S2.1 Methodology for estimating the prevalence of undernourishment for 2020 to 2023

As is always the case in this report, the prevalence of undernourishment (PoU) and the number of undernourished people (NoU) referring to the most recent years are nowcasted due to lack of direct information on the most recent values of each of the elements that contribute to their computation; in other words, they are predictions of the very recent past.

As already noted in the last two years’ editions of this report, 2020, 2021 and – to a lesser extent – 2022 were unique in many respects due to the COVID-19 pandemic and its lingering effects. This demanded special considerations when nowcasting the values of the PoU, especially with respect to estimating the likely change in the coefficient of variation (CV), considering the very special conditions under which food systems operated during the pandemic.

The strategy used to project values of the CV|y from 2019 to 2022 was based on assumptions regarding the way in which inequality in access to food contributes to rates of undernourishment, as fully described in last year’s edition of the report. This was not changed this year; however, as the world slowly returned to more normal conditions, the operation of household surveys was resumed in many countries. As a result, microdata from household surveys conducted after 2020 in nine countries became available to the Food and Agriculture Organization of the United Nations (FAO), which allowed for a reassessment of the values of the CV that had previously been modelled. Moreover, to nowcast the CV|y in 2023, the way in which inequality in access to food may have contributed to the levels of food consumption and undernourishment was reconsidered to reflect the gradual reversion towards normal estimation procedures.

One consequence of this is that the extent of uncertainty surrounding the estimates of PoU and NoU in 2021 and 2022 decreased, and the one remaining around the estimates for 2023 is considerably smaller compared to the years immediately following the COVID-19 pandemic.

Table S2.1 presents the lower and upper bounds of the PoU in 2022 and 2023 at the global, regional and subregional levels.

TABLE S2.1Ranges of prevalence of undernourishment and number of undernourished nowcasted in 2022 and 2023

NOTES: PoU = prevalence of undernourished; NoU = number of undernourished; n.r. = not reported, as the prevalence is less than 2.5 percent. For NoU, regional totals may differ from the sum of subregions, due to rounding and non-reported values. For country compositions of each regional/subregional aggregate, see Notes on geographic regions in statistical tables at the end of the main report.
SOURCE: Authors’ (FAO) own elaboration.

S2.2 Methodology for projections of prevalence of undernourishment to 2030

To project PoU values to 2030, the three fundamental variables that enter in the PoU formula (dietary energy consumption [DEC], CV and minimum dietary energy requirement [MDER]) are projected separately, based on different inputs, depending on the scenario considered.

The main source of information is the output of the MIRAGRODEP recursive, dynamic computable general equilibrium model, which provides series of projected values, at the country level, for:

• real per capita gross domestic product (GDP) (GDP_Vol_pc);
• income Gini coefficient (gini_income);
• an index of real food prices (Prices_Real_Food);
• extreme poverty headcount rate (that is, the percentage of the population with real daily income below USD 2.15 (x215_ALL); and
• daily per capita dietary energy supply (DES_Kcal).

The MIRAGRODEP model was calibrated to the pre-pandemic situation of the world economy in 2019 and was used to generate projections of macroeconomic fundamentals into 2020–2030 under two scenarios: i) “before COVID-19”, which aims to capture the implications for food availability and access (and therefore the PoU) of the world economic prospects as seen before the eruption of the pandemic by the International Monetary Fund (IMF) World Economic Outlook published in October 2019; and ii) “current prospects”, which is based on the latest World Economic Outlook published in April 2024.¹ A more detailed description of the MIRAGRODEP model, as well as the assumptions used to build the various scenarios, can be found in Laborde and Torero (2023).²

In addition, we use the median variant projections of total population (both sexes), its composition by gender and age, and the crude birth rate as provided by the 2022 revision of the United Nations Department of Economic and Social Affairs World Population Prospects.³

Projections of dietary energy consumption

To project the series of DEC we use the following formula:

where T = 2019 for “before COVID-19”,

and T = 2023 for “current prospects”.

