Challenges for Causal Inference in Obesity Research

This chapter reviews the empirical strategies that social scientists commonly use to make causal inferences in the absence of randomized experiments and then highlights particularly challenging issues in obesity research. It provides a non-technical summary of several approaches to making causal inferences when the researcher only has observational data available, that is, data in which the researcher cannot control the values of the treatment of interest. The instrumental variable methods in obesity research are described. Body weight is difficult to measure, in both conceptual and practical senses. It is difficult to explain, making the search for identification through instrumental variables even more difficult than usual and undermining efforts to characterize the causes of body weight itself. Body weight is a stock which changes slowly over time, such that small influences on body weight over time may eventually cause large changes in weight but will be very difficult to detect statistically.