Online Appendix: Fungibility and Consumer Choice∗

Online appendix figure I shows the dynamics of the effect of a change in the price of regular gasoline on the share of gasoline that is regular grade. The plot shows that a permanent increase in the regular price would increase the share of regular gasoline for at least six months. Online appendix figure II shows the distribution across households of the correlation between buying regular gasoline and the price of regular gasoline. This correlation is positive for 59 percent of households. Online appendix figure III shows that the premium-regular price gap tends to compress when the regulargrade price rises. Online appendix figure IV shows evidence on the dynamics of this compression. After an increase in the regular price, the gap in prices falls, but the decrease dissipates within two months. Online appendix figure V illustrates the findings of our placebo analysis. In each panel we estimate a logit model with a product-specific constant, allowing the marginal utility of money to depend on household income and to vary flexibly across weeks in the data. Using this model we can calculate the household income implied by purchasing behavior in each week: the household income level that would rationalize the average household’s behavior in a given week with the model estimated in a baseline week. For gasoline, the implied income series moves inversely with prices. This is simply another representation of our main finding: as the gas price rises, households act far poorer than can be reconciled with the plausible income effect of gasoline. For orange juice and milk purchases, we do not observe such behavior. Indeed, during the large fall in gasoline prices in the autumn of 2008, households act, if anything, slightly poorer when choosing orange juice or milk brands, consistent with the onset of the financial crisis but not with their behavior at the pump. ∗E-mail: [email protected], [email protected].