Abstract
In finance and economics, predictive regression models are widely used. It is known that the limit distributions of their least squares estimators are nonstandard, and depend on the properties of the predictors. In this paper, we consider the unified confidence region construction of a predictive regression model by using empirical likelihood. It turns out that the resulting statistic has an asymptotical chi-squared distribution regardless of the predictor being stationary or non-stationary. Simulations are also conducted to illustrate its finite sample performance.
Acknowledgments
We thank Professor Liang Peng for simulating this research. Our thanks also go to the reviewer’s comments, which have led to many improvements in this paper.