2.0 Introduction

This chapter considers survey methods suitable for a quantitative assess­ment of food avail­able for con­sump­tion / consumed at national and house­hold levels. Methods suitable for measuring food con­sump­tion at the indi­vidual level are discussed in Chapter 3. Methods designed to estimate the exposure to hazardous sub­stances, although some­times based on house­hold food supply, and those providing only a qualitative assess­ment such as the household dietary diversity score, or capturing only part of the avail­able food supply (e.g., using Universal Product Codes — UPCs), are not considered.

The food avail­able for human con­sump­tion by a popu­lation can be assessed through nationally rep­re­sen­tative surveys at the national level using Food balance Sheets (FBS), or alternatively at the house­hold level using either House­hold Food Con­sump­tion Surveys (HFCS) or House­hold Con­sump­tion and Expenditure Surveys (HCES). The resulting data from all these surveys are expressed as per capita food supply for human con­sump­tion, and are used to estimate per capita avail­able intake of foods, and energy and nutri­ents, when matched with appropriate food com­posi­tion data.

The term that should be used for data derived from most of these national and house­hold surveys is either “food supply” or “food avail­able for con­sump­tion” rather than “food con­sump­tion”. Only certain types of house­hold surveys measure actual foods consumed.

Data from FBSs provide no infor­mation on the distrib­ution of the food avail­able for con­sump­tion within the country (in terms of geographical areas, seasonality, or demographic characteristics), unlike the data gener­ated from house­hold surveys. However, house­hold surveys (HCS or HCES) do not provide infor­mation on the avail­ability of food/nutri­ents by age, sex, and physiological status. Moreover, the interpre­tation of the data from house­hold surveys in terms of dietary adequacy presents challenges and uncertainties unless complemented with indi­vidual-level dietary intake data.

Several indi­cators can be derived from FBS and HCES data to monitor and compare food insecurity at the global and national levels. These indi­cators can be based on the quantity of food avail­able for con­sump­tion (e.g., preva­lence of under­nourish­ment (PoU) based on the energy supply per capita), as well as the food quality (e.g., Household Dietary Diversity Score).

2.1 Food Balance Sheets

Food balance sheets are particularly impor­tant in coun­tries where food con­sump­tion surveys at the house­hold or indi­vidual level are not performed regularly on a rep­re­sen­tative sample of the popu­lation, as in some low-income coun­tries.

Several terms have been used to describe FBSs. These include “country's food supply”, “national food accounts”, “food moving into con­sump­tion”, “food con­sump­tion statistics”, “food disappearance data”, “food avail­able for con­sump­tion” and “con­sump­tion level estimates”. Here, the term “country's food supply” is used.

Food balance sheets present a comprehensive picture of the pattern of a country's food supply during a specified reference period: the calendar year, the agricultural year or the crop year (FAO, 2021c). The sources of the food supply and its utilization are reported for each food item potentially avail­able for human con­sump­tion, and includes primary com­modities such as ‘wheat, rice, fruit, veg­etables, and processed com­modities such as veg­etable oils and butter. FBS data cannot be disag­gregated to deter­mine the distrib­ution of a country's food supply spatially, seasonally, or by demographic characteristics.

Both official and unofficial data are used for the compi­lation of FBSs; National statistical offices con­stitute an impor­tant data source. However, data mea­sured directly on food avail­ability at national level may be difficult to obtain (FAO, 2021a). Instead, FBS compilers often derive estimates of food avail­ability by making certain adjustments based on other existing data sets that measure food produc­tion or con­sump­tion. For example, two official data sources that may assist in the estimation of a country's food supply avail­ability are Industrial Output Surveys and House­hold Con­sump­tion and Expenditure Surveys (HCES). See Section 2.2 for the description of HCES surveys.

The country's food supply is calcu­lated from national food produc­tion, plus imports, plus food taken from stocks. Exports and food added to stocks are then subtracted to obtain a gross estimate of a country's food supply. The estimates of food diverted for non-human uses (e.g. animal feed, seed, non-food use) and of food loss are then subtracted from the gross country's food supply to obtain the net country's food supply, i.e. the total food avail­able for human con­sump­tion in a country at the retail level (Figure 2.1, Nelson, 1984). Data on non-commercial food produc­tion and detailed infor­mation on processed foods are not avail­able in Food Balance Sheets.

For the purposes of the FBS, “food loss” closely aligns with “post-harvest / post-slaughter loss”. It represents the quantity of food that leaves the produc­tion / supply chain at any stage, e.g., during produc­tion, pro­cess­ing and distrib­ution, up to the retail level (FAO, 2021a). Food loss may be consid­erable in coun­tries where the agricultural products reach the consumer after travelling long distances and/or passing through several intermediaries prior to retail. In such cases, large amounts of food are lost because of damage, especially perish­able foods such as fresh fruit and veg­etables or processed foods with a short shelf life. Assumptions about losses may be based on expert opinion obtained in the coun­tries. Losses or waste at the retailer, food catering, and house­hold level are not considered in the compi­lation of FBS. Hence, total food supply estimates derived from the FBS are likely to be higher than average food consumed at the house­hold level.

An example of the upper portion of an FAO food balance sheet which also shows in detail the sources of domestic supply and utilization of two of the major food groups (i.e., Cereals and Starchy roots) is shown in Table 2.1.

In some cases, infor­mation on a country's food stocks is not avail­able. The effect of this data gap is reduced by preparing FBSs as a moving average for 3y periods for all coun­tries. For more details, see FAO (2021f).

FAO in its online database (FAOSTAT) pro­vides free access to FBSs for over 245 coun­tries and territories and all the FAO regional groups, from 1961 to date (FAO, 2021c). The FBSs reported are “standardized FBSs” in which the com­modity list is confined only to the quantity of the primary com­modities avail­able for con­sump­tion, with the quantity of processed com­modities converted into their original primary com­modity equivalent. For example, for bread, quantities are expressed in wheat equivalents. For more details on the standardization of Food Balance Sheets, see FAO (2021g).

The major food groups listed in the FAO FBS data are shown in Box 2.1. A list of food com­modities classified into these major food groups is avail­able from FAO (2021c) in the definition and standard section of the FBS website. As an example, the food com­modity “cassava” can be found in the food group “starchy roots”. The per capita supply avail­able for human con­sump­tion is obtained by dividing the food supply by the estimate of the partaking popu­lation in each country; it is expressed in grams per capita of indi­vidual food com­modities. FAO uses mid- year official popu­lation estimates released by the United Nations Devel­opment Program. FAO recom­men­dation for FBS com­pilers, the number of migrants and tourists must be calcu­lated as the differ­ence between country's outbound travellers and inbound visitors and then subtracted from the population estimate (FAO, 2021a).

Data on per capita food supplies are expressed in terms of quantity, and in terms of caloric value, referred to as the Dietary Energy Supply (DES) (kcal/capita/day). The pro­tein and fat content of each food group and com­modity calcu­lated by applying appropriate food com­posi­tion data are also avail­able and expressed in terms of g/capita/day (Table 2.1). In the future, FAO may also provide micro­nutri­ent data for each food group and com­modity.

