Food Balance Sheets (FBS) are one of the most important sources of data on food
availability for human consumption. This paper presents a method to improve the
information on food consumption patterns of FBS by using national household budget
surveys (HBS).
In this paper, food commodities are categorized into 16 major food groups. For each food
group, the contribution to the overall caloric intake is represented in shares. Item group
shares of 64 surveys from 51 low and middle income countries are compared with shares
from country-specific FBS. Given the countries represented in the data, the analysis
evaluates food consumption of over 3 billion persons worldwide.
A model based on a cross-entropy measure of information has been developed in order to
reconcile aggregate food consumption patterns suggested by FBS and HBS. The latter
model accounts for the fact that data from both data sources are prone to measurement
errors.
Overall, the results of the reconciliation suggest that average consumption of cereals, eggs,
fish products, pulses and vegetables are likely to be underestimated in FBS, while fruits,
meat, milk and sugar products are likely to be overestimated in FBS. Even though the
suggested changes in average food consumption are moderate, the results imply
considerable relative changes in the aggregate consumption of single food groups.
Furthermore, the results imply that the aggregate consumption of fats is 2% higher than
currently assumed.
The updated consumption patterns provide valuable information from an agro-industrial
perspective. Differences in updated consumption pattern with respect to the original FBS
might suggest a re-evaluation of FBS elements of the value chain, starting from production
and ending at food losses