History Dependence in Drug Demand: Identification and Implications for Entry Incentives

Abstract

Abstract I use temporal discontinuities in choice sets to identify history dependence in drug demand, finding large and long-term effects. Quasi-randomly assigning a patient to a drug today increases the probability she is taking the same drug four years later by 54 percentage points. History dependence is stronger in patients taking multiple medications and weaker in drugs that are significantly more effective than substitutes. It is also weaker in generics and line extensions, driven by switching from the reference branded drug. I use a pair-specific switching-cost model to capture these patterns and provide suggestive evidence that they affect manufacturers’ entry timing incentives.