Food balance sheets are derived statistics. Hence, their accuracy is depen­dent on the reliability of the underlying basic statistics. The coverage and accuracy of underlying statistics vary markedly across coun­tries. In some low-income coun­tries, the coverage and quality of the statistics are uncertain — especially for food diverted for non-human food uses and for food stocks. As well, in low-income coun­tries where subsistence agriculture is wide­spread, FBSs may result in an under­estimate of per capita food supply because the house­hold con­sump­tion of vegetal food products obtained through growing or gathering, and animal food products through breeding or hunting, are not considered in the FBS calculations. Further, as food systems become more sophisticated in coun­tries, systematic errors may increase.

In FAO documents, it is made clear that avail­ability for human con­sump­tion does not equate with con­sump­tion. The quantities of food avail­able for human con­sump­tion, as esti­mated in the FBS, reflect only the quantities reaching the consumer at retail level. The amount of food (and of nutri­ents) actually consumed may be lower than the quantity shown depending on the degree of loss and waste of edible food and nutri­ents in the house­hold. These may occur during storage, in preparation and cooking, as plate-waste, and when fed to domestic animals and pets, or when thrown away. Storage, preparation, and cooking will have a greater effect on the content of vitamins and minerals of the food com­modities than on the pro­tein and fat content. See: FAO (2021h).

Recently, FAO has changed its FBS method­ology, details of which are described in the document “The New Food Balances and the utilization variables” (FAO, 2021a). The key differ­ences between the new and old FBS include a proportional balancing mechanism, use of the 2019 UNPD popu­lation data, inclusion of a new food module, revised computations for losses, feed, and stocks, and addition of a new element on non-food for industrial use; for more details see (Lancet, 2024).

The release of FBS data for 2019 includes this new method­ology for the FBS data for 2014–2019. The new method­ology has also been applied backwards to cover 2010–2013, thereby giving consistency over time for the national per capita food supply data from 2010 to 2019. In the new method­ology, food avail­able for con­sump­tion by non-resident visitors has also been included for selected coun­tries in which tourism has a significant impact on the food supply (e.g., Small Island Developing States). Food Balance Sheet data pre­ceding 2010 will soon be recompiled using the new method­ology. FAO has devel­oped a FBS capacity devel­opment package for coun­tries that includes a compi­lation tool incorporating the new method­ology. Details are given in (FAO, 2022a).

2.1.1 Uses of Food Balance Sheet data

Food Balance Sheet data are easily accessible and provide coverage of per capita food supply across coun­tries and regions using a relatively consistent method­ology com­pared to many other sources of data on food supply.

Hence, they are often used to analyze overall trends, historic trends, and changes at country, regional, and global levels, in relation to both Dietary Energy Supply (Figure 2.2) and Food Groups (Figure 2.3). Asia has experienced the fastest growth in DES since 2010 (i.e., 6%), whereas in contrast the lowest historical DES is apparent in Africa, where there has been a slightly decreasing trend over the last decade (Figure 2.2).

Food Balance Sheets can also be used to formulate agricultural policies concerned with the produc­tion, distrib­ution, and con­sump­tion of foods.

In relation to food groups, at the global level in 2019, cereals accounted for the highest share of the total food supply, both in quantities (in tonnes) avail­able (25%) and dietary energy (44%), whereas although fruit and veg­etables represented 31% of the total quantities avail­able, they only accounted for 7% of the total dietary energy available due to their low kilocalorie content. In contrast, sugar, and fats and oils represented only 4% and 3% of the total quantities avail­able but 8% and 14% respectively of the dietary energy available, as shown in Figure 2.3.

In addition, FBS data have many other uses, including corre­lations with other data which have a global coverage. Thar et al. (2020) have summarized several health-related uses of FBS data, as shown in Box 2.2, based on their review of 119 FBS surveys. They identified 50 studies con­ducted from the earliest avail­able until 2016 that examined associations between dietary factors and mortality or health outcomes such as cancer, diabetes, or obesity. However, any interpre­tation in terms of causal/effect of such observed associations is highly ques­tion­able and should be discouraged. Further, given the impor­tance of the decisions that may be taken based on FBS data, users need to have a good under­standing of the strengths and weak­nesses of the source of the data. See Box 2.2 and Table 2.7 and the review of Thar et al. (2020) for a summary of some of the appropriate uses and potential weaknesses of FBS data.

Several investigators have examined the validity of the global dietary estimates gener­ated from FAO FBS data. For example, Del Gobbo et al. (2015) com­pared the FAO data with estim­ates based on nationally rep­re­sen­tative dietary data from indi­vidual-level surveys from 113 coun­tries. Substantial overestimates were reported for most food groups including fruits, veg­etables, whole grains, red and processed meat, fish and seafood, milk, and total energy, whereas beans and legumes, nuts and seeds were under-esti­mated based on FAO national food-supply data com­pared to the corres­ponding estimates from the national data at the indi­vidual-level. As a result, calibration equations were devel­oped to adjust the FAO estimates to improve their validity. The authors suggest that these calibration models could be used not only to improve estimates based on FBS data of per capita food intakes at the national level but also by age and sex.

One of the main uses of FBS is to monitor global hunger and food security by providing data on national diets and their nutri­ent adequacy. With these, insights into the coun­tries and regions that are most likely to be at greatest risk for low dietary supply of energy (DES)(kcal per capita) as well as selected nutri­ents, can be provided. Selected indicators derived from the food component data of FBS and used to monitor global hunger, food security, and nutrient adequacy are presented in Box 2.3. Details of some of their applications are given below.

2.1.1a Trends in global hunger

Ideally, the assess­ment of global hunger (and food security) should be based on national data on food con­sump­tion at the house­hold (see Section 2.2) or the indi­vidual level (see Chapter 3). However, only a few coun­tries conduct such surveys on an annual basis. Therefore, FBSs have a key role in the assess­ment of global hunger and food insecurity, published yearly in the report “The State of Food Security and Nutri­tion in the World” (SOFI) by FAO, IFAD, UNICEF, WFP and WHO (2022). This global monitoring report identifies coun­tries in which food insecurity is most prevalent, monitors global hunger and food security trends over time, and pro­vides projections of future global hunger and food insecurity.

In this report, the per capita Dietary Energy Supply (DES) (Box 2.3 and Table 2.1) esti­mated from FBSs is used as a proxy for the average energy intake of the popu­lation. The per capita Dietary Energy Supply (DES) is avail­able from FAOSTAT and calcu­lated using an FAO global food com­posi­tion database. The sum of the dietary energy content of the total food supply is then divided by the popu­lation size of the country and by 365 to calculate the per capita daily dietary energy supply (DES) avail­able for human con­sump­tion (See Table 2.1 and Figure 2.2). However, DES does not yield any infor­mation on the afford­ability, access, or con­sump­tion of dietary energy by different popu­lation groups within a country, although the contribution of energy from indi­vidual food groups (Table 2.1) avail­able for human con­sump­tion at the global level and within a country are calcu­lated.