In other words, we take the model-projected series of DES_Kcal and adjust its level so that the value for year T matches the actual value. (This is necessary as the MIRAGRODEP model has been calibrated to an older Food Balance Sheet [FBS] series.)

Projections of the minimum dietary energy requirement

To project the MDER, we simply compute it based on the data on the composition of the population by sex and age as projected by the 2022 World Population Prospects⁴ (medium variant).

Projections of the coefficient of variation

As explained in the methodological note on the PoU in Annex 1B of the main report, the total CV is computed as , where the two components refer to variability in the per capita habitual dietary energy consumption due to differences across households in terms of income level and variability across individuals based on differences in sex, age, body mass and physical activity level. The projected values for CV in 2025 and 2030 are obtained by applying the formula above to the CV|r and CV|y projected separately. Projected CV|r is computed based on the projected population structures by sex and age as provided by the World Population Prospects (similarly to what we do for the MDER), while the projected CV|y is computed as a linear combination of relevant projected macroeconomic and demographic variables as follows:

Table S2.2 reports the estimated regression coefficients.

TABLE S2.2Estimated coefficients from a regression of historical CV|y values on a set of covariates, 2000–2018

NOTE: CPI = consumer price index; CV = coefficient of variation; GDP = gross domestic product.
SOURCE: Authors’ (FAO) own elaboration.

The series of CV|y values predicted by the formula separately for each country for the years T + 1 to 2030 is then calibrated to the value for year T, similarly to what is done for the DES:

where T = 2019 for “before COVID-19”,
and T = 2022 for “current prospects”.

S2.3 Methodology for the analysis of food insecurity by degree of urbanization and by gender

The prevalence of food insecurity can be disaggregated by respondent/household characteristics when the data are collected directly from individual respondents in nationally representative samples. In Chapter 2 of the main report, food insecurity estimates are presented disaggregated by sex of the respondent (adult men or women) and by Degree of Urbanization (DEGURBA) (i.e. urban, peri-urban or rural residency).

The methodology to disaggregate the indicator by any individual or household characteristic is as follows:

• The cross-country comparable probability of food insecurity for each respondent is computed at two levels of severity: moderate or severe, and severe only. The probabilities are aggregated for each category of the characteristic of interest, by computing the weighted average (using sampling weights) across all respondents in that category, obtaining the prevalence of food insecurity within that group (e.g. among female respondents).
• The prevalence of food insecurity in a given category is weighted by the corresponding population (e.g. the number of female adults in the country) to obtain the subregional/regional/global estimate (e.g. the prevalence of food insecurity in the female adult population in Northern Africa), if reliable population data are available and if there is sufficient geographical coverage (at least 50 percent) in terms of percentage of the population.

The computation of the prevalence of food insecurity by sex is possible because data are collected from individual respondents (adults aged 15 years or older) by FAO via data collection service providers (see Annex 1B of the main report). For countries for which national government survey data are used to calculate the prevalence estimates of food insecurity (see Annex 1B of the main report), it is generally not possible to disaggregate the indicator by sex, as data are collected at the household level. Thus, in such cases, the same relative difference by sex estimated based on data collected by FAO is applied to the prevalence of food insecurity in the total population based on national data. This is an approximation, as the difference in the FAO data applies to adult respondents, and not to the whole population. However, the benefit is that the statistics by sex are consistent in terms of levels and trends with those of the overall population.

The disaggregation by DEGURBA is possible because Gallup^© began to georeference each interview in countries collected using face-to-face mode in 2021. Since 2022, countries covered by telephone interviews were also georeferenced, providing enough geographical representation to produce subregional, regional and global food insecurity estimates by DEGURBA.

