wordpress-themes.org wordpress themes wordpress themes

INTERTEMPORAL CHOICE AND THE CROSS-SECTIONAL VARIANCE OF MARGINAL UTILITY: Empirical results 2

The Consumer Expenditure Survey (CEX) is run by the US Bureau of Labor Statistics for the same reasons as the FES, namely to compute the Consumer Price Index, and the size of the sample is roughly similar (around 7,000 households per year). There are, however, two important differences between the two. First, in the CEX there is a short panel element, in that each household is interviewed 5 times over 15 months. The first interview is only a contact interview on which no information is disclosed in the public use tape; in each of the subsequent four interviews the household reports detailed information on expenditures incurred during each of the thiee previous months. Potentially, that is, each household provides twelve monthly expenditure accounts. Second, the households in the CEX do not compile a diary but answer retrospective questions.

The households in the CEX do not complete their cycle of interviews simultaneously. Since interviews are performed every month, households interviewed in different months report information on different periods. For example, a household that completes its last interview in January 1990 provides information on each of the 12 months of 1989; while a household that completes its cycle of interviews in June 1990 covers the period from June 1989 to May 1990. One twelfth of the sample is replaced each month. And of course not all households complete the cycle of 5 interviews.

This sampling structure complicates the computations of the cross sectional variances we are interested in. We cannot compute the cross sectional variance in a given year using the monthly observations as we do in the FES because several observations would refer to a single household. Variation across monthly observations would therefore reflect time series (seasonal or cyclical) rather than cross sectional variability. Estimating annual consumption for each household is complicated by the overlapping structure of the sample: the “year” over which we have information for each household would be different depending on the month in which the household completes its cycle of interviews. Annual estimates are further complicated by the fact that not all households complete the cycle.