Dietary Energy Supply is one of the three metrics used to estimate the preva­lence of under­nourish­ment (PoU), an indi­cator based on the per­cent­age of indi­viduals in the popu­lation who are in a condition of “under­nourish­ment” (Table 2.2). PoU is one of the indi­cators of Sustainable Devel­opment Goal 2 (FAO, 2021e) and is defined by FAO as the condition of an indi­vidual whose habitual food con­sump­tion is insufficient to provide, on average, the amount of dietary energy required to maintain a normal, active, and healthy life. The national PoU estimates are reported as three-year moving averages, whereas the regional and global aggregates are reported as annual estimates. Table 2.2 highlights the regional disparities in PoU, with Africa having the highest burden of under­nourishment as well as the region with the largest increase in the proportion of the population affected by under­nourishment during 2005 to 2021. Disparities at the subregional levels also exist and are summarized in FAO, IFAD, UNICEF, WFP and WHO (2022). PoU, however, does not provide infor­mation on which specific indi­viduals are under­nourished or on the quality or diversity of the diet. The three metrics used to estimate PoU are:

• Average dietary energy con­sump­tion (DEC). DEC is based on DES minus the incidence of caloric losses at retail distribution level
• Coefficient of Variation (CV) of energy intake. This is a measure of the level of inequalities in access to dietary energy con­sump­tion within a popu­lation
• Average minimum dietary energy require­ment (MDER) per person for the national popu­lation in each of the covered coun­tries. This represents the minimum amount of dietary energy per person that is considered adequate to ensure they maintain a minimum weight for health.

To estimate the CV of dietary energy intake, data on inter-house­hold variability from house­hold surveys such as House­hold Con­sump­tion and Expenditure Surveys (HCES) are used. Higher CV values represent larger levels of dietary inequality. Use of such data in the PoU calculation is crucial because food (and conse­quently energy) is never equally distributed among house­holds within a country. Hence, any com­par­ison of per capita DES with energy require­ments at the country level without considering the CV of energy intake would largely under­estimate the preva­lence of food insecurity at house­hold level in many coun­tries.

To derive the MDER, factors such as the age and sex structure of the population as well as height of the individuals within each country are applied. A coefficient is assumed for physical activity corresponding to the average values of the PAL category “Sedentary or light activity lifestyle” (FAO human energy require­ment 2001). Using the variation in intake and energy require­ments, under­nourish­ment can be calcu­lated. More details of how these three metrics are used to estimate PoU are given in FAO (2022).

Note the presence of persistent regional disparities in the proportion of the popu­lation affected by hunger from 2005 to 2021 (Table 2.2), with Africa having the highest burden of under­nourish­ment as well as the region with the largest increase in the proportion of the popu­lation affected by hunger during this time. Disparities also exist at the subregional levels: these are also summarized in the report.

FAO also pro­vides data on the Cost and Afford­ability of a Healthy Diet (CAHD). This has been shown to be strongly associated with PoU and with different forms of malnutri­tion, including child stunting and adult obesity. The cost of a healthy diet is defined by FAO as the minimum cost of food, using the least expensive avail­able items in each country, that people have to pay to secure a healthy diet. The latter is based on the cost of the average quantities of each food group recom­mended in national Food Based Dietary Guidelines (FBDG). On average, the healthy diet meets 95% of nutri­ent needs.

To deter­mine afford­ability of a healthy diet, the cost of a healthy diet is com­pared with country-specific income distrib­utions avail­able in the “Poverty and Inequality Platform” (PIP) of the World Bank. From these data, the per­cent­age and number of people in each country / region who cannot afford a healthy diet are esti­mated. The threshold for an unaf­ford­able healthy diet in 2020 was when its cost exceeded 63% of the income in a country. However, food prices have surged since 2020, so this threshold for unaf­ford­ability is likely to change. For more details on the derivation of CAHD and the estimates of the number of people in the world and by region and subregion, see Chapters 2 and 3 and ANNEX 3 in FAO, IFAD, UNICEF, WFP and WHO (2022).

2.1.1b Food security indi­cators at national level

FAO (2006) has emphasized that “Food Security” is

“a state in which all people, at all times, have physical and economic access to sufficient, safe and nutritious food to meet their dietary needs and food preferences for an active and healthy life”

Five indi­cators present in the FAO “Suite of indi­cators of food security” are derived from 3y moving averages of FBS data and are shown in Box 2.3. Note that none of these indi­cators provide infor­mation on the afford­ability, acces­sibility, or con­sump­tion by different popu­lation groups including nutri­tionally vulnerable groups. For more details, see: FAO (2022). The average dietary energy supply (DES) adequacy is defined as the dietary energy supply as a per­cent­age of the average dietary energy require­ment. The latter is defined as the amount of dietary energy mea­sured in kilo­calories that is required by an indi­vidual to maintain body functions, health, and normal activity. The estimates of DES adequacy consider the age structure, sex, and height in each country.

The dietary energy available from cereals, roots, and tubers (kcal/capita/day) (percent, 3y average) is an impor­tant nutritional indicator because items from these food groups are generally the least expensive. As a result, these staple foods are often a large part of the diet in low-income coun­tries. Conse­quently, the nutri­ent density, dietary diversity, and micro­nutri­ent supply of these diets are low as described in detail in Chapter 8c.

As noted earlier, average protein supply (g/capita/day, 3‑year average) provides insight into the nutritional quality of the food supply at the global, regional, or national level.

The average supply of pro­tein of animal origin (g/capita/day, 3‑year average) is an additional impor­tant indi­cator of diet quality as animal-based pro­teins contains all the essential amino acids. The supply of pro­tein from animal-based sources includes meat com­modities, eggs, and dairy products, and fish and seafood, all mea­sured in g/capita/day. Figure 2.4 presents the daily pro­tein supply from both animal and plant-based foods in 2020. The data from 1961 to 2019 show that there has been an increase in the share of total pro­tein from animal-based sources in most coun­tries since 1961 (Roser et al., 2023).

Additional indi­cators of the quality of the food supply based on FBS data have been devel­oped by INDDEX in their Data4Diets platform. They include meat con­sump­tion (expressed as total kilograms per capita at the national level), national energy avail­able from non-staples (i.e., all food items exclud­ing grains and tubers), and national fruit and veg­etable avail­ability in the food supply. For more details, see INDDEX Project (2018).

2.1.1c Macro­nutri­ent avail­ability at the national level

Macro­nutri­ents include carbohydrate, pro­tein, and fats, and are the major source of energy in the diet. Macro­nutri­ent supply per capita can be esti­mated by combining supply data from FBSs expressed in g/capita/day with a food com­posi­tion database for all primary and processed food products, as noted earlier. A common global food com­posi­tion database is avail­able from FAO and is used to estimate the nutri­ent supply data. The source of the nutri­ent com­posi­tion data in this global database is not cited and any variation in the nutri­ent com­posi­tion of food items across coun­tries is not considered.

