ABSTRACT
In this paper, we examine the CPI (consumer price index) measurement errors under the first wave of the COVID-19 spread in Japan. To address this question, we construct high-frequency quality-adjusted price indices by employing daily scanner data from retail stores in Tokyo. We demonstrate the importance of using price data with the wide-ranging coverage of products and outlets by making explicit adjustments for temporary sales effects and retail service quality in examining the retail price dynamics under the COVID-19 pandemic as the voluntary lockdown constrained household purchasing behavior. Note that the sources of the CPI measurement errors under the COVID-19 pandemic differ significantly from those in the US, observed as wide-ranging and long-lasting stockouts. We show that downward bias, not upward bias generally advocated, was observed during the first wave of the COVID-19 spread in Japan. The magnitude of the downward bias is estimated at from −0.6 to −0.3 points on the CPI for food less perishables and eating out on the basis of cumulative changes from January 2020 to June. The contribution of the estimates to the overall CPI is −0.3% to −0.15% points on an annualized basis, considering that the estimation covers half-year and the weights are about a quarter of the overall CPI. The magnitude of measurement errors is deemed limited, and the overall trend of the CPI remains unchanged even after incorporating the estimated downward bias. It should be noted that this downward bias arises mainly from the "one-specification-for-one-item” policy by weakening the price representativeness in the Japanese CPI.