Estimating Causal Effects with Longitudinal Data: Does Unemployment Affect Mental Health?

Abstract

This case study discusses some properties and advantages of longitudinal data analysis, by examining the exemplary research question: Does unemployment affect mental health? A fundamental problem in the analysis of observational data is unobserved heterogeneity, also known as omitted variable bias. Unobserved heterogeneity refers to unobserved differences in respondents that affect both the independent and the dependent variables. The presence of unobserved heterogeneity will cause our effect estimates to be biased—in other words, incorrect. The case study first discusses the basic structure of longitudinal data and how these data differ from the cross-sectional data. Second, it briefly sketches why unemployment may impact mental health. Third, it lays out the difficulties of drawing causal inferences from observational data by portraying the counterfactual approach to causality. Fourth, it discusses popular models for the analysis of longitudinal data (random effects models, fixed effects models, and hybrid models). Fifth, the application of these models is illustrated by investigating how unemployment affects mental health with longitudinal German data (German Socio-Economic Panel).

locked icon

Sign in to access this content

Get a 30 day FREE TRIAL

  • Watch videos from a variety of sources bringing classroom topics to life
  • Read modern, diverse business cases
  • Explore hundreds of books and reference titles