Generically speaking, “food consumption statistics” refer to estimates of what food people are eating – which foods, and how much. They may refer to actual foods, or types of foods, or the nutrients consumed in the foods. This sounds simple and straightforward, but in fact it is an enormous challenge – especially when aiming to produce food consumption statistics that are comparable across cultures and countries. There are many different approaches to quantifying food consumption, and depending on the source of data, the statistics produced refer to very different things. Rough estimates of the foods and nutrients consumed in a country can be derived from (1) national food supply data (i.e. availability), and (2) household food consumption or food acquisition data collected in household surveys (i.e., apparent intake). The most precise estimates of actual individual food and nutrient intake in a population come from detailed individual dietary intake surveys.

Three very different sources of data used to assess the food consumption in a population include: (a) individual quantitative dietary intake surveys (IQDIS), which use tools such as the 24-hour food recall; (b) household consumption and expenditure surveys (HCES); and (c) supply utilization accounts (SUA) and food balance sheets (FBS). The IQDIS capture individuals’ food intake, providing data that are crucial for gaining insights into dietary patterns, disaggregated by sex and age. Such data are key for developing evidence-based policies and programs for food, agriculture, health and nutrition. The HCES capture households’ apparent food consumption, based on reported food consumption or acquisition, and are often used by researchers in the absence of nationally-representative IQDIS, which tend to be less frequent. The SUA and FBS data provide an overview of a country’s food supply from agriculture, fisheries, and aquaculture sectors for a calendar year and produce information on foods available and used for human consumption.

The label “household consumption and expenditure survey” is used as an umbrella term for household-level surveys developed to produce poverty estimates and consumer price indices and to inform economic policies and planning. The list of surveys includes “Household Budget Surveys”, “Household Income and Expenditure Surveys” and “Living Standard Measurement Surveys”. These surveys collect information on household characteristics (e.g., region and urban-rural), household members characteristics (e.g., sex, age, education, food- and non-food expenditures), and food quantities consumed and/or acquired during a reference period. These surveys have representativeness at the national level and geographic areas and are conducted every one to ten years depending on the country. 

We refer to food consumption and nutrient intake statistics from household consumption and expenditure surveys (HCES) as “apparent” because they are based on food quantities consumed and/or acquired by households (without information on intra-household distribution), and not on individuals’ actual intake. Furthermore, HCES do not inform on households’ food (a) wasted; (b) given to employees, guests, relatives, or pets; (c) used to feed livestock; (d) used for small food businesses, or for resale; (e) cooking methods; and (f) fortification (i.e. whether the acquired foods were fortified with vitamins and minerals). Another reason food consumption and nutrient intake statistics from HCES are referred to as “apparent” is because the survey food modules typically do not collect very detailed information on the foods consumed or acquired. For example, it is common to use “closed” (non-open) food lists, non-detailed food labels (e.g. “Beef” or “Poultry”), and food combinations with different nutrient content as a single item (e.g. “Fresh and dried fish”). 

It is possible to express statistics on apparent intake per adult equivalent (based on energy requirements) when it is assumed that intra-household distribution is based on household’s member energy requirements. This is possible due to the expected high correlation between energy intake and energy requirements. However, the same assumption of correlation is not made for nutrient intake and requirements. Thus, there is really no value-added in expressing apparent nutrient intake per adult equivalent based on nutrient requirements. 

The information needed to produce statistics on apparent nutrient intake from HCES data include: (a) food quantities consumed and/or acquired; (b) gram or milliliter equivalents of local or non-standard food quantity units; (c) the non-edible portion of each food; (d) foods’ density values (i.e., grams per milliliters); (e) nutrient content of foods from food composition tables/databases; and (e) number of food partakers (taking into consideration the number of household members present and absent and the number of guests and non-family members that shared the meals during the reference period of food data collection); or, in the absence of this level of detailed information, the total number of household members. If for some foods, such as those prepared away from home, it is not possible to convert food quantities into nutrients, or only monetary values are collected, food monetary values for at-home foods are also needed to compute the unit cost of nutrients. These unit costs are then combined with monetary values of foods prepared away from home to estimate their nutrient content.

First, the FAO Statistics Division calculates, for each household, the nutrients from at-home food consumption and the monetary values of the respective foods. Then, the nutrient unit cost by household is computed by dividing the sum of monetary values by the sum of nutrients. In the next step, the median nutrient unit cost is computed for each combination of region, urban-rural area, and income quintile group. Households are classified in income quintile groups considering population weights (i.e., household size * household weight). Finally, the nutrient content of foods prepared away from home is estimated using the median nutrient cost of the household’s region, urban-rural area and income group and the foods’ monetary values.

In order to produce apparent nutrient intake statistics from data collected in HCES, information is needed on the nutrient content in foods, which are available in food composition tables and databases (FCTs/FCDBs). Thus, survey-specific nutrient conversion tables (NCT) containing information on the nutrient content of each food reported as consumed and/or acquired by households in the HCES are needed. The nutrient content is obtained matching foods in the HCES with foods in FCTs/FCDBs. The FCTs/FCDBs are published for a country or for a region and they contain information on the nutritional content of foods derived from quantitative analysis of representative food samples and/or compiled from other FCTs/FCDBs.

