C., Ide, T., Barnett, J. & Detges, A. Nat. Clim. Change (2018).

    Google Scholar 

  4. 4.

    Ban, K.-M. A climate culprit in Darfur. Washington Post (16 June 2007).

  5. 5.

    Obama, B. Remarks by the President at the United States Coast Guard Academy Commencement (The White House Archives, 2015).

  6. 6.

    Butler, D. Nature 447, 1038 (2007).

    CAS  Article  Google Scholar 

  7. 7.

    Selby, J., Dahl, O. S., Fröhlich, C. & Hulme, M. Pol. Geo. 60, 232–244 (2017).

    Article  Google Scholar 

  8. 8.

    van Weijen, D. Res. Trends 31, 7–8 (2012).

    Google Scholar 

Download references

Author information



Corresponding author

Correspondence to Cullen S. Hendrix.

Rights and permissions

Reprints and Permissions

About this article

Verify currency and authenticity via CrossMark

Cite this article

Hendrix, C.S. Searching for climate–conflict links. Nature Clim Change 8, 190–191 (2018).

Download citation

Further reading


Nature Briefing

Sign up for the Nature Briefing newsletter — what matters in science, free to your inbox daily.

Get the most important science stories of the day, free in your inbox. Sign up for Nature Briefing