Cost & opportunities in misspecified predictive regressions

This page contains research and presentation material on the topic of quadratic errors in predictive regressions.
This is joint work with Romain Deguest.

---

Files

• The theoretical contributions of the paper are decomposed into:
  • Comparisons of methods: tests between integral formulae and Monte-Carlo simulations. This requires R functions and C functions.
  • Sensitivity analysis: impact of persistence, forecasting horizon and sample size on loss / R^2.
    
  • Univariate learning: comparison between PRs and univariate learning, that is, regression on lagged values of the process.
• All our empirical results are based on the following RMarkdown notebook and the function file. The data can be found here (.RData format). It is based on that of Welch and Goyal (2008) plus one variable from Novy-Marx (2014).
  The source for the multiple analyses are the following:
  • impact of rho_y
    
  • value of economic constraints
    
  • aggregate variance forecasting

References:
- Welch, I., & Goyal, A. (2008). A comprehensive look at the empirical performance of equity premium prediction. Review of Financial Studies, 21(4), 1455-1508.
- Novy-Marx, R. (2014). Predicting anomaly performance with politics, the weather, global warming, sunspots, and the stars. Journal of Financial Economics, 112(2), 137-146.

---

---

DISCLAIMER: the data and code are meant for research and pedagogical use only.

---