The Dog and the Straw Man: Response to “Dividend Growth Does Not Help Predict Returns Compared to Likelihood-Based Tests: An Anatomy of the Dog”

I respond to Hjalmarsson and Kiss (2021). Cochrane (2008) does not advocate a “data-dependent null,” a testing methodology in which one always uses a sample value as null hypothesis for a test statistic. In fact, I calculate sampling distributions for a wide variety of fixed null values of the parameters. How best to impose prior information that the dividend yield is stationary remains an important question highlighted by both papers.