Lower-Level Substitution Bias in the Japanese Consumer Price Index: Evidence from Government Micro Data

This paper explores measurement errors in the Japanese Consumer Price Index (CPI) stemming from lower-level substitution within items. The CPI is widely used as a measure for inflation or the cost of living. The Japanese CPI employs the one-specification for one-item policy in surveying individual prices. The policy specifies a few most popular specifications for each item and continuously surveys their prices at specific outlets. As a result, the price homogeneity is generally maintained, limiting the impact of the differences in the elementary aggregation formulas, which corresponds to the narrow definition of the lower-level substitution bias. In contrast, the price representativeness becomes difficult to be maintained for highly heterogeneous and differentiated products. That is another aspect of the lower-level substitution bias particular to the Japanese CPI, encompassed by the broad definition of the lower-level substitution bias. However, quantitative assessments on the lower-level substitution bias in the Japanese CPI are very limited since the detailed CPI data at individual price observations was not readily available for a long time. This paper is the first trial on a quantitative assessment of the lower-level substitution bias using the micro data for the Retail Price Survey (RPS), which is the primary source data for the Japanese CPI. Empirical evidence confirms that the lower-level substitution bias in the Japanese CPI differs from that for the U.S. CPI. On the one hand, the one-specification for one-item policy in the price survey succeeds in keeping price observations homogeneous, limiting the elementary aggregation bias. On the other hand, the policy also weakens price representativeness, which requires additional quantitative assessments using alternative data sources, such as scanner data and web-scraping data.