This question is asked in a 2022 NBER working paper by Pierre Dubois, Ashwin Gandhi, and Shoshana Wasserman.
Methodology: Demand Side Model
The authors used IQVIA data on revenues and volumes of drugs sold to hospitals in the US, Canada, France, Germany, Italy, Spain and the UK markets from 2002 to 2013. The authors define a market based on the anatomic therapeutic and chemical classification (ATC-4) level, taking into account the possibility of drug alternatives to treat similar diseases. They then create a representative consumer demand model that estimates indirect utilities relative to the external good (i.e., no drug purchases). Drug characteristics include the drug’s molecule identifier, patent status, and generic status. Heterogeneity is modeled using random coefficients to account for differential utility of the product, differential disparities from higher prices, and differential preferences for brand versus generic drugs. The demand model is estimated using our demand model according to the standard BLP method with instrumental variables for prices. The specific instruments used are the number of products in the same ATC‑4 class, the number of generic and off-patent brands, and the number of countries where the drug is sold. This helps determine price as a function of competition because demand is presumably a function of price; The level of competition does not have a significant impact on demand except perhaps price and thus constitutes an appropriate instrument in this case. Although the authors have data from several countries, the demand model is estimated for the US and Canada.
Methodology: Supply Side Model
The authors model equilibrium pricing using the Nash bargaining model. In the model, companies maximize profits while government regulators maximize consumer welfare. The Canadian bargaining system is designed as a single payer entity; It is assumed that American negotiation and pricing occurs through Bertrand competition. Negotiations take place on a market-by-market basis (i.e., at the ATC-4 level).
Result
International reference pricing is likely to have no significant impact on US prices, but prices in Canada will increase significantly.
In the main specification, an international reference pricing policy where the price in the US cannot exceed that in Canada, Canadian prices become the effective price ceiling for similar drugs sold in the US. This barrier is imposed when companies negotiate prices with the regulator in Canada, but the impact on reducing expenses in the US is relatively small. Furthermore, we find that while baseline reference pricing slightly lowers prices in the US, prices in Canada increase dramatically because companies’ disagreements in negotiations tie payments to unrestricted US profits. Therefore, expenditure on pharmaceuticals increases significantly in Canada but there is no significant change in it in America.
What if the number of countries in the reference basket changes? Nevertheless, prices in the US have declined only marginally.
…referring to an index of countries or increasing the size of the country, both of which refer to lower equilibrium prices for US consumers. However, the price cuts are surprisingly small compared to the status quo price gap between the US and Canada. A major reason is that the US market is so unmatched in its size that even when referencing larger countries or more countries, profitability in the US market still drives negotiations in the referencing countries rather than the other way around.
You can read the full paper Here.
Appendix: The BLP Method Explained.
blp method (Berry–Levinsohn–Pecks) is a structural approach to estimating Random-coefficient discrete option demand model For differentiated products. It combines:
- A random-coefficient logit (or other different options) utility specification on products,
- A Nonlinear “inverse” From observed market shares to utilities, and then
- Instrumental Variables (IV) Estimating demand parameters, treating prices as endogenous
In practice, the BLP estimator is a nonlinear IV-GMM that:
- Allows flexible replacement patterns,
- allows preferences to be random, and
- Handles price endogeneity through instruments that alter prices (e.g., cost shifters, BLP-style instruments based on competitors’ characteristics)
