Measuring Long-term Performance: a Regression Based Generalization of the Calendar Time Portfolio Approach

We present a new, regression-based methodology for decomposing the risk-adjusted performance of private investors, firms, and mutual funds. Our technique allows for the inclusion of multivariate and continuous subject characteristics in the analysis and it ensures that the statistical results are robust to cross-sectional and temporal dependence. Considering a unique dataset on 40,000 European private investors, we apply our methodology to reassess some of the most popular hypotheses on the performance of private investors. By testing the various hypotheses on a stand-alone basis, we are able to confirm the results of previous studies. However, when we apply our methodological framework to perform a joint test of the hypotheses, our results question several findings from previous research on private investor performance. More generally, our results indicate that (1) testing for a specific hypothesis separately and (2) erroneously ignoring cross-sectional dependence in microeconometric data can both lead to severely biased statistical results.