Food systems fundamentally shape lives, well-being and human and planetary health, and they are central to tackling some of the most pressing global challenges of our time1. The United Nations (UN) held its first-ever Food Systems Summit (UNFSS) in 2021, which demonstrated the interconnectedness of food systems with the Sustainable Development Goals (SDGs) and provided a space for countries to develop national pathways towards food system transformation. Food systems also featured prominently at the 26th and 27th UN Climate Change Conference2 and in the Kunming-Montreal Global Biodiversity Framework targets3. This context offers growing momentum to influence public policy, private sector and civil society actions to transform food systems from their current unsustainable and inequitable trajectories to a healthier, more equitable, sustainable and resilient future4,5,6. Rapidly progressing towards the 2030 expiration of the SDGs and amid mounting social, political, health and ecological challenges, transforming food systems to support healthy diets in sustainable, resilient, just and equitable ways is more urgent than ever1,7,8. Yet while the contributions of food systems to global goals are recognized and the clear need for monitoring has been articulated9, decision-makers across sectors lack a means to assess their food systems, guide action or evaluate progress. Furthermore, without monitoring, bright spots and success stories go unrecognized when they can offer important lessons for other places.

In 2021, the Food Systems Countdown to 2030 Initiative (FSCI) emerged from the UNFSS as an interdisciplinary collaboration of dozens of scientists with the ambition to fill this monitoring gap. The authors first published the conceptual foundation in which they described the goal of food system transformation as “a future where all people have access to healthy diets, produced in sustainable, resilient ways that restore nature and deliver just and equitable livelihoods”1. They developed a monitoring architecture comprising five thematic areas, each with three to five indicator domains1. Building on the architecture, this paper presents the indicator selection process and the resulting indicator framework and global food systems baseline. To select indicators that capture all elements of the architecture, we surveyed additional scientific experts and conducted consultations with hundreds of policy stakeholders in a multi-stage indicator selection process. The process was restricted to existing indicators—or feasible modifications thereof—and aimed to align with other indicator frameworks, such as the SDGs, where sensible.

The consultative process selected 50 indicators and identified several data gaps, of which many are expected to be fillable in the near term (before 2030). We applied the 50-indicator framework to provide a harmonized baseline dataset as an initial descriptive analysis of the world’s food systems, the starting point to track change and an essential first step in a global food systems research agenda. For the next seven years (2023–2030), the FSCI will publish annual updates, incorporate new indicators to fill the remaining gaps and carry out further analyses. Specifically, in the next two years, publications will concentrate on understanding country-level performance and the dynamic interactions across indicators, domains and themes.

The fundamental contributions of this paper are (1) an application of the recently developed global architecture to monitor food systems1, (2) the selection of a set of indicators legitimated through consultative process, (3) the identification of the most critical data and information gaps for global food systems monitoring and (4) a harmonized baseline dataset to track food systems and their changes. These contributions are relevant to the government officials responsible for developing food system transformation pathways coming out of the UNFSS, who have expressed clear demand for guidance on indicators10,11,12,13. African countries are working to adapt the Comprehensive African Agricultural Development Programme to incorporate a broader food systems perspective, also requiring additional indicators for the Biennial Review process14,15. The intent is not to create another set of indicators that countries have to track but rather to offer a menu that can be useful for the food system transformation goals that countries are establishing, providing a mechanism for accountability to stated commitments where existing suites of indicators (for example, SDGs) are insufficient for food systems (Supplementary Fig. 1.1 contains the theory of transformation)16. Basing the framework on feasibility, existing indicators and available data lends further practicality and usefulness to leaders acting now. At the global level, the framework enables policymakers, advisors, private sector actors and civil society actors to monitor food systems worldwide.


Indicator selection

We employed a multi-stage, multi-stakeholder process to select the suite of indicators for food systems monitoring. In the first stage, we developed a long list of all possible indicators. This list was screened for feasibility, coverage and transparency (defined in Extended Data Table 2). The result was a shorter list of candidate indicators to be evaluated against the criteria of relevance, high quality, interpretability and usefulness (the operational definitions are provided in Extended Data Table 2). In stage two, a survey was fielded to all authors and additional experts to quantitatively score the indicators against the criteria and identify any alternative indicators or data sources and indicator gaps. Qualitative consultations were held with over 500 policy stakeholders across the world focused on gathering input on usefulness and gaps. In the final stage, we examined the indicator scores, additional suggestions to address gaps, and gaps that could not be filled to identify the list of 50 indicators presented in this baseline. Figure 1 presents the flow of indicators through the selection process, and full reports of the survey results and policy consultations are provided in Supplementary Appendix 3.

