A switching model with flexible threshold variable: With an application to nonlinear dynamics in stock returns

This paper proposes an extension to threshold-type switching models that lets the threshold variable be a linear combination of exogenous variables with unknown coefficients. An algorithm to estimate the model's parameters by least squares is provided and the validity of the methodological framework is assessed by a Monte Carlo study. The empirical usefulness of the proposed specification is illustrated by an application to US stock returns.