Revisiting taste change in cost-of-living measurement

Robert S. Martin

Abstract

Several recent papers aim to incorporate the pure effects of changing tastes into cost- of-living indexes (COLI), arguing that traditional consumer price indexes otherwise suffer from bias. Their proposals, however, reflect a different conceptual target (an unconditional COLI) from a consumer price index (a conditional COLI), and a strong implicit assumption (cardinal utility). Using similar scanner data to Redding and Weinstein (2018), and the same constant elasticity of substitution model, my analysis suggests that for food products, the purported contributions of tastes can dominate those of prices in an unconditional COLI estimate, whereas the choice of reference taste vector has a relatively small impact on a conditional COLI. Using CPI elementary item- area indexes, I find that differences between conditional COLI vary over time, depend on the substitution elasticity parameter chosen, and tend to accumulate when the index is chained.