Hedonic Regressions, Matched Models and Economic Theory

Quality adjustment of price indexes affects the analysis of many sensitive economic issues, such as real growth, productivity, international competitiveness, real wages, per-capita consumption and poverty, other than inflation. Hedonic methods are often recommended and increasingly used in the compilation of consumer price indexes. Nevertheless many official statistical agencies continue adopting traditional methods considering only the dynamics of prices of products matching in two adjacent periods of time. Indeed, a number of studies have even recently remarked that hedonic methods sometimes provide results very similar to the traditional matching models approach, particularly when models included in price index sample are replenished frequently. This paper briefly surveys the economic theory behind hedonic and traditional quality adjustment methods, and demonstrates that average price changes estimated by hedonic regressions differ from matched models estimation only because of the sum of regression residuals associated to disappearing and new models included in the sample. Thus, hedonic regressions including among the explanatory variables some indicators of the novelty and oldness of models provide exactly the same results of traditional methods. This fact casts some doubt on the overall effectiveness of hedonic methods in quality adjustment. The paper also focuses on that some economic and statistical hypotheses underlying hedonic methods possibly conflict with the assumptions and practices embodied in compiling the harmonised index of consumer prices for European countries.