Estimating Daily Inflation Using Scanner Data: A Progress Report

We construct a T¨ornqvist daily price index using Japanese point of sale (POS) scanner data spanning from 1988 to 2013. We find the following. First, the POS based inflation rate tends to be about 0.5 percentage points lower than the CPI inflation rate, although the difference between the two varies over time. Second, the difference between the two measures is greatest from 1992 to 1994, when, following the burst of bubble economy in 1991, the POS inflation rate drops rapidly and turns negative in June 1992, while the CPI inflation rate remains positive until summer 1994. Third, the standard deviation of daily POS inflation is 1.1 percent compared to a standard deviation for the monthly change in the CPI of 0.2 percent, indicating that daily POS inflation is much more volatile, mainly due to frequent switching between regular and sale prices. We show that the volatility in daily inflation can be reduced by more than 2daily inflation rate 0 percent by trimming the tails of product-level price change distributions. Finally, if we measure price changes from one day to the next and construct a chained T¨ornqvist index, a strong chain drift arises so that the chained price index falls to 10-10 of the base value over the 25-year sample period, which is equivalent to an annual deflation rate of 60 percent. We provide evidence suggesting that one source of the chain drift is fluctuations in sales quantity before, during, and after a sale period.