Georgiev, I., Harvey, D.I., Leybourne, S. J. and Taylor, A.M.R.
Predictive regression is a widely used tool in applied finance and economics, and it is common to test for such predictability, for example testing whether future (excess) stock returns are predictable by current information, such as the dividend yield or the term structure of interest rates. Due to the fact that the predictor variables are often highly persistent processes, it is possible that recently proposed tests for predictability can identify a spurious predictor variable whenever an unincluded persistent variable is present in the process driving the predicted variable.
In this publication, Iliyan Georgiev, David Harvey, Stephen Leybourne and Robert Taylor demonstrate theoretically and by means of simulation the potential for spurious predictability findings to arise when using standard predictive regression tests. To guard against this possibility, they propose a diagnostic test for predictive regression invalidity based on a stationarity testing approach. In order to allow for an unknown degree of persistence in the putative and unincluded predictors, and to allow for both conditional and unconditional heteroskedasticity in the data, a fixed regressor wild bootstrap test procedure is proposed and its asymptotic validity established. The behaviour of the proposed test is examined using Monte Carlo simulations, and the test is applied to data on US stock returns.
A bootstrap stationarity test for predictive regression invalidity, Journal of Business and Economic Statistics (2018, forthcoming) https://doi.org/10.1080/07350015.2017.1385467
View all featured publications
Posted on Monday 26th February 2018