What To Do (and Not To Do) with Time-Series Cross-Section Data
We examine some issues in the estimation of time-series cross-section models, calling into question the conclusions of many published studies, particularly in the field of comparative political economy. We show that the generalized least squares approach of Parks produces standard errors that lead to extreme overconfidence, often underestimating variability by 50% or more. We also provide an alternative estimator of the standard errors that is correct when the error structures show complications found in this type of model. Monte Carlo analysis shows that these "panel-corrected standard errors" perform well. The utility of our approach is demonstrated via a reanalysis of one "social democratic corporatist" model.
© 1995 American Political Science Association. We would like to thank Michael Alvarez, Geoffrey Garrett, Peter Lange, Alexander Hicks, and Duane Swank for generously providing their data for replication purposes. William Greene, Gary King, and Glenn Sueyoshi deserve more than the usual thanks for helping us to figure out both what we were doing and how to communicate it. We also thank Michael Alvarez, Charles Franklin, Ronald Gallant, Elizabeth Gerber, Sung Hahm, William Heller, Mark Kamlet, Brian Loynd, Glenn Mitchell, Chris Mooney, Jimmy Sanders, Renee Smith, James Stimson, and Michael Thies for helpful comments and conversations. We are grateful to Peter Williams for providing new LATEX styles. Katz thanks the National Science Foundation for a graduate fellowship that funded his work on this project while he was at the University of California, San Diego. Earlier versions were delivered at the 1993 annual meetings of the American Political Science Association in Washington, the Political Methodology Group in Tallahassee, and the Midwest Political Science Association in Chicago. All computer codes and data related to this article may be obtained via ftp to weber.ucsd.edu.