but cannot compare the patterns of relative prices of consumed quantities and the
relative transformed energy extracted by IHBS. Thus, the assumption that individ-
uals of lower economic status cover calorie requirements more easily by purchasing
high-calorie products, which are often cheaper than low-calorie ones, is at work.
(Drewnowski and Specter, 2004)4.
Figure 3 illustrates these points by dividing the sample into several socio-economic
characteristics - gender, age, education and economic status - and collecting house-
hold data for each subsample. Naturally, gender is shared by males and females,
education by household heads with a university degree, Master or PhD with respect
to a lower level of education, age by four classes of adults (18-34; 35-49; 50-64; ≥ 65)
and economic status by households which are below or above the relative poverty
line.
Panel a) shows that shifts in relative prices (healthy/unhealthy) for males are re-
lated to an increase in unhealthy food consumption and, on average, of total calories;
the influence on energy is less important for females. These results also find confir-
mation in educational and economic status. Greater sensitivity to healthy/unhealthy
price changes regarding the consumption of unhealthy foods is shown for households
with lower education (panel b) and lower income (panel c), with quicker growth of
the sub-samples in the last few years. Panel d) shows the patterns of price changes
by age. Although all the classes respond quickly to changes in healthy food prices,
the age classes in which the potential substitution of healthy foods with unhealthy
ones is more evident are those between 35-49 and 50-65.
These observations impose testable extensions on demand models: if the data
predict that significant price substitution effects of healthy versus unhealthy food
consumption partly justify the pattern increase in Italian overweight, then: (a)
the use of a regional price index for the categories of healthy and unhealthy foods
increases the variability of food response in estimating demand equations, which are
particularly severe when dealing with the aggregate price index; (b) the existence of
specific patterns of the elasticity of substitution may be assessed within demographic
and socio-economic groups as well as comparative estimations with respect to those
of the whole economy.
This extension therefore provides a general framework for testing the hypotheses
4The subsample for approximating low income households was obtained following the ISTAT procedure of calcu-
lating the relative poverty threshold, according to which, a household is considered under the threshold if its total
per-capita expenditure (total expenditure over the number of components) is lower than the average per-capita ex-
penditure. An equivalence scale was used to determine the relative poverty threshold for households with a number
of components other than two.