Research

• Vertical Integration and Market Structure: Evidence from Tesla's Charging Network (JMP)

slides   slides (old)  paper  

Abstract:  Theory offers ambiguous predictions on the welfare effects of vertical integration, and empirical evidence remains limited. I study Tesla's exclusive Supercharger network in the UK electric vehicle (EV) market, estimating a structural model of demand and supply with three channels: price competition, Tesla's charging investment, and incumbents' EV entry incentives, which account for the cannibalisation of their internal combustion engine profits. Demand is estimated with a new two-level inversion method that jointly uses national product shares and local technology shares, allowing local charging access to be instrumented. Opening the network with its size held fixed would lower Tesla's prices and raise consumer surplus by £473 million per year. However, once Tesla's investment is allowed to adjust, the predicted gain falls to £9 million. Exclusivity discourages rival entry, but only modestly. Replicating Tesla's contribution to EV adoption through public charging subsidies would cost roughly  twice as much as the actual spending by 2022. Overall, in this early-stage market, the investment benefits of exclusivity largely offset its static costs. 

• Estimating Cross-market Interactions in Demand 

 slides  paper  slides (old)  

Abstract:  Discrete choice models are widely used to estimate demand in differentiated product markets, and a growing literature extends this framework to study cross-market interactions. In many applications, product market shares are observed at an aggregate level while interacting variables vary locally, creating an aggregation mismatch that complicates identification and raises endogeneity concerns. I develop a two-level inversion approach that jointly inverts aggregate product-level shares and local group-level shares, leveraging the insight that cross market interactions operate at the level of product groups. This procedure recovers location specific unobserved demand and enables instrumental variables estimation without requiring highly disaggregated product-level data. I apply the method to the UK electric vehicle market from 2014 to 2023, where model level shares are national but charging infrastructure varies across local authorities. The results reveal positive but heterogeneous complementarities between electric vehicles and charging station and show that failing to instrument leads to substantially biased estimates. 

• The Evolution of Aggregate Markups, with Steve Bond 

Abstract:  Is there evidence of rising markups across countries? In this paper, we implement a new method that leverages the relationship between output returns to scale, revenue returns to scale, and firms’ profit-maximizing behavior. We compute upper and lower bounds for revenue elasticities using the sales-weighted arithmetic, geometric, and harmonic means of observed revenue shares. Applying this framework, we obtain semi-parametric bounds for aggregate markups for G7 countries, excluding the United States, under the assumption of constant returns to scale in production technology. Our findings suggest that the increase in aggregate markups is considerably more modest than previously reported and is accompanied by a slight rise in markup dispersion. We also implement a hybrid markup estimation approach, which further supports the conclusion that aggregate markups have risen only modestly, or not at all, across our sample of countries.