Title: IVX-based panel predictive regressions for stock returns
Authors: Christoph Hanck - Universität Duisburg-Essen (Germany) [presenting]
Matei Demetrescu - University of Kiel (Germany)
Abstract: New panel tests are proposed for predictability of, e.g., stock returns through regressors which are allowed to be persistent or stationary. The panel units may be heterogeneous and cross-sectionally dependent. Building on previous work, we employ IVX instruments to this end, which are generated within the panel and require no outside exogeneity assumptions. We show the test statistics to follow standard $\chi^2$ distributions under the null of no predictability as the number of time series observations $T$ and panel units $N$ jointly go to infinity. Simulations indicate good size and power in panels of size commonly encountered in empirical practice.