Latent class models with persistence in regime changes: a distributed lag analysis[Modèles de classes latentes avec persistance dans les changements de régime : une analyse des retards distribués]

One of the approaches to address the issue of production heterogeneity is to use latent class models. In most of these models, class membership either does not vary or might change freely over time. While the first assumption becomes increasingly untenable as the number of observed periods becomes larger, the second assumption is difficult to justify if important factors exist that prevent firms from switching classes back and forth several times. The present paper aims to develop a latent class model that allows firms to change from one class to another over time while permitting some degree of persistence in class membership. Our model can be used in settings with more than two classes and estimated using unbalanced panel datasets. An application of the empirical model in the context of dairy farm intensification is also provided. We find evidence of moderate resistance to replacing one milk production system with another in this sector, especially for small farms. Despite this, the standard latent class model performs reasonably well in terms of class-membership probabilities and temporal patterns.