As shown in Table 2.1, the supply per capita for energy is expressed as kcal/capita/day, whereas the macro­nutri­ent supply per capita for both pro­tein and fat are expressed as g/capita/day). Data are freely avail­able by year and country from FAO (2021c).

Figure 2.5 presents the global per capita supply of protein and shows that in lower-income coun­tries the per capita pro­tein supply in 2020 is about 60‑80g/capita/day, whereas in most coun­tries across Europe, North America, and Oceania, the per capita pro­tein supply is greater than 100g/capita/day.

The quality of the diet improves with the increased consumption of protein-rich foods. At the national level, the average pro­tein supply pro­vides an estimate of the pro­tein supply per capita and thus insight into the nutri­tional quality of the national food supply. However, no infor­mation on the avail­ability of pro­tein supply in nutri­tionally vulnerable groups is provide by the FBS data.

In contrast to both energy as kcal/capita/day and pro­tein g/capita/day, the per capita fat supply g/capita/day has increased across all regions during the period 1961 to 2014, although over the last decade, the increase has slowed in North America, Europe, and Oceania. Regional differ­ences in supply for fat are larger than that for either energy or pro­tein, with the average capita fat supply in North America being almost three times as large as in Africa in 2020 (Figure 2.6, Roser et al., 2023).

Macro­nutri­ents are sometimes expressed as per­cent­age of total energy (e.g. per­cent­age of kilo­calories from pro­tein or fat) due to uncertainties in the data and process used to develop FBS. Values expressed in this way (i.e., as nutri­ent density) are said to be much less influenced by age and sex in a popu­lation. Moreover, assess­ment of pro­tein supply as pro­tein density allows the quality of the diets to be com­pared across coun­tries which have differing levels of accuracy of the food supply data, and thus differing levels of under / over estimation of total nutri­ent per capita supply. Never­the­less, even com­par­ison of the values for per­cent­age of energy as well as per­cent­age of energy from pro­tein and fat (and any other nutri­ent) derived from each food group across coun­tries can be jeopardized by a differing level of under / over estimation between food groups. In some cases, values for the ratios of poly­unsatu­rated fatty acids to satu­rated fatty acids (P:S) and un­satu­rated fatty acids to satu­rated fatty acids (U:S) have also been used (Sasaki & Kesteloot, 1992) rather than per capita nutri­ent supply.

2.1.1d Micronutri­ent avail­ability at national level

Efforts have been made by several investigators to use FBS for the assess­ment of the adequacy of micro­nutri­ent supply at the global level, in different regions of the world, and in indi­vidual coun­tries. To accomplish this aim, FBS have been matched not only to energy, pro­tein, and fat, but also to a full range of micro­nutri­ents. However, because the FAO FBS do not include data on the micro­nutri­ent supply per capita at the present time, investigators have compiled their own data on the micro­nutri­ent com­posi­tion of the food com­modities listed in the FAO FBS using data from food composition tables.

As noted earlier, FAO FBS provide only national-level estimates per capita. This precludes the assess­ment of per capita micro­nutri­ent supply that takes into account age and sex and the differing micro­nutri­ent needs of the national population. However, by applying data from estimates on the age and sex distrib­utions of the national popu­lation in each country, the theoretical popu­lation mean require­ment for each micro­nutri­ent at the national level can be esti­mated. From this, the per­cent­age of the popu­lation at risk of inad­equacy has been calcu­lated, defined as the proportion of the popu­lation with micro­nutri­ent intakes below the theoretical average micro­nutri­ent require­ment of the popu­lation. The popu­lation distrib­ution of intakes for each micro­nutri­ent is inferred based on the assumed mean per capita micro­nutri­ent intake and coefficients of variation (CV) for each micro­nutri­ent (taken from the literature). The latter takes into account inter-indi­vidual variation in micro­nutri­ent intakes arising from the inequities which affect access and avail­ability of certain foods. This approach of assessing risk of inad­equacy is related to the Esti­mated Average Require­ment (EAR) cut-point method and described in Chapter 8b for a popu­lation.

Several investigators have employed this approach to estimate risk of micro­nutri­ent inad­equacy based on FBS data. Wuehler et al. (2005) were the first investigators to compare the adequacy of the food supply to meet the theoretical popu­lation mean require­ment for zinc, based on FBS for 176 coun­tries.

In this analysis, estimates of the absorb­able zinc content of the national food supplies were derived from a global food com­posi­tion database compiled by the investigators, with zinc absorption predicted using a model devel­oped by the International Zinc Nutri­tion Consultative Group (IZiNCG). In this study, the investigators com­pared the absorb­able zinc content of the food supply with the popu­lation theoretical mean physio­logical require­ment to estimate the per­cent­age of the popu­lation at risk of inadequate zinc intake. This earlier analysis was revised by Wessells & Brown in 2012. Figure 2.7 shows the esti­mated country-specific preva­lence of inadequate zinc intake gener­ated from this revised analysis. A companion article high­lights the major sources of uncertainty in the analysis (Wessells et al., 2012). They also com­pared their esti­mated preva­lence of dietary zinc inad­equacy with the national preva­lence of stunting in children less than 5y in 138 low‑ and middle-income coun­tries (Figure 2.8).

In later publications, this same approach has been applied to assess the per capita avail­ability of several other micro­nutri­ents in regions (e.g., Africa) (Joy et al., 2014) and South Asia (Mark et al., 2016), as well as in specific coun­tries (Broadly et al., 2012; Arsenault et al., 2015; Watts et al., 2015). For example, in a study by Arsenault and co-workers FBS data from three coun­tries (Bangladesh, Senegal, and Cameroon) were used to assess the per capita avail­ability for eight micro­nutri­ents (vitamin A, vitamin C, vitamin B₆, ribo­flavin, niacin, folate, calcium and zinc). Although a common food com­posi­tion database was compiled and used for all three coun­tries, additional country-specific adjustments were made in this study to take into consid­eration the in-country pro­cess­ing of staple foods, which lead to differ­ences in the nutri­ent com­posi­tion of ingested foods (e.g. per­cent­age of refined wheat at country level). Two indi­cators of adequacy were assessed in this study:

1. Per capita nutri­ent supply expressed as a per­cent­age of the theoretical mean popu­lation-adjusted nutri­ent require­ments for each country;
2. Per­cent­age of the popu­lation that had an adequate usual intake, based on the assumed inter-person variability in micro­nutri­ent intakes obtained from the literature.

Each nutri­ent gap in the food supply was also calcu­lated as the differ­ence between the current amount in the food supply minus the amount of the nutri­ent needed to achieve 80% preva­lence of adequate intakes. In this study, linear optimization modeling was also used to deter­mine the optimal mix of crops to fill the micro­nutri­ent gaps identified in each county and thus increase their micro­nutri­ent adequacy. Table 2.4 shows the optimal set of crops required to address the micro­nutri­ent gaps identified in Senegal while minimizing land require­ments.