Fig. 1: Multi-stage indicator selection process.
figure 1

The process of indicator selection and the number of indicators included and excluded at each stage. The excluded indicators are listed in Supplementary Appendix 2.

Table 1 presents the indicators and their global distributions, while Extended Data Table 1 contains the definitions, sources, rationale for inclusion, coverage and notable limitations. Many indicators have long time series available, while those without are expected to be collected or computed globally going forward and therefore are applicable for monitoring. Given our objective to work with existing data, there are limitations to these indicators, with several serving as imperfect proxies given data availability to be replaced as improved data and indicators are available (further details are given in Extended Data Table 1).

Table 1 Indicator list and global baselinea distributions

Diets, nutrition and health

Supporting human health is one of the three fundamental goals of food systems. The three indicator domains in this theme are food environments (the interface between individuals and the food system), food security and diet quality. One important aspect of food environments is the availability of different kinds of foods, reflected by the availability of fruits and vegetables and per capita sales of ultra-processed17 foods. Access to sufficient, safe, nutritious food and clean water is a core piece of food systems monitoring. Access to food is in part determined by the cost of a healthy diet—that is, the cost of purchasing the least expensive locally available foods to meet requirements for energy and food-based dietary guidelines. The affordability of that diet (cost relative to income) is one of three food security indicators alongside the prevalence of undernourishment and the percentage of the population experiencing moderate or severe food insecurity. Access to clean water is essential for avoiding food-borne and water-borne illnesses. No adequate available indicators exist for food safety, a priority data gap. Diet quality indicators capture what individuals actually eat, and they reflect diversity, adequacy and moderation. Indicators include minimum dietary diversity for women and children, consumption of the five food groups typically recommended for daily consumption in food-based dietary guidelines (fruits; vegetables; pulses, nuts or seeds; animal-source foods; and starchy staples), dietary factors that either protect against or increase risk for non-communicable diseases, and unhealthy dietary practices over the life cycle, aligned with international guidance18,19,20.

Environment, food production and natural resources

Food systems are a major contributor to environmental degradation, but they can also protect and restore environmental outcomes if managed appropriately. The six domains of environmental indicators address the multiple environmental impacts of food systems: greenhouse gas emissions, land, biosphere integrity, water, pollution (conceptually including nutrient runoff, chemical exposure and solid waste) and agricultural production, which interacts with all other domains.

Indicators of greenhouse gas emissions include total emissions (from production through consumption and waste disposal) and emissions intensities (emissions per unit of primary product) of major foods. Land use change is measured by cropland expansion and water use, expressed by how much agricultural water withdrawals place pressure on renewable freshwater resources. Overuse of pesticides and sustainable nitrogen management capture pollution; additional indicators of solid waste and chemical pollution attributable to food systems are wanting. Functional integrity—the capacity for biodiversity to support sustainable food production and other ecosystem services—and the integrity of fishery stocks capture biosphere integrity. Yields interact with all other domains; increases are directly tied to the observed declining trends in emissions intensities.

Livelihoods, poverty and equity

Poverty is most prevalent in rural areas where people earn substantial income shares from agriculture (including marginalized groups such as Indigenous Peoples and female-headed households)21,22,23. Food systems provide employment for 1.23 billion people and (including household members) support over 3.83 billion livelihoods, in all stages of the value chain across rural and urban areas24. Four indicator domains capture their well-being: income and poverty, employment, social protection and rights. Compared with other themes, the available data are more limited due in large part to lack of disaggregation to distinguish food system livelihoods from others.

