Abstract. As introduced in Chapter 1, longitudinal and panel data are often heterogeneous and may suffer from problems of attrition. This chapter describes models for handling these tendencies, as well as models designed to handle omitted-variable bias.
Heterogeneity may be induced by (1) fixed effects, (2) random effects, or (3) within-subject covariances. In practice, distinguishing among these mechanisms can be difficult, although, as the chapter points out, it is not always necessary. The chapter also describes the well-known Hausman test for distinguishing between estimators based on fixed versus random effects. As pointed out by Mundlak (1978aE), the Hausman test provides a test of the significance of time-constant omitted variables, certain types of which are handled well by longitudinal and panel data.
This ability to deal with omitted variables is one of the important benefits of using longitudinal and panel data; in contrast, attrition is one of the main drawbacks. The chapter reviews methods for detecting biases arising from attrition and introduces models that provide corrections for attrition difficulties.
Heterogeneity is a common feature of many longitudinal and panel data sets. When we think of longitudinal data, we think of repeated measurements on subjects. This text emphasizes repeated observations over time, although other types of clustering are of interest. For example, one could model the family unit as a “subject” and have individual measurements of family members as the repeated observations.