Table 2.4 Optimal set of crops to address micro­nutri­ent gaps in Senegal while minimizing land require­ments^a
a The amounts are additional amounts needed to achieve target levels of all nutri­ents so that the food supply pro­vides sufficient amounts of nutri­ents (except calcium) for at least 80% preva­lence of adequate intakes in the popu­lation
^b Serving sizes are in edible form, i.e., all veg­etables are cooked, grains are dry but nutri­ents adjusted for cooking losses. Data from Arsenault et al. (2015).

Crops	Servingsb	Grams	Vit. A (µg RAE)	Vit. C (mg)	Ribofl. (mg)	Folate (µg)	Calcium (mg)	Zinc  (mg)	% crop land
Barley	0.50	23	0	0	0.07	4	8	0.6	1.8
Broccoli	0.50	39	30	25	0.05	42	16	0.2	1.0
Cabbage	0.50	38	2	14	0.01	11	18	0.1	0.4
Carrots	0.40	18	157	1	0.01	3	6	0.1	0.3
Groundnuts	0.50	14	0	0	0.02	27	10	0.5	3.8
Okra	0.50	40	6	7	0.02	18	31	0.2	0.8
Pumpkin	0.25	31	90	1	0.02	3	5	0.1	0.4
Spinach	0.25	23	118	2	0.05	33	31	0.2	0.6
Total	3.40	226	403	50	0.25	141	123	1.8	9.0
%adequate intakes	—	—	81	%95	%85	%84	<1%	%80	—

In an effort to overcome some of the lim­ita­tions of FBS data, notably their lack of infor­mation on both indi­vidual foods that are impor­tant in diets and actual intakes by age or sex, Smith et al. (2016) have con­structed a new model: the Global Expanded Nutri­ent Supply (GENuS) model to estimate per capita supply for 23 indi­vidual nutri­ents across 225 food groups for 34 age‑sex groups in 152 coun­tries. They also provide trends in national level data over50 years (1961‑2011).

To achieve these objectives, the GENuS model combines FBS data with some ancillary data from indi­vidual and house­hold surveys at country level. This approach yields food supply estimates which consider both the edible weight of each food com­modity and level of food fortification in 225 food groups. In each country an ad-hoc food com­posi­tion database was compiled based on six national and regional food com­posi­tion tables in which nutri­ent losses during the pro­cess­ing of cereals to flour were considered. Age and sex-specific supplies of foods and nutri­ents across 26 age groups were calcu­lated for each country. The distrib­ution of nutri­ent per capita supply by age and sex was derived through a probabilistic method (Monte Carlo simulations). Finally, to assess validity, a com­par­ison of the GENuS estimates of nutri­ent supplies against USDA estimates was performed, which revealed very good agree­ment for 21 of 23 nutri­ents. For a discussion of the lim­ita­tions of the data used in the GENuS model, see Smith et al. (2016).

2.2. Household Surveys

Household surveys assess food available for consumption at the household level. This is defined as the total amount of food and beverages avail­able for con­sump­tion in the house­hold plus any foods prepared and consumed outside the house­hold. House­hold surveys are of two types:

1. House­hold Food Con­sump­tion Surveys (HFCS). These were devel­oped in the 1980s as an alternative to surveys at the indi­vidual level, and provide measurements of food con­sump­tion in the house­hold.
2. House­hold Con­sump­tion and Expenditure Surveys (HCES). In these surveys, food avail­able for con­sump­tion is mea­sured. These household surveys are widely used to inform on global food security and poverty, particularly in low‑ and middle-income economies. Such surveys can also be con­ducted in institu­tions such as those for institu­tionalized elderly.

In the past, some surveys similar to HFCS or HCES were con­ducted in high-income coun­tries with the objective of assessing food con­sump­tion at the house­hold level. They were based on a record of food quantities entering the house­hold, either purchased, received as gifts, or produced for house­hold use, over a reference period, often seven days. They used the “food account method” Burk & Pao, 1976; or the “food inventory method&rdquo. (Turrini et al., 2001)

Food con­sump­tion by indi­vidual members of the house­hold was often recorded as an additional component of the survey. This approach led to an increased burden on the sampled house­holds, resulting in some cases in a low response rate. Such surveys are now rare and have been replaced by food con­sump­tion surveys at the indi­vidual level.

During HFCS and HCES, data on acquisition and/or expenditure and/or con­sump­tion of com­modities including foods and beverages are collected. HFCS and HCES are less expensive than indi­vidual food con­sump­tion surveys and require a lower burden for participants. However, many of the technical problems and lim­ita­tions of house­hold surveys are the same as those of dietary surveys conducted at the individual level, which are discussed in detail in Chapters 4 and 5. For example, diet might be altered by the design of the survey or the recording process.

Typically, HFCS and HCES are con­ducted on a large nationally rep­re­sen­tative sample of house­holds (Fiedler et al., 2012), and include the collection of infor­mation on demographic and socio­economic characteristics of the house­hold, thereby enabling data to be presented in terms of income level, family size, region of the country, etc. Attention must be paid to the sampling design of these surveys to ensure that a rep­re­sen­tative national sample is obtained; see details on sampling methods in Chapter 1, section 1.4.2. The sampling design should account for the influence of season, holidays, weekends, socio­economic status, and region on food con­sump­tion patterns.

2.2.1. House­hold Food Con­sump­tion Surveys (HFCS)

House­hold Food Con­sump­tion Surveys measure all food and beverages consumed within a house­hold during a specified period. They require careful supervision by the interviewer and good cooperation of the respon­dents. In general, these surveys are more complicated and costly to undertake than HCES and hence were often con­ducted infrequently. Such surveys are rarely per­formed today. Instead, they have been replaced by surveys of food con­sump­tion at the indi­vidual level which are more informative, even if more costly (see Chapter 3). Two methods — weighed food records and house­hold 24hr recalls — were used in the past to measure the house­hold food con­sump­tion.

Weighed house­hold food records

For this method, the food eaten by the house­hold is recorded using weighed food records (also called food diaries). These are usually completed over at least a 1wk period, by either the house­holder or a fieldworker. During the survey period, the weight or volume of each food consumed at each meal is recorded, before subdivision into indi­vidual helpings USDA. Detailed descriptions of all foods, including brand names, and their method of preparation are recorded. For composite dishes, the amount of each raw ingre­dient used in the recipe and the final weight of the prepared composite dish is also recorded. In some surveys the plate waste from each meal is collected and separated so that waste for indi­vidual food items can be weighed and recorded. Generally however, kitchen and plate waste, and food fed to pets, is not accounted for in this method, and instead an arbitrary wastage factor is applied.

