Therefore, we take an alternative approach in evaluating whether changes affect
body weight. The growing availability of microdata from household expenditure
surveys and its role in allocating food expenditure allows us to model the dynamic
of food consumption (or its categories) consistently and to exploit the richness of
survey data information. This approach has also recently been used to estimate the
relationship between household food consumption and obesity in France (Bonnet
et al., 2008). In our case, it is worth noting that the choice to analyse changes
in food behaviour based on household consumption also appears to be consistent
with the statistical representativeness of food consumption. Indeed, the share of
food expenditure of households allocated for food away from home has remained
constant during the last decade, and ranges from 14 to 16 percent of total food
expenditure1 .
In this paper, we use the monthly consumption data of nine Italian Household
Budget surveys (IHBS) (1997-2005), to model a micro-economic demand system
for foods (and non-foods). In particular, by separating a composite category of
expenditure on healthy food from unhealthy ones (and residual foods and non-
durables), we shed light on the latent mechanisms of substitutability, based on the
differences between the price patterns of these goods, and the conditions in which
total calories rise when the relative price of the unhealthy food category - associated
with energy-dense foods - falls.
Our research is consistent with the hypothesis of Cutler et al. (2003), i.e. tech-
nological changes have lowered the relative full prices of mass-produced foods and
favoured the consumption of energy-dense ones. Our work is also related to the
work of Auld and Powell (2009), who model food consumption in such a way that
changes in relative prices change their contribution in terms of total calorie intake
solely by their compensated (or Hicksian) price elasticities. Our work is methodolog-
ically close in spirit to the work of Zheng and Zhen (2008) which, within a demand
system, with three categories of consumption for the United States and Japan, com-
prises the components of healthy and unhealthy foods and tests the influence of
substitution effects on these food categories.
This paper makes a number of contributions. First, to answer the important
1 Recent research programs have also used scanner data from grocery stores to estimate price elasticities, with
observation of prices paid for individual transactions. Although within-category substitution of food would be
much more meaningful as consumers can substitute more easily towards similar goods, it is doubtful whether this
segmentation would lead to more effective policy implications. In a plethora of food products on offer, many items
are generally not consumed by households, so that the problem of possible inefficiencies in implementing food policies
to reduce the frequency of overweigh, while recognized in the literature, needs to be resolved. One way of doing this
should involve modelling households’ zero expenditure.