wordpress-themes.org wordpress themes wordpress themes


Our results are presented in three parts. We first describe the data sets used, then provide a graphical illustration of the main feature of the cross-sectional variance of consumption and marginal utility, and finally present the econometric results. Some of the details of the data treatment are given in the appendix.

Data description

We estimate equations (9) and (10) using three sets of average cohort data. The UK Family Expenditure Survey is the largest dates, including 20 annual surveys (1974-1993). The US Current Expenditure Survey is currently available for 13 years (1980-1992). The Italian Survey of Household Income and Wealth is available from 1984 to 1995; but given the characteristics of this survey, we use only data from 1987 to 1995.

Our choice to focus on the US, the UK and Italy was guided mainly by the availability of cohort data for these countries.


The main motivation for the collection of the Family Expenditure Survey (FES) on the part of the UK Department of Employment is the computation of the weights for the Retail Price Index. In recent years the data set has been used extensively to describe the behavior of UK households and to estimate structural models of consumption behavior. The sample includes around 7,000 households per year (the survey has no panel element). Each household stays in the sample for two weeks, during which it compiles a diary with its expenditure. At the end the diary is collected and further information on various expenditure items during the previous three months (typically durable goods and utilities) is gathered. The survey also takes information on several economic and demographic household characteristics, which range from labor supply to household composition. The quality of the consumption data, thanks especially to the use of the diary method rather than retrospective interviews, is remarkably good.

As each household stays in the sample for just two weeks, consumption figures are heavily affected by seasonality. This problem is particularly serious for the cross sectional variances we compute. For instance, part of the difference in consumption between a respondent interviewed in December and someone interviewed in August will certainly be due to seasonal effects, not to genuine cross sectional variability. To account for this, we deseasonalize the individual consumption observations before computing the cross sectional variance. The seasonal adjustment is described in the appendix.

This post was written by , posted on May 30, 2014 Friday at 2:41 pm