Inference in Predictive Regression Models with Persistent Regressors
This thesis comprises three papers on predictive regressions with persistent regressors. Standard approaches such as OLS suffer from second-order bias and corresponding test statistics show nonstandard limiting distributions in the presence of endogenous and persistent regressors. I discuss three different approaches to deal with this issue as well as further pitfalls such as nonlinearity and model uncertainty within this thesis. Two M-based tests are aggregated in Chapter 2, yielding a standard distribution for stable as well as persistent regressors. The setup is extended to nonlinear predictability and the performance of nonparametric and IV-based tests are compared in Chapter 3. Finally, the focus in Chapter 4 lies on multivariate predictions of different tree-based methods.