CCP Estimation of Dynamic Discrete Choice Demand Models with Segment Level Data and Continuous Unobserved Heterogeneity: Rethinking EV Subsidies vs. Infrastructure.

This article presents a novel methodology for estimating dynamic discrete choice demand models for durable goods using aggregate sales data segmented by consumer groups, addressing challenges posed by continuous unobserved consumer heterogeneity, unobserved product characteristics, and nonrandom attrition (dynamic selection). The approach leverages conditional choice probabilities (CCPs) as functions of unobserved heterogeneity across multiple consumer groups within the same market, enabling identification and estimation without resorting to value function approximations or state space reduction. Empirically, the method is applied to estimate consumer demand for electric vehicles (EVs) in Washington state from 2016 to 2019, revealing heterogeneous price sensitivities across income groups and the significant influence of factors such as electric range, charging infrastructure, and gasoline prices. Counterfactual policy simulations suggest that incentivizing EV adoption through expanded Level 3 charging infrastructure yields greater increases in battery electric vehicle (BEV) sales and larger reductions in CO2 emissions than consumer tax credits, while a tax credit based on electric range rather than battery size also improves environmental outcomes. The methodology's reliance on group-level market share data rather than individual-level panel data makes it broadly applicable to markets with many differentiated durable products and aligns with emerging data privacy trends.