Hedonic regressions: An application to video recorders using scanner data

This study provides results for an hedonic regression for VCRs in the UK using relatively exhaustive scanner (bar-code) data. Specific issues addressed are sample selectivity bias, given a substantial proportion of the models have limited sales, incorporation of a proxy for the price-cost margin, the bundling of characteristics in new models and the incorporation of models of different vintages coexisting in the market. In particular it is shown that current formulations of vintage effects could prove misleading and more complex non-linear representations specific to individual makes are found to be quite successful. In addition an effect specific to the season (month) of launch is identified.