House­hold 24-h recalls

In this method, the house­hold member responsible for the food preparation is interviewed to obtain infor­mation on both house­hold com­posi­tion and house­hold food con­sump­tion over the previous 24h period. In the first stage of the interview, infor­mation is collected on the dishes and ingredients con­sumed, followed by details on the quantity, focusing particularly on those foods that are impor­tant sources of energy. A technical guide for measuring house­hold food con­sump­tion using a 24h recall was devel­oped by Swindale & Ohri-Vachaspati (1999). This guide pro­vides detailed instructions and sample questionnaires that can be used to collect the data, quantify the portion sizes of food consumed, and analyze the results.

2.2.2 Household Consumption and Expenditure Surveys

House­hold Con­sump­tion and Expen­diture Surveys (HCES) are particularly impor­tant and typically performed within House­hold Expen­diture Surveys (HES), i.e. economic surveys designed to inform national economic policy and usually imple­mented by national statistical agencies.

HCES are less costly than HFCS and are widely used in low-income settings (FAO, 2023). They are usually con­ducted every 3‑5y. HCES refer to a heterogeneous group of complex surveys which include House­hold Income Expen­diture Surveys (HIES), National House­hold Budget Surveys (NHBS), and Living Standards Measure­ment Surveys (LSMS). The latter (i.e, LSMS) are multi-topic surveys for which technical assistance is usually provided by the World Bank's Living Standard Measure­ment Study (LSMS) group.

HCES were designed primarily to measure house­hold food avail­able for con­sump­tion in a reference period through the expenditure approach (i.e., the monetary value of the food). Over time, quantities of foods and beverages have also been collected in an effort to repur­pose the surveys so they can serve other needs such as monitoring house­hold-level food security. This has led to improvements in survey design to capture food data of a higher quality by including a more detailed food component module.

For some HCES based on the concept of acquisition, data are collected on food entering the house­holds acquired through purchases, own-produc­tion, and in‑kind, on the assumption that there were no major changes in house­hold stocks during the survey period. An impor­tant lim­ita­tion of this simplified method­ology is related to the fact that some foods (e.g. grains) are not perish­able and can be stored. Conse­quently, if the survey was con­ducted in a period of drawing-down stocks to meet current con­sump­tion, house­hold con­sump­tion would be underesti­mated. On the contrary, if the survey was con­ducted in a period of accumulating stocks for later con­sump­tion, house­hold con­sump­tion would be overesti­mated (Smith et al., 2014). As a result, use of these simplified HCES is no longer recom­mended. Alternatively, a combination of acquisition and con­sump­tion data can be collected in the HCES so that both food acquired through purchases and consumed from own-produc­tion and transfer are reported.

Food con­sump­tion estimates gener­ated from acquisition data or a combination of both acquisition and con­sump­tion data are usually referred to as “apparent con­sump­tion” or “food avail­able for con­sump­tion” to distinguish it from actual con­sump­tion (Fiedler & Mwangi, 2016). Data collected in HCES on food avail­able for con­sump­tion are quite often used as a proxy for the quantities eaten.

HCES vary in their complexity and respondent burden. Data may be collected through a food diary, recall, or a list, with a reference period usually ranging from seven days to one month. When a list is used, the food and food groups specified varies, and when the list is short, the aggregated food items often lack the details needed to match the food items correctly with a food com­posi­tion database. Food and beverages wasted, spoiled, or fed to pets or livestock are frequently not taken into account, although in some cases, waste may be weighed or a wastage factor applied.

In general, in the past the capture of food and beverages prepared and consumed outside the home has also been poor in most HCES, although there is now a trend to improve the recording of the con­sump­tion of these items. For example, in some recent HCES a special module has been added to collect personal expenditure on snacks, meals, sweets, and drinks consumed outside the home. Adjustments are also made in some HCES for the presence of non-house­hold members during the survey period. Ideally HCES should report all sources of food con­sump­tion, either purchased, received as gifts, or produced for house­hold use.

Variations in the survey designs of HCES across coun­tries and over time within coun­tries has created many challenges. To understand the implications of these variations on the accuracy and consistency of con­sump­tion measurements, a field study was con­ducted in Tanzania in 2007 and 2008 in which con­sump­tion estimates in 4,000 house­holds was mea­sured via eight different HCES survey designs, selected to capture the most common designs used in practice (Beegle et al., 2010). The effect of variations according to survey length, use of a house­hold diary versus recall format, and level of detail of food items listed in the HCES on over- and under-reporting was investigated, using frequently supervised personal con­sump­tion diaries as the “gold standard”.

The characteristics of the study popu­lation, notably whether they were illiterate or largely urban, was found to affect the magnitude of the under­estimate of con­sump­tion using the HCES diary method. For the HCES recall method, the investigators recom­mended the use of a full long list of food com­modities to recall within a reference period of 1 or 2 weeks rather than using a shorter recall list with few aggregated food categories. In summary, over‑ and under-reporting appeared to depend on the specific setting and varied with traits such as the literacy of the household, the share of con­sump­tion from home produc­tion, dietary diversity, and fraction of meals eaten in the house­hold as opposed to consumption by individuals at restaurants and catering establishments (Beegle et al., 2010).

The findings of Beegle et al. (2010) highlighted the need to improve the quality and utility of the food data components of the HCES surveys so they were more relevant to nutri­tionists and food security analysts, as well as the impor­tance of harmonizing the HCES survey method­ology. Conse­quently, a desk review of survey questionnaires and methods used in 100 avail­able HCES surveys from low‑ and middle-income coun­tries was con­ducted to identify key areas for improvement (Smith et al., 2014).

Five areas in the food data component of HCES surveys were selected for inves­ti­gation; these are summarized in Box 2.4. For a detailed discussion of the measure­ment error patterns that were identified within each of these five areas, see Zezza et al. (2017). As an example, a major challenge has been the rise in the con­sump­tion of food away from home, a trend that has accom­panied the nutri­tion transition over the last decades in both devel­oping and devel­oped coun­tries. Such data may include food that is purchased (e.g., from a street stall or restaurant), and/or received in kind (e.g., provided via food assistance or a gift etc). Snacks have also become an increasingly impor­tant part of the diet and may make up a large proportion of food consumed away from home.

Food away from home (FAFH) can refer to food produced outside the home such as takeout meals, irrespective of whether the food is consumed outside or inside the home. Alternatively, food away from home can refer to food consumed outside regardless of the origin of the food. Examples of this scenario includes homemade meals consumed at work or school. The FAO and the World Bank 2018 have outlined a method for defining food away from home, as shown in Figure 2.9.

For nutri­tion and food security analysis, an additional challenge, not captured in Figure 2.9, is acquiring infor­mation on what was eaten. Such details must often be secured from other sources. Several low‑ and middle-income coun­tries have devel­oped innovative approaches to secure these details; see Annex 2 in FAO/World Bank (2018) for more details.

Table 2.5 Food away from home data collection.
^a N=100 surveys. ^b Calculations are only for surveys for which any data were collected on food consumed away from home (N=90).
Data from Smith et al. (2014).

