together with other non-durable goods, so that the link between healthy - energy-
light foods and unhealthy - more energy-dense ones is unambiguous The empirical
section then provides a sensitivity analysis of the estimates of substitution effects
by first including bread, pasta and olive oil in the unhealty food category and then
in healthy food one.
Figure 1 shows the overlapping paths of the fraction of overweight and obese
adults and the relative changes in healthy and unhealthy food prices. For this pur-
pose, we match two sources of data. First, we use the annual multipurpose survey of
Italian households (2002-2005) conducted by the Italian Institute of Statistics (IS-
TAT) to obtain an average of the Italian body mass index (BMI). We then integrate
these measures of individual BMI with data from two more general ‘’multipurpose
surveys” of 1994-1995 and 1999-2000, and interpolate the values for the missing an-
nual surveys. Note that these breaks do not allow us to use the BMI indicator as a
direct (long-run) body weight response of changes in food price categories 3.
Instead, the annual mean of expenditures and prices for these two categories of
consumption are obtained by extracting microdata from the IHBS, released annually
by ISTAT. Because the latter surveys are the main sources for the empirical section,
we present detailed data below, merely noting that the relative prices of unhealthy
and healthy foods are obtained by the ratios between nominal and real expenditures,
and that real expenditure, as a consumption index, aggregates the ratios between the
current expenditure for each individual item and its index price (at national level).
The graph, shows a constant increase in BMI over time, although upward trends
are not recorded. As an aggregation of the repeated Italian household surveys, the
picture which emerges shows that the prices of healthy foods rise more quickly than
those of unhealthy ones. Descriptively, the increase in healthy food versus unhealthy
food prices is, on average, 1 percent per year.
According to the hypothesis that, as unhealthy food becomes relatively cheaper,
people are expected to substitute away from healthy foods towards unhealthy ones,
with greater consumption in the latter category. Relative consumption, expressed
as a constant expense, is reported in the continuous line of Figure 2 (i.e., vice versa
with respect to relative prices). The predictions of the theory are supported by the
slight upward positive trend in the growth of unhealthy food consumption during
3The body mass index (BMI) is a measure of body fat based on height and weight, which applies to both adult
men and women. Four categories are generally used to classify adults: i) Underweight BMI ≤ 18.5; ii) Normal weight
= 18.5-24.9; iii) Overweight = 25-29.9; iv) Obesity = 30 or over. It is known that BMI is not the most accurate
measure of body fat and that self-reported weight produces measurement errors for young and adult people. For a
critical discussion of this indicator, see Burkauser and Cawley (2008).