In explaining monetary policy to the public, central banks employ price indexes that cover goods and services consumed by households, especially the consumer price index (CPI). The CPI, however, is generally thought to overstate changes in the true cost of living. Accurate price measurement is particularly important in a country like Japan where CPI inflation is running close to zero.
Shiratsuka (1999) first showed a point estimate of the upward bias in the Japanese 1990 base CPI of 0.9 percentage point. Thereafter, the Japanese CPI introduced various revisions to improve its accuracy, including the application of hedonic methods for PCs and digital cameras. Shiratsuka (2006) made a follow-up assessment, suggesting that the upward bias had substantially narrowed in the 2000 base index.1
Imai and Watanabe (2014) make an important contribution to enhancing our understanding of the measurement problem of the Japanese CPI, particularly in relation to quality adjustments using unit-price changes. By employing daily scanner data for more than 300,000 products sold at about 200 supermarkets from 2000 to 2012, Imai and Watanabe show that Japanese consumers tend to respond in parallel to both size/weight changes and price changes. Imai and Watanabe thus argue that the Japanese CPI may be downwardly biased rather than upwardly biased, since appropriate adjustments are not made for the quality downgrade associated with product downsizing.
I would draw different implications from the empirical findings in Imai and Watanabe (2014). It is possible that the Japanese CPI is downwardly biased because it fails to reflect hidden price changes in product downsizing, as they discussed. If this is the case, however, such a bias would be introduced not through the quality adjustment method, but through the price survey method of “one-specification for one-item.” In addition, the direction of the bias is uncertain, depending on economic conditions.
The “one-specification for one-item” policy specifies a single and generally the most popular specification for each item, and continuously surveys the prices of this item. That policy has an advantage in that monitoring the specification changes for the items surveyed, including product downsizing/upsizing, becomes easier. The policy, however, has a disadvantage in that the specified item is not necessarily representative of all items in an expenditure category.
Figure 1 plots the chocolate CPI, which has a very detailed specification of products, like “Meiji milk chocolate” or “Lotte Ghana milk chocolate.” Imai and Watanabe (2014) report that per-gram prices for “Meiji milk chocolate” increased 27% from May 2008 to October 2012, which almost corresponds to the 26% increase in the chocolate CPI during the corresponding period. That suggests that unit-price adjustments are properly implemented in surveying chocolate prices. It is still unclear whether the current chocolate CPI properly reflects the overall movements in chocolate prices.
In closing, I should emphasize that measurement errors in the CPI are unavoidable to some extent, since the CPI is not based on census data, but rather on surveys with many statistical adjustments. It is crucially important to examine constantly whether the CPI is reliable enough. Statistical agencies are then required to allocate their limited resources efficiently to create better statistics. In the context of Imai and Watanabe (2014), I suggest exploring how to improve price survey methods so as to reflect the overall price movements of each product category in a more appropriate manner.