Research

We examine whether international pharmaceutical prices are higher where demand is less price responsive. Using data from 63 countries over 2008–2020, we adapt double machine learning to estimate own- and rival-price elasticities that vary across countries, medicines, and years. We establish conditions under which a rich price-response specification approximates responses from identified logit-based demand models and characterize identification limits of unrestricted structural specifications. Estimated responsiveness varies substantially across markets and is not consistently captured by GDP per capita. Robustness checks against elasticities from structural models are presented. Higher branded-drug prices are associated with less responsive demand, consistent with inverse-elasticity ordering, while, as expected in a perfect competitive environment, generic products show no similarly stable relationship. A historical exercise motivated by U.S. Most-Favored-Nation pricing illustrates why these findings matter for policy. On the same sample, the median U.S.-to-reference price ratio is 6.95 using the lowest eligible foreign price, compared with 1.62 using a foreign benchmark that accounts for U.S. demand responsiveness and country characteristics. Our results show how directly estimated elasticities can inform international price benchmarks beyond comparisons based on income alone.

We examine the causal effect of market size on pharmaceutical innovation in a context where observed revenues are endogenous equilibrium outcomes jointly determined by demand, competition, and innovation. To address this endogeneity, we use major drug recalls as a novel instrument for expected market rewards, relying on their sharp and unanticipated nature at the therapeutic-class level.

Combining U.S. sales, recall, and clinical trial data at the relevant submarket level, we estimate the effect of market size on innovation within a control-function IV Poisson framework that addresses both idiosyncratic and heterogeneity-driven endogeneity. The identifying variation comes from recall-driven contractions in class-level revenues that persist beyond the recall year.

We find that a 10% increase in market size leads to approximately a 6% increase in clinical trial activity, an elasticity at the upper range of existing estimates. Importantly, when innovation is restricted to trials that result in marketed products, the estimated elasticity falls substantially. This finding suggests that standard measures of innovation effort overstate the responsiveness of commercially successful innovation to market-size incentives and underscores the importance of distinguishing between innovation effort and realized outcomes for both empirical measurement and policy design.

Using administrative data on the universe of inter-firm transactions in Spain, we show that firms learn to import from their domestic suppliers and customers. The identification exploits the panel structure of the data, the firm-time variation across import origins, and the network structure. We find evidence of both upstream and downstream network effects, even after accounting for sectoral and spatial spillovers. A 10 percentage point increase in the share of suppliers (customers) importing from a given region raises the probability of starting to import from it by 10.7% (19.0%). Connections with geographically distant domestic firms provide more useful information to start importing. Larger firms are more responsive to this information but less likely to disseminate it. Counterfactual simulations based on our estimates show that peer effects roughly double import entry over five years, and that an import-promotion policy induces an order of magnitude more entrants when directed at high-spillover firms than when assigned at random.

This work belongs to the strand of literature that combines machine learning, optimization, and econometrics. The aim is to optimize the data collection process in a specific statistical model, commonly used in econometrics, employing an optimization criterion inspired by machine learning, namely, the generalization error conditioned on the training input data.

Our paper presents a causal Machine Learning (ML) methodology to study the heterogeneous effects of economy-wide shocks and applies it to the impact of the COVID-19 crisis on exports. This method is applicable in scenarios where, due to the pervasive nature of the shock, it is difficult to identify a control group that is not affected by the shock and to determine ex ante differences in shock intensity across units. In particular, our study investigates the effectiveness of different machine learning techniques in predicting firms’ trade and, by building on recent developments in causal ML, these predictions are used to reconstruct the counterfactual distribution of firms’ trade under different COVID-19 scenarios and investigate the heterogeneity of treatment effects.

Advanced therapies represent a paradigm shift. They offer potentially curative solutions to life-threatening diseases through a single administration.