question as to whether shifts observed in relative prices affect consumer behaviour in
food choices, we examine whether the estimated cross-price elasticities of unhealthy
and healthy foods are significant for Italy.
Second, unlike Zheng and Zhen (2008), who estimate a demand system by an
aggregate national price index, we use a monthly regional price index associated with
data of household expenditure and demographic and socio-economic characteristics.
This choice reflects the fact that the variability of the demand system obtained
from regional prices is much larger than that of prices derived at national level. Our
empirical strategy addresses the inference problem in estimating demand equations
which may arise from insufficient price variations - a problem which is particularly
severe when dealing with macro-data.
Third, we test the hypothesis that the rise in unhealthy food consumption is
partly responsible for the increased body weight of Italian adults. In turn, this
hypothesis assumes, that unhealthy foods are more dense in energy. A nutritional
method is used to transform the consumption of healthy and unhealthy food cat-
egories and to clarify the relationship with light and more energy-dense foods, re-
spectively.
A final contribution is to extend the estimations of cross-price demand elasticities,
explaining how they vary according to gender and socio-economic factors. Collecting
time-series of cross-sections allowed us to determine consistently whether changes
in food prices alter dietary intake patterns sufficiently to explain weight gains by
sub-groups2 . Previous research, exploring the rise of disparities in overweight and
obesity among social groups, supports the hypothesis that, the less education an
individual receives, the greater the overweight recorded (Klesges et al. 1998; Dennis
et al., 2000; Hardy et al., 2000; de Saint-Pol, 2009). These heterogeneous responses
among groups appear to be confirmed according to individual standards of living (de
Saint-Pol, 2009), whereas estimations of household income are not conclusive (Ball
and Crawford, 2005; Villar and Quintana-Domeque, 2009). The existence of the
confounding effects of age and gender in individual body weight identified by Bray
(1987) and de Saint-Pol (2009), respectively, justify our choice to test the patterns
of substitution elasticities separately.
Our main results are that the estimated elasticities of substitution do have the
expected effects. We cannot reject the hypothesis that falling relative unhealthy food
prices is one probably cause of the prevalence of overweight in Italy, leading people
2For a review of longitudinal studies of socio-economic status and weight, see Ball and Crawford (2005).