brand dummy variables in hedonic regressions a study using stereo receiver scanner data teague ruder and ted to us bureau of labor statistics january 2004 abstract time are typically used as controls for omitted the operating assumption made is that dummies control unobserved factors affecting price dummiescontrolforunobservedfactorsaffectingquality weinvesti gate validity latter by explicitly controlling supply demand scannerdata wefindthatthecharacteristicandbrandcoefficientsre main roughly unchanged if industry level vari ables place interacting with produce manufacturer con trols although coefficients physical product characteristics not significantly changed become quite unstable we conclude poorly specified their use quality ad justment may introduce errors into index calculations keywords measurement market structure jel classification codes l1 d1 l6 thank dan sichel seminar participants from 2003 nber criw summer institute views opinions contained herein be construed
