Causal inference in conservation

A CEED Workshop, University of Queensland, July 2015

Workshoppers discuss the finer points of causal inference at the CEED workshop at the University of Queensland in July. (Photo by Jane Campbell)

Workshoppers discuss the finer points of causal inference at the CEED workshop at the University of Queensland in July. (Photo by Jane Campbell)

How can we be confident that a conservation intervention had a positive impact? This was the central question at a recent workshop at the University of Queensland on causal inference.

We’ve all heard the adage that ‘correlation does not imply causation’ – so we wanted to find out what does! We brought in Paul Ferraro, an international expert on the topic to train 30 CEED staff and students on the art (and science) of causal inference. (Paul is the Bloomberg Distinguished Professor at Johns Hopkins University; he’s also an advisor to the Global Environmental Facility). Paul has over a decade of experience in environmental policy evaluation, including vibrant contributions to both the peer-reviewed literature and conservation practice.

Participants in the workshop included representatives from three CEED nodes (Brisbane, Melbourne, Perth), CSIRO, as well as CEED visitors from the University of British Columbia, the University of Cambridge, the Basque Centre for Climate Change, and James Cook University.

And what did we discover? It turns out that the key to identifying causal effects is to eliminate rival explanations that may mimic or mask a relationship between a cause and an effect. Logical, yes, but this consideration requires a greater attention to analysis design than is often given. It can have significant implications for interpreting the effectiveness of conservation actions. For example, naively comparing deforestation rates inside and outside parks (and not considering that parks are usually biased towards areas of low conflict with other uses) may lead to an overestimation of the effectiveness of protected areas in delivering avoided deforestation – by over 65%! Lucky we now know about causal inference!

While this classic approach to causal inference is great for post hoc analyses, much of the work we do is planning for the future. The second half of the week narrowed the workshop to thirteen participants, nutting out the question of how we can incorporate these methods into our planning work. Keep an eye out for our upcoming paper on this!

More info: Elizabeth Law

Leave a Reply