Carlo Andrea Bollino, Federico Perali and Nicola Rossi, "Linear Household
Technologies," Journal of Applied Econometrics, forthcoming.

This file contains some additional information about the data used in this
paper. See the readme file.

These data are based on published data from the Italian Central Statistical
Office (ISTAT). The individual unit is the cell defined in terms of family
size and income decile aggragating individual observations of the surveys
undertaken during the period 1973-1992.

Expenditure surveys were undertaken by ISTAT in 1968 and carried out
regularly in subsequent years. However, in 1973, the methodology of the
survey went through a series of major revisions. Therefore, results refer
only to the homogeneous series of surveys starting in 1973 and used in this
paper. The survey covers all private consumers' expenditures including "self
consumption" of agricultural products, imputed rents and consumer goods
received as wage. However, it does not cover purchases of houses or land as
well as expenditures related to production.

As far as consumer durables are concerned, expenditures are recorded at the
time of purchases. Hence, for example, payments by installments are not
accounted for. The survey's central unit is the household defined as set of
persons living together characterized by the common use of their incomes.
The household, as defined in the surveys, includes servant and nurses.

The methodology of the survey can be briefly described as follows. 

In a first stage, 700 towns and cities are selected and divided into two
groups: group 1 includes capitals of province and towns with at least 50,000
inhabitants, group 2 includes the rest. A predetermined number of households
in towns in group 1 are surveyed each month of the year while towns in group
2 are divided in three further subgroups and each subgroup participates the
survery in the first, second or third month of each quarter of the year.
Summarizing, each month approximately 320 towns are surveyed: 140 of them
belonging to group 1 above and 180 to group 2. In a second stage 3,000
households per month are randomly selected from the registrar's offices of
the selected towns, thus giving a total of 36,000 households surveyed in the
year. The rate of participation in the survey has been approximately
estimated around 85%. The initial results are then related to the population
by ISTAT.

Empirical estimates are based on a data set extracted from the ISTAT annual
household survey, for the period 1973-1992. Budget expenditures are grouped
by expenditure classes and family size for a total of 1,815 observations;
exclusion of non reporting and/or meaningless cells reduces the sample size
to 1,729. It is relevant to emphasize that household size is the only
demographic variable available from published data. This informational
limitation does not allow us to fully account for systematic heterogeneity
in preferences. Furthermore, it must be noted that model adequacy can be
significantly affected by the high level of data aggregation in terms of
aggregation both across households due to the fact that the published data
are in cell means and across goods. The price vector has been derived from
the corresponding consumption breakdown available from NIA. In order to
limit the computational complexity, only three categories have been
analyzed: "food," "housing and fuels," "miscellaneous." The "miscellaneous"
item includes clothing, health care, transportation, communications and all
other goods and services.

Here is the data section of the TSP program used to aggregate the data:

READ (FILE='cab.WK1')
     ANNO INC FAM PCT
     X1 X2 X3 X4 X5 X6 X7 X8 X9 X10
     PP1 PP2 PP3 PP4 PP5 PP6 PP7 PP8 PP9 PP10
     CC1 CC2 CC3 CC4 CC5 CC6 CC7 CC8 CC9 CC10;

? ------------------ GENERAZIONE   DATI ----------------------------
 SMPL  1  1815 ;
     GENR WGT=  1 ;
     GENR XT = X1+X2+X3+X4+X5+X6+X7+X8+X9+X10      ;
?
 SMPL  1  1815 ;
? 1 FOOD
     GENR W1 = X1 / XT  ;
? 2 HOUSING AND FUELS
     GENR W2 =(X4+X5+X6+X8) / XT ;
? 3 MISCELLANEOUS
     GENR W3 =(X2+X3+X7+X9+X10) / XT ;
?
? ---------------- CONTROL
?
   GENR  WRES = 1-W1-W2-W3       ;
?   PRINT  WRES ;
?

Legend: 
	Anno= year
	INC=income
	FAM= household size
	PCT= cell weight
	X1-X10=expenditures
	PP1-PP10= prices