Jeremy Piger (University of Oregon) presented new work analyzing sample properties and estimation procedures for N-State Markov Switching models. Typically the correlation between the error terms and the states is assumed to be zero, but in some cases this may not be natural. For instance if shocks to the economy drive business cycles then we should allow the transition process to be affected by the shocks themselves. A small body of work has examined endogenous Markov-switching but has focused on estimation when the number of states is small i.e. N=2. Consider the case for N>2 allows the econometrician greater flexibility in estimating business cycle models.

Slides for the presentation can be found here.