THE FINAL REPORT of the Advisory Commission to Study the ConsumerPrice Index (CPI) represents the most influential critique of the CPI in decades. I This report, in conjunction with a number of other reviews of CPI bias, has focused the attention of policymakers and economists on the limitations of price index numbers, in particular, and of other measures of economic activity, more generally.2 The commission's estimate of overall bias in the CPI is 1. 1 percent per year, of which 0.4 percent is attributed to the failure of the fixed weight index to account for consumer substitution as relative prices change, 0. 1 percent is attributed to failure to account for discount stores and other innovations in retailing, and 0.6 percent is attributed to inadequate measurement of improvements in quality and of new goods. In contrast to the commission's estimates of substitution bias, which have been relatively uncon-The authors thank troversial, the estimates of quality and new goods bias have been criticized by several economists.3There are two general categories of quality errors: failure to detect a change in quality and failure to make the appropriate adjustment for a change that has been detected. The new products bias can also be considered in two forms: failure to include new products in the sample without long lags and failure to include the consumer surplus generated by a new product.4 In contrast to the quality bias, which can, in principle, go in either direction, the new product bias is theoretically known to be an upward bias (although it may be offset to some extent by the downward bias that occurs when a product disappears).Many discussions of quality bias have begun with the premise that much quality change goes undetected, that such quality change is predominantly improvement, and that the result is an upward bias of the index. We concur with Jack Triplett that this is of doubtful validity as a general proposition, although it may hold true in specific cases.5 The data collectors and commodity analysts at the Bureau of Labor Statistics 3. suggest that the quality bias estimate may be too large, whereas Diewert (1997, p. 95) describes the commission's estimates as "perhaps a bit conservative. " 4. These are not actually separate biases, since the consumer surplus generated by the new good should, in principle, include the value generated by any price decline early in the product's life cycle. The case of the new good may be treated in the theory of the cost of living index by using the reservation price of the item; that is, the price at which the consumer's demand for the item is just equal to zero. (See Hicks, 1940; Rothbarth, 1941; and Hausman, 1997.) Let P,,, represent the price of the new good after introduction and P;* ,, represent its reservation price before introduction. Let the prices of the other n -1 goods be p, = (P,,, P,2 . P, I, ), and similarly forp, . The constant-utility cost of living index, I, evaluated at the period t level of expenditures is where the expenditure function gives the mini...