ditures and purchased quantities on food, alcoholic and non-alcoholic
drinks and tobacco products have to be noted in absolute detail.
For our purposes the detailed log book is of great value. By dividing the
expenditures by the quantities, we know the individual prices at which the
household purchased the commodities.
So we have a price variation in our cross section that enables us to estimate
demand systems with data from only one year. In cases where the individual
price data is not available, an econometric estimation of a demand system
would only be possible by using time series of aggregated cross section data.
The time series approach is only feasible if an income and expenditure survey
is conducted every year, as it is the case in the US and the UK. The only
yearly survey available in Germany is the already mentioned LWR, which is
less detailed than the EVS and which is explicitly not recommended for any
time-series analysis.
Taking into account the above described data limitations, the demand
systems are estimated for nine staple foods listed in the detailed log book
of the EVS: Milk, cream, eggs, butter, margarine, apples, bananas, mineral
water and coffee. The choice of the specific commodities was driven by
several considerations. First, the level of commodity detail available in the
detailed log is very high. Second, the staple foods chosen are purchased by
the majority of the
households, so that a sufficiently large sample size is achieved without
making assumptions about the households with zero consumption of an item.
Third, the staple foods chosen are homogenous goods, without considerable
quality changes, new types appearing or wide differences in product char-
acteristics. Fourth, food is purchased frequently which is not the case for
durables. Fifth, the food consumption is assumed to be separable from the
non-food consumption in consumers utility function, so that the cross price
reactions between food and non-food commodity groups should be rather
small. And finally the availability of absolute price data from the consumer
price statistic is also better for the food-, beverage- and tobacco-group than
for other commodity groups. To ensure the homogeneity of the consumers
analyzed, the socio-demographic variables of the EVS are used to include
only households in the sample that are consisting of couples with one or
more children younger than 18 years old.
For research purposes the research data centre of the GFSO provides so-
called Scientific-use-files containing household level micro data from the EVS.
28