Model selection in under-specified equations facing breaks

Abstract

When a model under-specifies the data generation process, model selection can improve over estimating a prior specification, especially if location shifts occur. Impulse-indicator saturation (IIS) can 'correct' non-constant intercepts induced by location shifts in omitted variables, which leave slope parameters unaltered even when correlated with included variables. Location shifts in included variables induce changes in estimated slopes when there are correlated omitted variables. IIS helps mitigate the adverse impacts of induced location shifts on non-constant intercepts and estimated standard errors, and can provide an automatic intercept correction to improve forecasts following location shifts.