This dissertation empirically studies demand complementarities in automobile purchases using newly collected Japanese household-level panel data, Keio Household Panel Survey. It is motivated by the observation that approximately one third of Japanese households own more than one automobile and they tend to hold particular combinations of products, which cannot be captured by the prevalent single choice model in this literature. The dissertation develops a structural model where consumers can purchase up to two differentiated products, where I allow for flexible complementarities which depend on consumer attributes and product characteristics. In the model, firms set the prices for their products, given other firms’ pricing strategies and consumer demand. I then estimate the model using two types of data: micro-level data on household automobile purchasing decisions and macro-level data on market share. My estimates suggest that strong complementarities arise when households purchase a combination of one small automobile and one regular-sized automobile, or one small automobile and one minivan as their portfolio. The estimates also indicate that households are more likely to purchase two automobiles as their numbers of earners increase or if they are located in rural areas. Ignoring such portfolio effects would lead to biased counterfactual analyses. For example, my results suggest that a policy proposal of repealing the current tax subsidies for eco-friendly small automobiles would decrease the demand for those automobiles by 12%, which is less than the 17% drop predicted by a standard single discrete choice model. Similarly, model simulations indicate that the presence of positive portfolio effects significantly influences firms’ pricing behavior: firms potentially have incentive to use a mixed bundling strategy when the number of products and firms in the market is small.