Sound theory and methods form the backbone of good empirical research. We strive to contribute to new theories in land system science and to develop and apply sound statistical methods.  In particular, we have been engaged with the on-going theoretical debate centered around land sparing and land sharing.  Methodologically we have worked to improve causal inference in non-experimental settings as in the case of protected area effectiveness.

Land Sparing & Land Sharing

The Land Sparing / Land Sharing debates is centered on a simple question: If humans desire a landscape that produces both food and biodiversity, is it more efficient to produce food intensely on a small portion of land and protect the rest for natural ecosystems, or is it better to produce food at a lower intensity, and integrate biodiversity into the food system. This simple question has lead to fierce debates within the land use / conservation community. Our approach to the problem is to model the mathematical foundations of this theory. Our theoretical work has shown quite clearly (in our opinion at least) that the answer to this question is “it depends.”

Causal Inference

To better understand the world around us, we often want to know the effect an action on an outcome. That is, we want to know the causal impact of a treatment (policy, economic shock, natural disaster…) on an outcome (land use decision, economic growth, biodiversity). Causal inference is the science of quantifying the impact of a treatment on an observation. In some natural sciences, the impact of a treatment can be deduced using experimentation and sound experimental design. In social ecological systems, experiments are rarely possible, and yet the need to understand the impact of a treatment (for instance a land use policy) on an outcome is equally as important. We use innovative statistical techniques to uncover the causal impacts of treatments in situations where experimentation is impossible. We have used a variety of techniques including difference-in-differences, matching, panel regressions, and regression discontinuity design in our own work. These techniques are well known to many social scientists, but perhaps less well known to some with backgrounds in the natural sciences. We have written a guide to these techniques, including a tutorial with code in R and STATA.  The associated manuscript can be found here