Abstract
We propose a novel model for household choice of car and driving. The model extends existing discrete-continuous models of single car choice to allow for zero, one, or two cars. The key innovation is that complementarities arise from the driving utility, which households take into consideration when choosing their car portfolio. We use administrative micro data from Norway where data on driving in all cars is linked to detailed household demographics, including individual spousal work distances. Moreover, Norwegian policy incentives for electric vehicle adoption include exemption from road tolls, implying that there is substantial cross-sectional policy variation in monetary incentives. We show that portfolio considerations play a crucial role in understanding EV adoption that predictions from single-car models cannot explain and we document the implications for policy evaluation.