Dealing with missing and incomplete data

Abstract

Missing data is an issue frequently encountered by international relations researchers, and there's no single "right answer" to any missing data problem. Rather, there are numerous options, each with its own costs and benefits, and each of which relies on a set of assumptions about the data. This chapter starts by suggesting a general roadmap for thinking about incomplete data and choosing missing data handling approaches. Next, it introduces different types of missing data - in terms of the location of missing values and their relationship to the observed data. It then reviews deletion, single imputation, multiple imputation, and maximum likelihood estimation approaches to handling missing data, discussing the assumptions they rely on their potential costs and benefits. Finally, it highlights how thinking carefully about missingness can provide insight into other research design considerations, such as scope and generalizability, causal inference, and measurement error.