Within each country, it is possible to link each georeferenced observation to the DEGURBA dataset, defining whether the observation (respondent) is located in a city, town or rural area, based on population density and size, according to internationally comparable criteria developed by the Statistical Office of the European Union (EUROSTAT), the International Labour Organization (ILO), FAO, the Organisation for Economic Co-operation and Development (OECD), the United Nations Human Settlements Programme (UN-Habitat) and the World Bank and approved at the 51st session of the United Nations Statistical Commission in March 2020.⁵ The prevalence of food insecurity is computed for each category of urbanization and then aggregated at the subregional, regional and global level using the 2020 updated DEGURBA population distribution published by EUROSTAT.⁶ For countries where official food insecurity statistics are informed by national data, the same approximation method described for the disaggregation by sex is applied.

As no Food Insecurity Experience Scale (FIES) data were collected by FAO in China in 2022 and 2023, and the data collected in 2021 were not georeferenced, the estimates of food insecurity by DEGURBA in China were approximated as follows: the prevalence of food insecurity for 2021 was disaggregated by area of residence as defined in the Gallup© World Poll (GWP), where respondents report if they live in: a rural area or on a farm; a small town or village; or a large city or the suburb of a large city. Then, these categories were mapped to DEGURBA by considering people living in a rural area or on a farm as part of the “rural” population, those living in a small town or village as part of the “peri-urban” population, and those living in a large city or the suburb of a large city as “urban” residents. This mapping was justified with the rationale that DEGURBA classifies areas with increasing urbanization based on population density and size. To ensure that no significant bias was induced by this approach, the same mapping was validated as accurate for other Asian countries where data were collected in 2022.

S2.4 Methodology for the analysis in Box 3 of the main report: Is food insecurity associated with the properties of a healthy diet? Preliminary analysis from 28 countries

The objective of the analysis presented in Box 3 of the main report was to examine the association between food insecurity severity and selected properties of a healthy diet, using food insecurity and dietary data collected from the same respondents in 28 countries between 2021 and 2022.

Datasets

The food insecurity data was collected using the FIES survey module (individual-referenced module with a one-year reference period). Food security data have been collected annually by FAO using the FIES survey module since 2014 through the GWP. The dietary data were collected using the Diet Quality Questionnaire (DQQ), developed by the Global Diet Quality Project, a collaborative effort of Gallup©, Harvard University and the Global Alliance for Improved Nutrition.⁷ The DQQ gathers data on intake of 29 food groups using a sentinel foods list. Beginning in 2021, the DQQ has been included in the GWP in a growing number of countries.

Only countries and data collection rounds in which both types of data were collected from the same respondents aged 15 years and over were considered for the analysis. Datasets from 28 countries were used, including 16 countries in Africa, seven in Asia, three in Latin America, and two in Northern America and Europe. Of these, 20 are low- or lower-middle-income countries and eight are upper-middle-income or high-income countries, as based on the World Bank country income classification for fiscal year 2024.

The data considered in these analyses were collected in the GWP in 19 countries in 2021 and nine countries in 2022. In 2021, the countries were Benin, Bolivia (Plurinational State of), Burkina Faso, Cameroon, Ecuador, Egypt, Gabon, Ghana, Kenya, Mozambique, Nigeria, Senegal, Sierra Leone, South Africa, Türkiye, Uganda, the United Republic of Tanzania, the United States of America and Viet Nam. In 2022, they were Afghanistan, Albania, Armenia, Honduras, Kyrgyzstan, Malawi, Palestine, Tunisia and Uzbekistan.

Definition of variables

For food insecurity, a trichotomous variable was created using the respondent-level estimated probabilities of food insecurity, based on the FIES global reference scale. Respondents were classified as:

• food secure or mildly food insecure if the probability of moderate or severe food insecurity was less than 0.5;
• moderately food insecure if the probability of moderate or severe food insecurity was equal to or greater than 0.5 and the probability of severe food insecurity was less than 0.5;
• severely food insecure if the probability of severe food insecurity was equal to or greater than 0.5.

