ABSTRACT
We employ a new panel-based testing procedure that is robust to the uncertain persistence of regressors, time-varying volatility and cross-sectional error dependence in studying the predictive dynamics between conventional US monetary policy surprises and firm-level stock returns. We find that accounting for cross-sectional dependence by means of (estimated) factors considerably alters the predictive significance of monetary policy surprises depending on the sample period being studied. Concretely, during the period 1990–2000, monetary policy has no influence on future stock returns when cross-sectional dependence is accounted for by means of common factor augmentation. By contrast, the predictive power of monetary policy is even boosted when introducing common factors into the model when the period of analysis covers 2002–2007.
Acknowledgements
The author gratefully acknowledges the support of the Deutsche Forschungsgemeinschaft (DFG) grant DE 1617/4-1. The author would like to thank Kai Carstensen, Matei Demetrescu, Maik Wolters and the participants of the 2015 German Statistical Week and the 4th Joint Statistical Meeting of the Deutsche Arbeitsgemeinschaft for very helpful comments and suggestions.
Disclosure statement
No potential conflict of interest was reported by the author.
Notes
1 See Breitung and Demetrescu (Citation2015) for detailed proofs and simulation results.
2 We set a moderate persistence parameter at , although using alternate values ranging from only resulted in marginal differences in the estimated coefficients.
3 We omit the 2001 period where the Fed successively reduced the Fed funds target rate due to the 11 September 2001 event.
4 We also employ the cross-sectional average of stock returns as an alternative estimator of and find that the general findings remain the same.