We investigate several measures for the rates of change in price levels at the household level based on daily scanner data. The major findings are as follows: (1) when the data frequency is annual, the price index based on the actual transaction prices is more accurate than the traditional index based on information on some representative commodities only; (2) large heterogeneity in the rate of changes of price level across households are observed regardless of the differences in data construction procedure; (3) heterogeneity in the rate of price level changes across different age groups is negligible, while rich households tend to face higher rate of price change; (4) finally, about one percent difference between Laspeyres and Paasche indices at annual level is observed, indicating significantly lower substitution across commodities within each category.