The following metrics for the properties of a healthy diet, derived from the DQQ, were considered:

• Minimum Dietary Diversity for Women (MDD-W), computed only for women 15 to 49 years of age: equal to one if the woman had consumed foods from at least 5 out of 10 listed food groups^a (indicating a minimally acceptable level of dietary diversity) during the 24 hours preceding the interview, and zero otherwise.
• Zero vegetable or fruit (ZVF), computed for all respondents: equal to one if the respondent had not consumed any fruits or vegetables during the 24 hours preceding the interview, and zero otherwise.
• Animal-source food (ASF), computed for all respondents: equal to one if the respondent had consumed at least one animal-source food during the 24 hours preceding the interview, and zero otherwise.
• NCD-Protect score, computed for all respondents: ranging from 0 to 9, based on intake of foods from nine food groups^b composed of foods containing dietary factors that are protective against non-communicable diseases (NCDs). A higher score indicates inclusion of more health-promoting foods in the diet.
• NCD-Risk score, computed for all respondents: ranging from 0 to 9 based on intake of foods from eight food groups^c containing dietary components that should be limited or avoided as per global dietary recommendations. A higher score reflects higher consumption of foods and drinks to avoid or limit.

Analysis

To study the association between food insecurity and properties of a healthy diet, all the data were pooled, and two distinct analyses were conducted:

• Associations between food insecurity severity and adherence to properties of a healthy diet were studied by computing each of the following according to three categories of food insecurity (food secure/mildly food insecure; moderately food insecure; severely food insecure):

• –the weighted proportion of women aged 15 to 49 years achieving MDD-W;
• –the weighted proportion of all respondents consuming ZVF;
• –the weighted proportion of all respondents who consumed any ASF;
• –the weighted average of NCD-Protect score;
• –the weighted average of NCD-Risk score.

• Regression models, to account for potential confounding effects:

• –Separate logit regression models were estimated using MDD-W, ZVF, ASF (all binary indicators) as dependent variables and food insecurity status as the independent variable, along with income quintiles, education, sex, country and urban/rural residence of the respondent considered as potential confounding variables.
• –Separate ordered logit regression models were estimated using NCD-Risk and NCD-Protect (considered as ordinal nominal variables) as dependent variables, and the same set of independent and potential confounding variables described in the point above.

S2.5 Methodology for updating the estimates of the cost of a healthy diet

The International Comparison Program (ICP) coordinated by the World Bank is currently the only source of retail food price data for internationally standardized items, with new data only available once every three to four cyears. In previous editions of this report, the reference ICP data series was the one published in 2020, reflecting 2017 prices.⁸ This year, the cost of a healthy diet (CoHD) indicator is based on the last ICP release of 2024, which reports on 2021 prices.

The latest ICP 2024 round is chosen as the reference to update the cost this year because it includes the most recent food prices, reflecting food price patterns in post-COVID-19 pandemic years. Furthermore, the list of food items in the ICP 2024 wave is more comprehensive compared to the previous wave, as it collects prices for additional food items, including green leafy vegetables, which are a relatively cheap and nutritionally rich options for vegetables, especially in poorer countries.

Switching from the 2020 to the 2024 ICP data release to compute the CoHD may therefore also affect the composition of the reference Healthy Diet Basket (HDB) based on the new price information. For each country, the new composition of the HDB may therefore differ from the one previously used, since the list of items for which prices are collected in the two waves may differ and the prices reported for each item may also differ. Although the energy and nutritional contents of the HDB remain unchanged, the least-cost locally available foods that are selected among those that belong to a given food group may differ due to the expanded coverage of the ICP 2024 round, or because the least expensive items have changed. For this reason, readers should avoid directly comparing the series published this year with those in previous editions of the report.

