Speaker: Ryan Barnhart, York University
                Department of Psychology

Title: Fixed versus Random Effects Models for Longitudinal/Panel Data Analysis. There's Really Only One Estimator and How the Hausman Specification Test Should Be Abandoned

Abstract: In the social sciences, there have been two very strong arguments concerning the adoption of either Fixed or Random Effects specifications to address the dependencies within observations in longitudinal and panel data analysis. In particular, the question of exogeneity/endogeneity of the random or fixed intercepts and the issue of consistency of the random effects estimator. In this talk I will introduce the strengths and weaknesses of each approach as well as the Hausman specification test used to determine whether to use the Fixed or Random Effects specification. I will finally demonstrate that under correct model formulation the Hausman test is aimless and obsolete as the Within Estimator of the Fixed Effects model and the Generalized Least Squares Estimator of the Random Effects model are algebraically equivalent.

Suggested Readings:
        Clark, T. S., & Linzer, D. A. (2012). Should I use fixed or random effects? ( Working paper 1315).
       Clarke, P., Crawford, C., Steele, F. & Vignoles, A. (2010). The choice between fixed and random effects models: Some considerations for education research (Working Paper). 
        Snijders, T. A. B., & Berkhof, J. (2008) Diagnostic checks for multilevel models. In J. de Leeuw, & E. Meijer (Eds.), Handbook of Multilevel Analysis, (pp. 141-175). New York: Springer.
        Troeger, V. E. (2008) Problematic choices: Testing for correlated unit specific effects in panel data (Working Paper).