This paper addresses two critical limitations of repeated cross-sectional (RCS) designs: autocorrelation and the need to choose between aggregate or individual-level analysis. Unlike traditional panels, RCS data features unique units observed once per time period.
First, it introduces a solution to autocorrelation by combining ARFIMA filtering for longer time series with multilevel modeling. This dual method allows simultaneous estimation of both aggregate and individual parameters.
Second, the approach is validated through Monte Carlo simulations and three empirical applications across different countries. By demonstrating how this technique improves inference accuracy while preserving nuanced analysis capabilities, it offers a valuable tool for political scientists working with single-wave survey data.