In terms of how the series is constructed, the CoHD for 2021 is directly computed using ICP data, while it needs to be estimated for the years 2017, 2018, 2019, 2020 and 2022, when direct information on food items’ prices is not available. Estimated costs are obtained by inflating or deflating the 2021 prices using the consumer price index (CPI) for food and beverages.⁹

Specifically, to estimate the CoHD in year t, expressed in purchasing power parity units (c(PPP)_t), the 2021 CoHD expressed in the local currency unit (c(LCU)₂₀₂₁) is first multiplied by the ratio between the food CPI in year t and that in 2021 (we denote this as FCPI ratio^t₂₀₂₁), and finally divided by purchasing power parity conversion factors in year t (PPP_t):

where t = 2017, 2018, 2019, 2020, 2022 and:

Due to data limitations, the CoHD is updated using the aggregate CPI for food and beverages in the years when ICP food price data are unavailable. However, the food CPIs reflect average price changes for a basket of various food items defined in each country which may not accurately represent the price changes of foods in the HDB. In fact, by definition, the HDB is designed to include the cheapest nutritious foods that compose a healthy diet. This means that using the aggregate food CPI may lead to an overestimation of the CoHD. Further research is being conducted to construct a price index that reflects well the HDB composition.

S2.6 Methodology for estimating the unaffordabilty of a healthy diet

Conceptually, affordability is the condition by which households or individuals control sufficient resources to be able to procure the foods necessary to sustain consumption of a healthy diet. To operationalize this concept, the percentages and number of people in a population unable to afford a healthy diet are estimated by contrasting the distribution of incomes in the population with a fixed, normative cost threshold representing the amount of money needed to acquire the lowest cost combination of the locally available foods that are needed to compose a healthy diet, as well as all other non-food goods and services that are essential to conduct a dignified life.

In principle, as prevailing prices for both food and non-food goods and services vary by location, the ideal unit of analysis is the largest possible (usually subnational) area where the cost thresholds can be deemed equal for all the residing population.

Formally, the prevalence of unaffordability in an area s (PUA_s) can be estimated as follows,

Equation 1:

where r_s represents the fixed normative cost threshold, composed of the sum of the cost of a healthy diet (c_s) and the cost of non-food essential needs (n_s), and where f_s(x) is the distribution of incomes among residents of the area considered. Then, the number of people unable to afford a healthy diet is simply computed as the product between the PUA_s and the population size N_s:

NUA_s = PUA_s × N_s

Then, national estimates of NUA can be obtained by summing NUA_s over all relevant areas s, and PUA as the ratio between NUA and the national population size:

In practice, often the assessment can only be done for the entire national population or at the level of subnational geographic areas (such as urban vs rural, or administrative area units) that are larger than the ideal. This is because either the income distributions or the average costs – or both – are only available at that level. In such cases, one may still conduct the assessment using the formula in Equation 1, with reference to the national income distribution, f(x), and the national average threshold , as follows:

Equation 2:

recognizing that the formula in Equation 2 will yield an unbiased estimate of the true PUA only if the distribution of subnational area level values of r is statistically independent of the distribution of incomes across the same subnational areas. In all other cases, the threshold as used in Equation 2 will generate a bias (see Box S2.1). Given that the existence and the sign of the spatial correlation between incomes and the appropriate cost thresholds are empirical questions, research is being conducted on a large number of datasets from recent surveys to determine the best approach to adjust the threshold and correct for the potential bias that is affecting the current estimates.

BOX S2.1Why using the national average cost of a healthy diet and of essential needs may yield biased estimates of the prevalence of unaffordability

If there is a systematic relation between income levels and the sum of the cost of a healthy diet plus the cost of non-food essential needs, using the average as a threshold (as in Equation 2), will over- or underestimate the true prevalence of unaffordability (PUA), depending on whether the relation is positive or negative and whether the average falls above or below the modal income. This is because, when a correlation exists between incomes and costs across the different locations in a country, use of the average threshold instead of the appropriate differing thresholds will induce misclassifications in each of the areas that will not cancel out in the aggregate. The reason is quite simple: in these models, where the issue boils down to evaluating areas under a probability density function, the probability of misclassifying units at the two opposite sides of the threshold is not equal, except in the (very special) case that the threshold happens to fall in a region where the probability density function is flat. When the true thresholds are located in areas of the distributions where the income probability density functions are increasing, using the average as a threshold will lead to an overestimation of the PUA, if average incomes and food and essential needs costs are positively correlated, and to an underestimation in the (less likely) opposite case.

