Brand Dummy Variables in Hedonic Regressions: A Study Using Stereo Receiver Scanner Data∗

Time and brand dummy variables are typically used as controls for omitted variables. The operating assumption made is that time dummies control for unobserved factors affecting price and brand dummies control for unobserved factors affecting quality. We investigate the validity of the latter assumption by explicitly controlling for supply and demand factors in hedonic regressions for stereo receiver scanner data. We find that the characteristic and brand coefficients remain roughly unchanged if industry level supply and demand variables are used in place of time dummies. Interacting industry level variables with brand variables to produce manufacturer level controls for supply and demand, although the coefficients for physical product characteristics are not significantly changed, the coefficients for brand variables become quite unstable. We conclude that brand dummy variables are poorly specified and their use for quality adjustment may introduce errors into price index calculations.