Panel data of household consumption generally exhibits large volatility and weak serial correlation, which is inconsistent with the standard household consumption model. This paper investigates whether (1) measurement errors and (2) transitory components in consumption data can explain the observed volatility of household consumption. Using several micro data including daily scanner data from Needs-Scan/Panel, we found that the short length of data collection period, that is, the existence of transitory consumption, is the main source of the observed large volatility. Compared to the aggregation period, the measurement errors play minor roles in the consumption volatility. Household consumption becomes close to a random walk only when we use quarterly or longer aggregation periods. We also found that not-perishable goods such as canned foods are close to an i.i.d. process regardless the aggregation length. Our findings suggest that we need to be very careful when testing consumption smoothing or random walk hypotheses with typical household-level panel data based on weekly or monthly surveys.
