Many research and policy questions surrounding migration are causal questions. What causes people to migrate What are the consequences of migration for the migrants, their families, and their communities Answering these questions requires dealing with the self-selection inherent in migration choices. Field and natural experiments offer methodological approaches that enable answering these causal questions. This paper discusses the key conceptual and logistical issues that face applied researchers when applying these methods to the study of migration, as well as providing guidance for practitioners and policymakers in assessing the credibility of causal claims. For randomized experiments, this includes providing a framework for thinking through what can be randomized
discussing key measurement and design issues that arise from issues such as migration being a rare event, and in measuring welfare changes when people change locations
as well as discussing ethical issues that can arise. The paper then outlines what makes for a good natural experiment in the context of migration, and discusses the implications of recent econometric work for the use of difference-indifferences, instrumental variables (and especially shift-share instruments), and regression discontinuity methods in migration research. A key lesson from this recent work is that it is not meaningful to talk about "the" impact of migration, but rather impacts are likely to be heterogeneous, affecting both the validity and interpretation of causal estimates.