To illustrate, Figure A shows income distributions for two hypothetical subnational geographic regions in a country, one where incomes are systematically higher and the other where they are systematically lower. The figure shows why errors in estimating the PUA, using an average as a threshold, do not cancel out if the correct values to use as thresholds are correlated with incomes.

FIGURE A Overestimation and underestimation of the prevalence of unaffordability when using the average as a threshold if incomes are correlated with costs

NOTE: The green line refers to the income distribution in a poorer geographical subregion and the red line refers to the income distribution in a comparatively richer subregion.
SOURCE: Authors’ (FAO) own elaboration.

In this edition of the report, estimates of the PUA are computed using Equation 2, inferring income distributions from querying the World Bank Poverty and Inequality Platform, which provides an estimate of the percentage of the population with income below any specified threshold. For each year, the threshold provided is generated as the sum of the country-specific and of an estimate of the amount of money needed for essential non-food goods and services:

where the superscript c indicates the country and g the country income group.

The latter is estimated based on the values of the poverty lines used by the World Bank to compute estimates of extreme poverty, assuming a certain share of income spent on food that differs by country income group. Four different values of n are used depending on the country’s latest classification as low, lower-middle, upper-middle or high income by the World Bank, as per Table S2.3.

TABLE S2.3Calculation of the component of the cost threshold that corresponds to essential non-food goods and services

NOTE: PPP = purchasing power parity.
SOURCE: Bai, Y., Herforth, A., Cafiero, C., Conti, V., Rissanen, M.O., Masters, W.A. & Rosero Moncayo, J. (forthcoming). Methods for monitoring the affordability of a healthy diet. FAO Statistics Division Working Paper. Rome, FAO.

S2.7 Methodology for projecting global nutrition indicator estimates to 2030

Methodology for stunting, anaemia, low birthweight, overweight, exclusive breastfeeding and wasting

The World Health Organization (WHO) and the United Nations Children’s Fund (UNICEF) use the following approach to project estimates beyond the latest year of available data. This methodology is based on applying the average annual rate of reduction (AARR) from the baseline year to the latest year with available data. For these indicators, the baseline year is 2012 in accordance with the baselines set forth in World Health Assembly (WHA) Resolution 65.6 – Comprehensive implementation plan for maternal, infant and young child nutrition.¹⁰ For stunting, overweight, exclusive breastfeeding and wasting, the latest year is 2022. Anaemia and low birthweight have the latest years of 2019 and 2020, respectively.

Calculating the average annual rate of reduction

The AARR is calculated using a linear regression analysis. The dependent variables are the natural log transformations of all data points from the baseline to the latest year. The independent variables are all the years of available data. The coefficient of this linear regression (β) can be translated into the average annual rate of reduction by using the formula:

Equation 1: AARR = 1 – e^β

Projecting estimates to 2030 based on average annual rate of reduction

The following formula is used when projecting estimates forwards beyond their latest estimate based on the AARR. The projected prevalence for year t_n, baseline year is t₀:

Equation 2:

Methodology for adult obesity

The WHO uses the following approach when projecting estimates for adult obesity beyond the latest year of available data. The methodology is based on applying the average change in probit-transformed prevalence from the baseline year to the latest year with available data. For adult obesity, the baseline is 2010, in accordance with the baselines set forth in WHA Resolution 66.10 – Follow-up to the Political Declaration of the High-level Meeting of the General Assembly on the Prevention and Control of Non-communicable Diseases¹¹ and the Global Action Plan for the Prevention and Control of Non-Communicable Diseases.^{12, 13} Estimates are based on the outputs of the Bayesian model, which WHO uses to estimate prevalence of adult obesity for each year, country, age and sex. First, the age-standardized prevalence of obesity is computed for ages 18 years and over for each year, each sex and both sexes, and for each country, for every Bayesian model iteration. Then following regression is fit to the estimates for 2010–2022 separately for each country-sex-iteration unit:

Equation 3: probit(prev) = α + β × year

The probit transformation is the inverse cumulative standard normal distribution function.