	Interview surveys (%)	Diary surveys (%)	All (%)
Whether any data collected on food consumed away from homea	83.3	100.0	90.0
Detail of data collectionb
Only one line item (e.g., “Restaurant food”)	36.0	7.9	23.9
Data collected for multiple places of con­sump­tion	14.0	35.0	23.3
Data collected on food received in-kind	46.0	65.0	54.4
Data collected on specific food items	28.0	40.0	32.9
Snacks explicitly referred to	26.0	35.1	29.9
Alcoholic beverages explicitly referred to	36.0	32.4	34.5
Data collected at the indi­vidual level	12.0	23.7	17.0
Recall period b
Less than one week	60	100.0	47.8
One week	48.0	0.0	26.7
Two weeks	12.0	0.0	6.7
One month	14.0	0.0	7.8
Greater than one month	20.0	0.0	11.1

Table 2.5 presents data on food consumed away from home based on data collected from the 100 HCES surveys examined by Smith et al. (2014). Such data may include food that is purchased (e.g., from a street stall or restaurant) and/or received in kind (e.g., provided via food assistance or a gift etc). Snacks have also become an increasingly impor­tant part of the diet and can make up a large proportion of food consumed away from home. As shown in Table 2.5, although most of the surveys collected data on food consumed away from home, the detail with which the data were collected was very poor.

In view of the many challenges experienced in collecting the food data component of HCES surveys, and the recom­men­dations proposed by several investigators (Carletto, Zezza & Banerjee, 2013; Smith et al., 2014; Zezza et al., 2017), guide­lines to improve the quality of data collected have been devel­oped by FAO and the World Bank (2018) and are itemized in Box 2.5.

These guidelines take into consid­eration the best balance between accuracy and cost effec­tive­ness. For more details on the recom­men­dations specified for each domain and their justification, see FAO and World Bank (2018). Low‑ and middle-income coun­tries are recom­mended to adopt these guidelines to improve the quality of the food data component collected in HCES and standardize the method­ology across coun­tries.

2.2.3 Uses of house­hold surveys

Increasingly, HCES are being repur­posed so that the data gener­ated are not only used by economists for monitoring poverty, calculating national accounts, and as an input for consumer price indices, but also by food security analysts and nutri­tionists for nutri­tion-related analyses, based on the food con­sump­tion module of the HCES surveys (Smith et al., 2014; FAO and the World Bank, 2018). The World Bank has devel­oped the most comprehensive repository of publicly available worldwide HCES data. An additional resource on HCES surveys is also avail­able at the International House­hold Survey Network (HSN).

Several resources are avail­able to derive food security and nutri­tion indi­cators as well as nutrient-based indicators based on the food module of HCES surveys. Of these, two resources itemized below provide details of their methods of con­struction, uses, strengths and weaknesses.

• Data4Diet platform compiled by the INDDEX project (2018) is freely avail­able at: Data4Diet.
• ADePT-FSM Food Security Module, a stand-alone application, based on the research of Moltedo et al. (2018) and devel­oped by the World Bank and FAO.

Readers are advised to consult these two resources. A selection of indi­cators designed to assess food security and nutri­tion are summarized in Box 2.6, whereas details of indicators designed to characterize dietary micronutrient supply are presented in Box 2.7. When compiling these indi­cators, the term “apparent consumption of foods” and “nutrient supply” at the household level and “apparent nutrient intake” at the indi­vidual level rather than “intake” should be used. None of these indi­cators permit assess­ment at the indi­vidual level.

Indicators designed to characterize household dietary nutrient supply Additional indi­cators designed to quantify, characterize, and evaluate the adequacy of household micronutrient supply across populations have also been compiled. These indi­cators are calcu­lated by first multiplying the edible amounts of the food com­modities listed in the food supply module of the HCES survey and said to be avail­able for con­sump­tion (not the actual intake) for each house­hold, with their corre­sponding nutri­ent content derived from food com­posi­tion tables (FCTs) (mostly for the raw form of the food before preparation). The data generated are termed “micro­nutrient supply”. The nutri­ents included are the essential amino acids, vitamins B₁, B₂, B₆, B₁₂, and C, folate, total vitamin A (expressed as both Retinal Equivalents and Retinol Activity Equivalents), zinc, calcium, and total iron. In addition, heme iron, expressed as a per­cent­age of total iron, can also be esti­mated.

Several investigators have com­pared the estimates of apparent nutrient intake derived from HCES data with those obtained with individual dietary assessment methods. In a systematic review of five studies, Tang et al. (2022) concluded that in general nutrient intakes were overestimated using HCES food consumption data. For some of the studies, discrepancies were large com­pared with the estimates from indi­vidual dietary assessment methods. For a discussion of the possible reasons for the poor agree­ment, see Tang et al. (2022).

Apparent micronutrient intake per capita per day and per adult-male equivalent per day as well as micronutrient density can be esti­mated at the national level, and disaggregated into categories when they can be used to monitor or identify target popu­lations for policy makers. The categories that can be included depends on the house­hold socio-economic characteristics collected in the HCES in the country and may include education, income, geography, and ruralness. Examples of these indi­cators disaggregated by region, urban-rural areas, and quintile of income are given in Box 2.7.

In the past, attempts to assess micro­nutri­ent adequacy from HCES survey data esti­mated ratios of apparent micro­nutrient intake per capita to the average weighted micro­nutri­ent require­ment at the popu­lation level. These ratios provided a gross indication of issues with regard to meeting require­ments of selected micro­nutri­ents at the popu­lation level, but provided no infor­mation on the preva­lence of inadequate micro­nutri­ent intakes in the popu­lation (Moltedo et al., 2018).

Conse­quently, recently efforts have been made to estimate the preva­lence of nutri­ent inad­equacy (PoNI) in the popu­lation based on HCES survey data. The new method extends the FAO probabilistic cut-point method used earlier to estimate chronic dietary energy inad­equacy to include micro­nutri­ents. To achieve this objective, the data gener­ated on apparent micronutrient intake per day at the indi­vidual level must first be pre-adjusted to yield the distrib­ution of apparent usual micro­nutrient intakes by removing the excess variation due to both within-person day-to-day variability and seasonal variability. After these preadjustments, the distrib­ution of apparent usual intake levels is com­pared to a threshold to estimate PoNI. The threshold is based on the weighted average of the esti­mated average require­ment (EAR) for the micro­nutri­ent under study for each sex-age group in the popu­lation.

The prevalence of inadequacy for eight micronutrients: vitamins A,  B₁, B₂, B₆, B₁₂, and C, and calcium and zinc based on pre-treated household survey data versus individual level dietary data from Bangladesh has been com­pared by Moltedo et al. (2022). Their estimates of inadequacy from the household survey data com­pared with the individual-level data in which the conventional EAR cut-point method was applied, appeared promising. Nevertheless, more research is needed to guide the use of HCES to estimate apparent nutrient intake at the individual level and prevalence of inadequacy. Detailed instructions for estimating these indictors based on nutri­ent analysis from FCTs and the new FAO probabilistic cut-point method, respectively, are avail­able at:

• Upgrading ADePT-FSM for analyzing food security statistics using house­hold con­sump­tion and expenditure survey data
• Moltedo et al. (2022)

Special attention must also be given to the choice of the food composition tables used for the micro­nutri­ent analysis. For discussion of some of the lim­ita­tions of the micro­nutri­ent values in food com­posi­tion tables, consult Section 1.2 in the above ADePT-FSM Version 3 Software. Note that this version of the ADePT software also permits an assess­ment of the micro­nutri­ent content in foods consumed away from home.

