robustness of results below, to assess the sensitivity of these estimations. Demand
for healthy food is a luxury, whereas unhealthy food is a necessity. The estimated
expenditure elasticities are in line with the findings of Zheng and Zhen (2008) in
the United States although a different classification (and habits) were responsible
for differences in impact measures. Lastly, the cross-price elasticities show that
shifts in healthy or unhealthy food prices enter consumers’ choices to substitute
the relatively expensive food category for the cheaper one. According to estimated
cross-elasticities, they are statistically significant and have well-defined sizes at the
estimation point.
However, we cannot make direct inferences regarding substitution effects on pat-
terns of prevalent obesity in Italy because, as reported in Section 2, the prices of both
unhealthy and healthy foods rose in the sample period. The asymmetric responses
of the elasticities evaluated at the sample means show that a 1 percent increase in
the price of healthy food increases the budget share in the unhealthy food category
by 0.536 percent, whereas those of unhealthy food increases the budget share of
healthy food by 0.402 percent. The implications are threefold. The net effects of
changes in food consumption, given shifts in relative prices, indicate a slight but
significant impact on the growth of unhealthy food consumption. According to the
net estimated elasticity of substitution, we can also reproduce the data reported
in Figures 2 and 4, in which it was shown that quicker changes in healthy versus
unhealthy food prices increased relative unhealthy food consumption and its ex-
penditure share. Lastly, because the unhealthy category is more energy-dense, the
demonstrated increases in total calorie intake and overweight patterns are therefore
partly modelled by the channel of convenient food purchases.
If our attention is concentrates on the dynamics of the elasticities of substitu-
tion, η21 and η12 , an interesting implication of the non-stationarity of prices is that
elasticities may change over time. This point is illustrated in Figure 6. The cross-
price elasticity for healthy foods, η12 , appears to change little over time, whereas a
slight recent increase in η21 is recorded since the end of 2001, when unhealthy foods
became a much larger share of total spending (see Figure 4) or a rising category of
food expressed in terms of relative quantities (see Figure 2).
The robustness of our estimates are shown by moving bread, pasta and olive oil
into the sectors of healthy (specification I) and unhealthy foods (specification II),
respectively. The implicit price index was then used to estimate the parameters of
19