Projecting estimates to 2030

The following formula is used to project estimates forwards beyond their latest estimate for each year t_n, country, sex and iteration. The projected prevalence for year t_n, latest year (2022) is t₀:

Equation 4:

Projected values for each year, country and sex are computed as the average of all iterations.

S2.8 Methodology for assessing country-level progress towards the global nutrition targets

Universe of countries

The analysis of countries’ progress towards achieving the 2030 target(s) is based on the 195 countries that are common to the universe of countries for the indicators that are modelled (i.e. stunting, anaemia, low birthweight, overweight, adult obesity) and the reporting universe for the indicators that are based on primary data (i.e. exclusive breastfeeding, wasting).¹⁴ This is done to ensure a consistent comparison across all indicators. The countries and areas in the analyses are: Afghanistan, Albania, Algeria, Andorra, Angola, Antigua and Barbuda, Argentina, Armenia, Australia, Austria, Azerbaijan, Bahamas, Bahrain, Bangladesh, Barbados, Belarus, Belgium, Belize, Benin, Bhutan, Bolivia (Plurinational State of), Bosnia and Herzegovina, Botswana, Brazil, Brunei Darussalam, Bulgaria, Burkina Faso, Burundi, Cabo Verde, Cambodia, Cameroon, Canada, Central African Republic, Chad, Chile, China, Colombia, Comoros, Congo, Cook Islands, Costa Rica, Côte d’Ivoire, Croatia, Cuba, Cyprus, Czechia, Democratic People's Republic of Korea, Democratic Republic of the Congo, Denmark, Djibouti, Dominica, Dominican Republic, Ecuador, Egypt, El Salvador, Equatorial Guinea, Eritrea, Estonia, Eswatini, Ethiopia, Fiji, Finland, France, Gabon, Gambia, Georgia, Germany, Ghana, Greece, Grenada, Guatemala, Guinea, Guinea-Bissau, Guyana, Haiti, Honduras, Hungary, Iceland, India, Indonesia, Iran (Islamic Republic of), Iraq, Ireland, Israel, Italy, Jamaica, Japan, Jordan, Kazakhstan, Kenya, Kiribati, Kuwait, Kyrgyzstan, Lao People's Democratic Republic, Latvia, Lebanon, Lesotho, Liberia, Libya, Lithuania, Luxembourg, Madagascar, Malawi, Malaysia, Maldives, Mali, Malta, Marshall Islands, Mauritania, Mauritius, Mexico, Micronesia (Federated States of), Monaco, Mongolia, Montenegro, Morocco, Mozambique, Myanmar, Namibia, Nauru, Nepal, Netherlands (Kingdom of the), New Zealand, Nicaragua, Niger, Nigeria, Niue, North Macedonia, Norway, Oman, Pakistan, Palau, Palestine, Panama, Papua New Guinea, Paraguay, Peru, Philippines, Poland, Portugal, Qatar, Republic of Korea, Republic of Moldova, Romania, Russian Federation, Rwanda, Saint Kitts and Nevis, Saint Lucia, Saint Vincent and the Grenadines, Samoa, San Marino, Sao Tome and Principe, Saudi Arabia, Senegal, Serbia, Seychelles, Sierra Leone, Singapore, Slovakia, Slovenia, Solomon Islands, Somalia, South Africa, South Sudan, Spain, Sri Lanka, Sudan, Suriname, Sweden, Switzerland, Syrian Arab Republic, Tajikistan, Thailand, Timor-Leste, Togo, Tonga, Trinidad and Tobago, Tunisia, Türkiye, Turkmenistan, Tuvalu, Uganda, Ukraine, United Arab Emirates, United Kingdom of Great Britain and Northern Ireland, United Republic of Tanzania, United States of America, Uruguay, Uzbekistan, Vanuatu, Venezuela (Bolivarian Republic of), Viet Nam, Yemen, Zambia and Zimbabwe.

