The Problem of Extrapolation in Economics

This thesis explores the problem of extrapolating causal claims in the social sciences, particularly economics. The problem of extrapolation is the problem of inferring something about a phenomenon of interest in one context, based on what is known about it in another. For example, we may want to infer that a medicine works in population $Y$, based on the fact that we know it works in population $X$. Extrapolation is the inferential process of generalizing or transporting claims about a phenomenon of interest to new populations or settings. The answers to the problem of extrapolation in philosophy of science aim to explain how successful extrapolation is possible, as there will always be relevant differences between the two systems. I study extrapolation from the viewpoint of philosophy of science, which aims to both analyze and complement science and scientific knowledge. I also use a case study with two examples to further illustrate the relationship between the theoretical approaches to extrapolation in philosophy of economics and actual studies in experimental economics. I focus on comparative process tracing, a general account of extrapolation developed by philosopher of science Daniel Steel, and its success in extrapolating causal claims from field experiments in economics. The first chapter introduces central concepts and key questions. The second chapter discusses external validity, a concept typically used in economics to describe the potential of causal claims to be extrapolated. The third chapter introduces comparative process tracing, which explains how and why extrapolation can be based on knowledge about causal mechanisms. Next, I discuss field experiments in economics and methodological issues of extrapolation particular to them. The fourth chapter consists of a case study, which shows the limitations of approaching extrapolation in economics with comparative process tracing. The last chapter concludes. The central conclusion of this thesis is that even though comparative process tracing is meant as an account of extrapolation that can explain and apply to extrapolation across disciplines, applying it to economics faces methodological challenges. Nevertheless, the issues it faces with regard to field experiments in economics do not refute it as an account of mechanistic extrapolation. I propose that comparative process tracing is a theoretically comprehensive epistemological account of extrapolation in the social sciences, but it must be complemented with a systematic methodological account of problems of extrapolation in practice. This methodological account complements and enhances epistemological analysis of extrapolation.