2.3 Comparison of Food Balance Sheets (FBS) and
House­hold Consumption and Expenditure Surveys (HCES)

Both FBS and HCES are prone to measure­ment errors, as itemized in Table 2.7. For more details, see Thar et al. (2020) and FAO (2001) for FBS, and Fiedler et al. (2012) and Smith et al. (2014) for HCES.

Table 2.7. Comparison of Food Balance Sheets and Household Consumption and Expenditure Surveys. From FAO (2022a).

	Food Balance Sheets	Household Consumption and Expenditure Surveys
Objective	Assessment of food available for consumption at national level	Assessment of food available for consumption at the household level
Advantages	• Not expensive • Publicly available for over 245 countries    & territories from 1961 • Snapshot of overall agri-food situation • Updated regularly with standardized    method­ology so compatible for com­par­ison    and trends of per capita food supplies    in terms of quantity, caloric value,    protein & fat content	• HCESs conducted regularly &    publicly available • More detailed in sub-national,    sex, economic strata breakdown • Fortfiable foods can be identified    from percentage of HHs consuming    & purchasing indi­vidual food • Report metrics that characterise    food security, dietary quality,    household nutrient supply, & apparent    nutrient intake at indi­vidual level
Disadvantages	• Quality of FBSs limited by completeness    and accuracy of in-country reports; data    on country's stocks not always available • Food loss, food waste, homegrown food,    & food fed to animals is under-recorded • Number of migrants & tourists not always    subtracted from population estimates • Wrong matching of foods with FCTs as    ambiguous details of foods listed • Nutrient supply data estimated from a    common global FCT so differences in    composition of foods across countries    not considered • Distribution of food at sub-national level    or population specific characteristics such    as age, sex, SES not provided so high risk    groups cannot be identified • No data on home-grown food or details    of processed foods	• Relatively expensive & requires insti-    tutional capacity & trained personnel • Household food supply module not    yet standardized across countries • Indi­vidual data on food purchases,    food consumed from home    production, food received in kind,    and details of food consumed away    from home not always collected • At-home food list varies (50-300    items) & not always specific leading    to incorrect matching with FCTs • Seasonality of food consumption    patterns not always considered • Food given to non-HH members    not always reported • HCESs do not provide information    on distribution of foods consumed    among household members    or by life-stage groups • Apparent male equivalent intake    estimates limited by lack of data on    physical activity, physiological state of    household members

Grünberger (2014) com­pared per capita food supply based on country-specific FBS with that of 64 HCESs from 51 low‑ and middle-income coun­tries, for 16 major food groups. Overall, the results suggested that the esti­mated differ­ences in the average total food supply per capita are moderate. However, under­estimates for the contribution of the average con­sump­tion of cereals, eggs, fish products, pulses and veg­etables are likely, whereas over­estimates for fruits, meat, milk, and sugar products are probable. These findings suggest that con­sider­able uncertainty may exist in the estimates for the avail­ability for con­sump­tion of single food groups at country and global level.

2.4 National and house­hold surveys in the global assess­ment of food security and diet quality by age and sex

Where feasible, it is preferable to use quantitative food con­sump­tion data collected at the indi­vidual level to assess the distrib­ution of usual intakes of both food and nutri­ents in different popu­lation groups within a country. In this way, the adequacy of the nutri­ent intakes at the popu­lation level can be evaluated using the methods described in Chapter 8b. Such quantitative data from food con­sump­tion surveys collected at the indi­vidual level can be accessed from the FAO/WHO GIFT platform.

However, currently, quantitative national food con­sump­tion data collected at the indi­vidual level are only avail­able for a few low‑ and middle-income coun­tries. Therefore, to fill this gap, alternative global databases based on infor­mation synthesized from FBS, HCES surveys, and indi­vidual-level surveys (where possible) have been compiled. Such databases are aimed at modeling worldwide indi­vidual intakes of foods and nutri­ents by age, sex and physiological status. Examples of these global databases include the Global Dietary Database (GDD) and a database devel­oped by the Institute for Health Metrics and Evaluation (IHME) discussed briefly below. For more discussion of these global databases, see Chapter 3.

The GDD database pro­vides modeled data in 188 coun­tries for 55 dietary factors including 14 foods, 7  beverages, 15 macro­nutri­ents, 19 micro­nutri­ents, and two indices of carbohydrate quality. The FoodEx2 categorization system is used to standardize the description and classification of foods into food groups. Modeled data are provided by country, year of primary data collection, age across the life span, sex, education level, urban or rural residence, and pregnancy or nursing status. The GDD modeled estimates can also be used to compile global patterns of healthy and unhealthy diets together with indi­cators such as the Healthy Eating Index and the Mediterranean Diet Score, Minimum Dietary Diversity for Women (MDDW) and for Infant and Young Child Feeding (IYCF). See Chapter 8c for more details of these indi­cators.

The IHME initiative database aims to provide rigorous and comparable measurements of the world's most impor­tant health problems and evaluates strategies used to prevent them. IHME uses FAO Food Balance Sheet estimates, national product sales, household surveys, and data based on 24 hr recalls (considered the gold standard). Datasets created by IHME are stored in the IHME data catalogue known as the Global Health Data Exchange and can be freely downloaded from the IHME website. The IHME initiative provides modeled data by country, age, sex and year, based on primary data collected in 204 coun­tries and 87 indi­cators, including 15 dietary indi­cators (9 foods and 6 nutri­ents). These dietary indi­cators are included in the Global Burden of Diseases (GBD), a worldwide observational epidemiological study that tracks the progress within and between coun­tries of the changing health challenges (Lim SS et al. 2012; GBD 2017 Diet Collaborators). For example, Qiao et al. (2022) investigated the global burden of non-communicable diseases attributable to dietary risks from 1990‑2019. They reported that a high intake of sodium and low intake of whole grains and fruits were leading dietary risks for deaths and disability-adjusted life-years (DALYs) worldwide, especially in devel­oping coun­tries and among males. Their findings highlight the need to raise public awareness of interventions and improve dietary practices aimed to reduce the disease burden caused by dietary risk factors. Data from the GBD study are also used to monitor progress in coun­tries at the national and subnational level towards meeting the United Nations 33 health-related Sustainable Devel­opment Goals. Key research papers based on the global health estimates from the analysis of GBD data are published each year in a special issue of The Lancet.

Acknowledgments

RSG is grateful to Michael Jory for the HTML design and his tireless work in directing the trans­ition to this HTML version.