Across-Regime Correlation in a Switching Regression Model: A FIML Approach

A Full Information Maximum Likelihood (FIML) procedure to estimate a two-regime endogenous switching model for cross-sectional data is here provided. The across-regime correlation of error terms, generally unidentified in the most widely used models, is also directly estimated. The results of a Monte Carlo experiment, assuming normally distributed error terms, confirm consistency of the estimated across-regime correlation. As an empirical application, both individual wage and reservation wage equations are estimated simultaneously using a sample of Italian graduates.