The idea that products can act as complements or substitutes is commonplace in economics. What is less well understood is how households trade off particular attributes of one product with the attributes of another.… One immediate implication of this “attribute substitution” is that the demand for products within a household will not be independent. A second implication is that attribute substitution may influence the efficiency or cost effectiveness of policies that influence the attributes of products, such as energy efficiency standards. Few papers have been able to empirically measure these forces.

… Our setting, a household’s demand for vehicles, allows us to overcome similar identification challenges. Attribute substitution may manifest itself in a number of dimensions. For example, households may prefer to have one SUV and one sedan, or one powerful vehicle and one fuel-efficient vehicle. Identification in this case is challenging because households may also fall into certain household “types.” Just as Anderson et al. (2015) and Mannering and Winston (1985) show that some households have a preference for a certain brand of vehicle, some households may also have a preference for certain attributes. In such instances, if households have a preference for, say, horsepower or fuel economy, the attribute will be present in high or low levels across all goods in the portfolio.

—Archsmith et al.

Using a rich data set that enabled them to track households over a number of vehicle replacement and purchase decisions, and through the use of an instrumental variables strategy, the researchers estimated how the fuel economy of a newly added vehicle depends on the fuel economy of the vehicle already owned by the household. The results have implications for policies that improve the fuel economy of light-duty vehicles, such as fuel economy standards.

A central objective of my research as an environmental economist is to identify ways to fight global climate change. Unintended consequences like this need to be taken into account when making policy. On average, fuel economy standards are putting more fuel-efficient cars in households. That can be good if it reduces gasoline use. But if it causes people to buy a bigger, less fuel-efficient second car to compensate, this unintended effect will erode the intended goals of the policy.

—author David Rapson, UC Davis

Broadly, they found evidence that households substitute attributes across the vehicle portfolio. Increases in the fuel economy of the kept car reduced the probability the household purchases a car in the lower quartile of gallons per mile (the highest fuel economy quartile); rather such increases raised the probability the household buys a car in the upper quartile (the lowest fuel economy quartile).

Attribute substitution erodes over 60% of the fuel savings from the fuel economy increase of the kept vehicle on net after accounting for all of these factors. As a specific example, consider a 10 percent increase in fuel economy from the average vehicle in our sample. Given the average miles driven (688 per vehicle in our sample), this 10 percent fuel economy increase would directly lead to a 69 gallon decrease in annual fuel consumption. However, due to attribute substitution the next vehicle the household purchases will be less fuel efficient than it otherwise would have been. This decrease in fuel economy of the newly-purchased vehicle reduces the fuel savings from our thought experiment to 40 gallons, holding usage of the two vehicles constant. But we also find significant changes in usage patterns that further reduce the net fuel savings.

Mileage of the kept car increases significantly. A large fraction of this increase is due to shifts in miles traveled from the now less fuel efficient purchased vehicle; however, we also find a net increase in overall mileage across the two-vehicle portfolio. Accounting for all of the changes, the net savings of the exogenous increase in fuel economy falls from the naive estimate of 68 gallons to 24 or 27 gallons, depending on whether the initial vehicle was the most fuel efficient vehicle in the household.

—Archsmith et al.

The research was supported by funding from the California Air Resources Board.

Resources

• James Archsmith, Kenneth Gillingham, Christopher R. Knittel, and David Rapson (2017) “Attribute Substitution in Household Vehicle Portfolios” DEEP WP 018