Estimations of generic drug entry in Brazil using count versus ordered models

Two major changes in Brazilian legislation during the 1990s reshaped the local pharmaceutical industry: the ratification of the TRIPS agreement including a provision for pipeline inventions in 1996, and a Generic Drug Act in 1999, which introduced bioequivalence tests and facilitated generic drugs' substitution for the pioneer drugs at dispensing. Genuine generic drug entry may be dated back to 2000, when the first applications were approved. Price controls were gradually resumed in the turn of the century. The present article estimates entry of generic versions of off-patent drugs into various therapeutic classes using both count data and ordered multinomial models. Results point out that a simple Poisson model fits the data poorly, calling for further modelling of overdispersion or of excess zeros by applying Negative Binomial and zero-inflated count models. Ordered models seem to provide a worse fit, even though the explaining variables display the same pattern of significance and signs. Most of theexplaining variables utilized are significant, in particular a proxy for potential market (lagged revenues of the drug), market concentration, and age of the pioneer drug. Multiple brands (rather than simply own-generics) and evergreening strategies seem to be effective in deterring entry.