Lacking a rural poverty indicator with sufficient coverage, the share of gross domestic product (GDP) from agriculture provides a proxy for a country’s overall level of development25. Declining GDP from agriculture and fewer people working in agriculture are hallmarks of the structural transformation process that is integral to poverty reduction and rural transformation25. Unemployment and underemployment capture employment, though not ‘decent’ work26. Though lacking sectoral disaggregation, the rural rates proxy the status of agricultural and farm-related labour markets27. Social protection systems increase access to food quantity and quality, reduce producers’ risk and incentivize productive investment28,29. Social protection programmes may be particularly impactful in breaking the cycle of poverty for small-scale food producers and informal workers who face chronic food insecurity and vulnerability to shocks29. Finally, among the many rights and issues of justice related to livelihoods, the indicators currently available capture women’s access to land and the specific human rights violation of child labour, of which an estimated 70% occurs in agriculture30.


Governance is foundational for inclusive food system transformation, encompassing not only the political commitment to adopt supportive policies but also promoting participatory processes and accountability to ensure that policies have legitimacy and reach the intended target group. Furthermore, governance involves strengthening capacities for implementation across sectors to ensure that aspirational goals are technically feasible. Three indicator domains collectively capture these dimensions of governance: shared vision and strategic planning, effective implementation and accountability. There are few indicators of governance specific to food systems, but broad indices of the governance landscape may have substantial impacts on food system choices and outcomes. Further research is especially needed in this area to develop more direct indicators of food system governance.

Indicators of shared vision and strategic planning include one broad indicator beyond food systems and three others reflecting intentionality by governments to pursue food systems objectives. The Civil Society Participation Index captures whether civil society organizations (for example, non-governmental organizations, unions and social movements) have opportunities to convey their views to policymakers. Food-system-specific indicators are the presence of a legal recognition of the right to food, the existence of a food system transformation pathway and the share of the urban population living in cities that have signed on to the Milan Urban Food Policy Pact (MUFPP). The MUFPP is an innovative policy mechanism that has rapidly become the leading international tool for urban food policy governance (37 recommended actions and specific indicators) as well as a platform for cooperation, organizing and political influence31.

Effective implementation is also measured by a combination of indicators that are contextual (broader than the food system but establish the governance regime within which food system actors can operate) and specific to food systems. The government effectiveness index reflects the quality of public services, civil service, policy formulation, implementation and credibility. Public tracking of investments for food systems requires transparency over budgets and guarantees for public information access, reflected in the Open Budget Index score and guarantees for public access to information, as well as the overall Accountability Index, which encompasses the existence of mechanisms to keep officials responsive to the public (for example, checks and balances, elections and press freedoms). Specific to food systems, available data can monitor two policy tools for achieving healthy food systems: health-related food taxes and food safety capacity (the number of specific mechanisms in place to detect and respond to food-borne disease and contamination).


We define food system resilience as “the ability of different individual and institutional food system actors to maintain, protect, or quickly recover the key functions of that system despite the impacts of disturbances”1. The COVID-19 pandemic and the conflict in Ukraine both demonstrated the imperative to better understand and strengthen the resilience of local and global food systems to numerous shocks and stressors—not just climate change. Assessing resilience requires a combination of indicators related to two domains: (1) the contextual elements of resilience (the level of exposure of the system to adverse events and the capacities of that system to anticipate, absorb or adapt to those events) and (2) the short- and longer-term outcomes of resilience—generally measured through individual and system well-being, ideally considered at multiple scales32.

A range of indicators is necessary to capture these different components of resilience and to better understand how to establish more efficient, inclusive and sustainable food systems in the face of increasingly complex and intertwined shocks. The indicators of resilience therefore cover five domains: exposure to shocks, resilience capacities, agrodiversity and food diversity, short-term resilience responses and long-term outcomes.

Exposure to shocks depends on the intensity, nature and frequency of shocks and stressors and can be proxied by the cumulative costs of those events relative to GDP. Resilience capacities are the different elements that can be used to buffer and respond to adverse events. Those capacities take many forms. In food systems, the diversity and redundancy of food sources, national infrastructure (proxied by mobile phone coverage) and social capital are some of the key elements that constitute resilience capacities. Also critical to food system resilience is the level of biodiversity on which food production relies, captured by the number of plant and animal genetic resources conserved for use. Understanding how actors react and respond in the short term to the impact of shocks is also a foundational element of resilience analysis. This element can be measured using the coping strategies index, while longer-term outcomes of food system resilience can be captured by the ability of the system to maintain low price volatility and low food supply variability.