Survey-specific nutrient conversion tables for household consumption and expenditure surveys processed by the FAO Food Security and Nutrition Statistics team are available upon request to the email: [email protected].

The most common units to express vitamin A are micrograms (mcg) of retinol equivalents (RE) and mcg of retinol activity equivalents (RAE). Different sources of vitamin A requirements are based not only on different criteria but also on different units of vitamin A intake. For instance, the FAO and the World Health Organization (WHO), as well as the European Food Safety Authority (EFSA) published vitamin A average requirements keeping RE as the expression of intake. However, the United States Health and Medicine Division published vitamin A average requirements keeping RAE as the expression of intake. While vitamin A in mcg of RE is equal to the sum of mcg of retinol, and mcg of beta-carotene equivalents divided by six, vitamin A in mcg of RAE is equal to the sum of mcg of retinol, and mcg of beta-carotene equivalents divided by twelve.

Yes, it is possible if the household consumption and expenditure survey includes an experience-based food security module, such as the Food Insecurity Experience Scale (FIES) or the “Escala Latinoamericana y Caribeña de Seguridad Alimentaria (ELCSA)”. An example of the type of analysis that could be performed can be found here: https://doi.org/10.4060/cb6217en.

The FAO Food and Diet Domain publishes statistics on apparent food consumption and apparent nutrient intake from HCES data (URL: https://www.fao.org/faostat/en/#data/HCES). 

The comparison across countries of food and nutrient statistics from household consumption and expenditure surveys (HCES) is only possible if a number of challenging criteria are met. First, the resulting statistics depend on a number of aspects related to the design of the food module, such as: (a) the length and level of detail of the food list; (b) whether consumption or acquisition is reported; (c) how information on foods prepared away from home is collected; (d) number of days for which food data is collected (e.g., seven or 14 days); and (e) the tool used to collect information: recall vs diary. Second, the survey data collection period might also impact the final statistics. For instance, a survey conducted throughout a year might capture changes in dietary patterns (e.g., in fruit and vegetable consumption) while a survey conducted during a short period might not capture changes due to seasonality. Furthermore, in surveys conducted during short periods, the period of the year might also impact the final statistics, especially whether it includes festivities such as Christmas or Ramadan, or if the survey is conducted before or after crop harvesting Finally, the matching of foods reported in the survey with fortified or unfortified foods impacts the nutrient apparent intake statistics.

For more information on the comparability of food and nutrient statistics from HCES across countries refer to: FAO. 2024. Food and diet – Statistics on dietary data. FAOSTAT Analytical Briefs, No. 82. Rome. https://doi.org/10.4060/cc9454en.

The comparison of food and nutrient statistics from different HCES, conducted in the same country, should be performed with caution. First, changes in the food module’s design might have an impact on the final statistics, especially regarding (a) the length of the food list; (b) the level of detail in the food labels; (c) if consumption of acquisition is reported; (d) the reference period for food data collection (e.g., seven or 14 days); and (e) the tool used to collect information: recall vs diary. Second, a change in the survey data collection period might also impact the final statistics. For instance, a survey conducted along a year might capture changes in dietary patterns (e.g., in fruit and vegetable consumption) while a survey conducted during a short period might not capture changes due to seasonality. Furthermore, in surveys conducted during short periods, the period of the year might also impact the final statistics, especially if it includes or not festivities such as Christmas or Ramadan, or if the survey is conducted before or after crop harvesting. Then, also changes in the module used to report on foods prepared away from home impacts the nutrient apparent intake statistics. Finally, nutrient apparent intake statistics are also influenced by using different food composition tables/databases between survey rounds or when one survey round involves fortified foods in the food matching, but the other survey round does not. 

The food and nutrient statistics produced from HCES data are not comparable with those derived from individual quantitative dietary intake surveys (IQDIS), supply utilization accounts (SUA) and food balance sheets (FBS). The statistics produced with each of these data sources provide different lenses on food consumption and should be interpreted and used in very different ways. Even when the same food group classification is applied to the data collected, the statistics by food group derived from HCES are not directly comparable to those derived from IQDIS and SUA/FBS data.

Important differences in the data sources that explain why the resulting food and nutrient statistics are not comparable include:

• HCES collect data on apparent food consumption from households, IQDIS collect information on actual food intake of individuals, and SUA and FBS data refer to food available for human consumption at the country level (including institutions);
• HCES rarely account for details such as foods prepared away from home and cooking methods, whereas IQDIS typically do; SUA and FBS are country-level statistics that cannot provide this level of detail;
• HCES and IQDIS typically account for consumption of foods from family and other small garden production, whereas SUA and FBS do not;
• SUA and the FBS data refer to food available over a year, whereas HCES and IQDIS reflect food consumed during the specific data collection period of the food consumption module;
  

For more information on the comparability of food and nutrient statistics from different data types refer to:  FAO. 2024. Food and diet – Statistics on dietary data. FAOSTAT Analytical Briefs, No. 82. Rome. https://doi.org/10.4060/cc9454en.