Drawing Accurate Inferences About the Differences between Cases in Time-Series Cross-Section Data

Abstract

Researchers with time-series cross-section (TSCS) data should be aware that different methods to analyze TSCS data are designed to produce inferences about different aspects of the data. Many commonly used methods consider only the variation over time. Some consider only the variation across cases, and others draw inferences by averaging the two dimensions of variance. A new method, called the between effects estimation routine (BEER), is developed to maximize information from the TSCS data to model the cross-sectional effects while allowing these effects to change over time. Individual regressions are run, and the results are combined using a Bayesian averaging process that places larger weights on time points which are similar and proximate to the time point under consideration. This method is applied to reconsider the effect of income on state voting in US presidential elections. Simulations demonstrate that BEER is more accurate than the other commonly used TSCS methods when the goal of the researcher is to model cross-sectional effects that change in a smooth way over time.

Keywords: time series cross section, elections, simulation, Bayesian