A Comparison of Consumption-Related Statistics

This paper provides an outline of the Family Income and Expenditure Survey (FIES), the National Survey of Family Income and Expenditure (NSFIE), the Comprehensive Survey of Living Conditions (CSLC), and the Survey of Household Economy (SHE) and compares the results of these major consumption-related governmental statistical surveys. Trends in the FIES figure for total consumption expenditure, which is the item of consumer data that attracts the greatest attention, are broadly consistent with those in the CSLC and the NSFIE, but clearly lower than those in the SHE. Accordingly, focusing on the difference between the FIES and the SHE, this paper examines the causes of this disparity. The difference between trends in the two survey can be broken down into the portion relating to goods and services specified in the SHE and the portion relating to the other goods and services, so this paper examined these portions separately. The disparity arising in goods and services other than those specified in the SHE would appear to be caused by the design of the survey; that is, the panel structure of the two surveys. The FIES has a gsurvey fatigue bias, h with which the expenditures becomes under-reported with repetition of the survey since households become tired of keeping records, while the SHE has an gattrition bias, h with which those who are willing to complete the survey are over-sampled since non-cooperative households drop from the survey. However, once we control these biases, the results are almost consistent in the both surveys. As for the disparity arising in goods and services specified in the SHE, the under-reporting of durables and other high-priced items in the FIES would play an important role. Whereas the FIES uses a free-entry format, or after-code method, in which respondents record the amount of expenditure in dairy, the SHE uses a pre-coded questionnaire format. It would appear that use of the free-entry format gives rise to substantial omissions, particularly of expenditure on high-priced items. Overall, it became clear that the difference between the two surveys stems from the survey methods used. However, it is neither easy nor necessarily desirable to change the method used, because the choice of survey method has an important role to play, both in practical and academic terms. In this sense, it is desirable for users to fully understand the distinctive characteristics of each survey and to make appropriate corrections for them before use.