This article treats the analysis of "time-series - cross-section" (TSCS) data, which has become popular in the empirical analysis of comparative politics and international relations (IR). Such data consist of repeated observations on a series of fixed (nonsampled) units, where the units are of interest in themselves. An example of TSCS data is the post-World War II annual observations on the political economy of OECD nations. TSCS data are also becoming more common in IR studies that use the "dyad-year" design; such data are often complicated by a binary dependent variable (the presence or absence of dyadic conflict). Among the issues considered here are estimation and specification. I argue that treating TSCS issues as an estimation nuisance is old-fashioned; those wishing to pursue this approach should use ordinary least squares with panel correct standard errors rather than generalized least squares. A modern approach models dynamics via a lagged dependent variable or a single equation error correction model. Other modern issues involve the modeling of spatial impacts (geography) and heterogeneity. The binary dependent variable common in IR can be handled by treating the TSCS data as event history data.
- Error correction
- Feasible generalized least squares
- Random coefficients
- Robust standard errors
- Spatial econometrics
ASJC Scopus subject areas
- Sociology and Political Science