A generalization of the calendar time portfolio approach and the performance of private investors

We present a regression-based generalization of the calendar time portfolio approach which allowsfor the inclusion of continuous and multivariate investor or firm characteristics in the analysis. Ourmethod is simple to apply and it ensures that the statistical results are heteroscedasticity consistentand robust to very general forms of cross-sectional and temporal dependence. Furthermore, ourregression-based technique also remedies several well-known weaknesses of the traditional calendartime portfolio approach. By considering a new, unique dataset on more than 40,000 Europeanprivate investors, we illustrate empirically that erroneously ignoring cross-sectional dependenceinherent in microeconometric panel data can lead to severely biased statistical results. Moreoverwe use our method to validate some of the most popular hypotheses on the performance of privateinvestors.