Targets

The targets for the various indicators are presented in Table S2.4.

TABLE S2.42030 global maternal, infant and young children nutrition targets

SOURCE: Authors’ (WHO, UNICEF) own elaboration.

Calculating the average annual rate of reduction

The progress assessments for all indicators except adult obesity are based on average annual rates of reduction. These rates of reduction are calculated using a linear regression analysis. The dependent variables are the natural log transformations of all data points. The independent variables are all the years for the data points. The coefficient of this linear regression (β) can be translated into the average annual rate of reduction by using the formula:

Equation 5: AARR = 1 – e_β

Stunting, anaemia, low birthweight and overweight are model based and as a result, there is a consistent country-level time series. For these indicators, the baseline is 2012.

Exclusive breastfeeding and wasting are based on primary data, which are predominantly nationally representative surveys. The timing of these surveys is not consistent across countries.

As such, the baseline is defined as follows:

1. If the country has data between 2005 and 2012 → Select the latest data point in this year range.
2. If the country only has data from 2013 onwards → Select the earliest data point in this year range.

Calculating the current average annual rate of reduction

The current AARR is calculated using all the data points from the baseline to the latest year of available data. For the indicators based on primary data, there must be at least two data points with one of the data points after the baseline period of 2005–2012.

Calculating the required average annual rate of reduction

The required AARR is the AARR required to reach the target in 2030. This is calculated using two data points: the baseline data point and the target prevalence:

Equation 6:

where:

On-track prevalence level

The prevalence thresholds presented in Table S2.5 are used to determine if a country is on track to achieve the target regardless of the current AARR and the required AARR. For all indicators except for stunting, this assessment is based on only the point estimate. Assessments for stunting based on prevalence level are based on the lower confidence interval limit.

TABLE S2.5On-track prevalence levels for six global maternal, infant and young children indicators

SOURCE: Authors’ (WHO, UNICEF) own elaboration.

Calculating the posterior probability of a true increase

Progress assessments for adult obesity are based on posterior probability that the prevalence of obesity is truly flat or decreasing. Posterior probabilities are a measure of certainty. They indicate – based on available data and assumptions – our estimated probability of a certain outcome being true (e.g. prevalence was flat or decreasing from 2010 to 2022). The regression analysis described in Equation 3 is used to compute these posterior probabilities. We report the posterior probability that an estimated change in obesity prevalence represents a truly decreasing trend as the percentage of Bayesian iterations for which β is less than or equal to 0. Countries are assessed as being on track if the posterior probability of a flat or decreasing trend is greater than 0.5, and otherwise are considered to be off track.

The rules for assessing progress towards the seven global nutrition targets are summarized in Table S2.6.

TABLE S2.6Rules for assessing progress towards the seven global nutrition targets

NOTES: n.a. = not available; AARR = average annual rate of reduction. * Progress assessments for exclusive breastfeeding are based on non-exclusive breastfeeding (100-exclusive breastfeeding).
SOURCE: Authors’ (WHO, UNICEF) own elaboration.

Population weights

For the estimates of percentage of total population living in countries that are on track, off track and no assessment, the analysis is based on the same population weights used to generate regional and global aggregates as defined in Annex 1B of the main report. These are as follows:

1. Stunting, overweight, wasting: children under five years of age (sum of children aged 0 through 4)
2. Anaemia: females aged 15–49 (sum of females aged 15 through 49)
3. Low birthweight: live births
4. Exclusive breastfeeding: children aged 0–6 months (half of children aged 0)
5. Adult obesity: adults aged 18–100+ (sum of adults aged 18 through 100+)

For consistent comparisons, the constant year of 2023 is used for all indicators.