How to avoid incorrect inferences (while gaining correct ones) in dynamic models

Abstract

Abstract A flurry of current interest in time series has focused on clarifying equation balance, fractional integration, and cointegration testing. Despite this, a number of recent suggestions may continue to lead scholars toward incorrect inferences. In this comment, I investigate the likelihood of drawing both correct and incorrect inferences under a variety of stationary and non-stationary data-generating processes. I extend previous work in this area by focusing on both short- and long-run effects using several popular model specifications. Given these findings, I conclude by offering a variety of recommendations to practitioners about how they can best specify their model.