Our Academic research on different volatility regimes presents a methodology to use a regression estimated in a sample of stable volatility (i.e., without heteroskedasticity) to conduct inference in a subsequent, out-of-sample period of unstable volatility. Critical values for out-of-sample test of significance use the empirical distribution of the t-tests rather than large-sample distribution.
This paper was published in the Journal of Forensic Economics (2024) 31 (1): 5-25, DOI: 10.5085/JFE-480, and a copy of paper is available here.
