Resumen de Misspecification Effects in the Analysis of Panel Data

Marcel D.T. Vieira, Peter W.F. Smith, Maria de Fátima Salgueiro

• Misspecification effects (meffs) measure the effect on the sampling variance of an estimator of incorrect specification of both the sampling scheme and the model considered. We assess the effect of various features of complex sampling schemes on the inferences drawn from models for panel data using meffs. Many longitudinal social survey designs employ multistage sampling, leading to some clustering, which tends to lead to meffs greater than unity. An empirical study using data from the British Household Panel Survey is conducted, and a simulation study is performed. Our results suggest that clustering impacts are stronger for longitudinal studies than for cross-sectional studies, and that meffs for the regression coefficients increase with the number of waves analysed. Hence, estimated standard errors in the analysis of panel data can be misleading if any clustering is ignored.