This resulting indicator framework partially overlaps with the SDGs, underscoring both the relevance of the overall development agenda for food system transformation and, conversely, the inadequacy (incompleteness) of the SDG framework for food systems monitoring16. Of the 240 SDG indicators, 81 are related to food systems and food system transformation. Only 11 are specific to food systems, and of those, only 5 meet the criteria for inclusion in the FSCI. The SDG indicators included are 2.1.1 (undernourishment), 2.1.2 (food insecurity), 6.1.1 (safe drinking water), 16.10.2 (access to information) and 2.5.1 (conserved genetic resources). Three SDG indicators are expected to be added as soon as data become available, including 2.4.1 (sustainable agriculture) and 12.3.1 (food loss and waste indices). Similarly, 5.a.1 (women’s agricultural land ownership) will replace the current data source (which will no longer be updated), and 5.b.1 (mobile phone ownership) will replace the current indicator of phones per 100,000 people, as soon as there is sufficient country coverage. Supplementary Table 4.1 documents which SDG indicators are relevant to food systems, the subset of those that are included in the FSCI framework and an explanation for why the others are not included.

Data gaps

Notable data gaps emerged through the indicator selection process. Several gaps cut across multiple themes such as the true cost of food, a cost that includes the externalities currently unaccounted for in the market price such as diet-related disease, pollution and natural resource degradation33. Similarly lacking are data on food loss and waste at the country level, and we await country-level data of sufficient quality for SDG 12.3.1 (food waste and food loss indices).

In the realm of diets, nutrition and health, food safety is an area lacking indicators (and data), though food safety capacity is captured in governance. Under environment, natural resources and production, a gap regarding sustainable agriculture will be filled with SDG 2.4.1 (agricultural area under sustainable management practices) when data for the recently developed indicator are available34. Other gaps include food production and supply indicators inclusive of aquatic and wild foods. Furthermore, environmental indicators predominantly relate to production and largely exclude loss and waste as well as pollution related to processes further down the value chain (for example, solid waste and material pollution from packaging35). Many gaps exist with respect to livelihoods, including the economic value of food systems, the magnitude and composition of populations working in food systems and their vulnerabilities, and productivity in the sector (for example, value-added as a share of GDP and per worker). In addition, indicators of livelihoods that can capture the welfare of food system workers beyond agriculture—especially measures of decent work, gender equity and violations of human rights in food systems—are needed. With respect to food system governance, data gaps include policy coherence (alignment across policy areas) for food system transformation and budgetary allocations to food systems. These gaps require substantial country-level data inputs to fill, but new machine learning methods may provide opportunities to develop estimates that can be added to the indicator suite in the near term. Additional indicators of governance and resilience specific to food systems are also lacking.

Gaps also pertain to the country and time series coverage of indicators. Figure 2 presents a data coverage heat map from 2000 forward showing that the indicators with the greatest country coverage and the longest time series are those associated with agricultural development such as yields and the share of agriculture in GDP. For other indicators—adult diet quality, biodiversity, and agrodiversity and food diversity—the country and year coverage remain sparse. Country coverage of diet quality indicators is expected to increase rapidly, but there are no adult diet quality data for any countries in Oceania, and there are data for only one country in the Caribbean—a priority gap, given the high burden of diet-related disease in these regions18,36,37. Environmental indicators have the greatest coverage, partly because so many derive from FAOSTAT indicators with a long history of collection38. Governance indicators also have good country coverage, but one third of the indicators in this theme are newly developed (right to food, presence of a food system pathway and urban population signed on to the MUFPP). Livelihood and resilience indicators have poorer geographic coverage across most regions, especially Oceania and northern Africa and western Asia.

Fig. 2: Data coverage, number of years per country–indicator, 2000–2021.
figure 2

Heat map illustrating the density of data points per country–indicator pairing, with the darkest cells illustrating more years of data between 2000 and 2021. Indicators with no data at all for that country are shown in white. The figure illustrates greater availability of data for food security and agricultural indicators and lesser availability for indicators of diet quality, livelihoods and resilience. Heat maps showing the indicator–country time series by region are available in Supplementary Figs. 1.31.11. The maximum country coverage is all UN member states, but coverage differs per indicator depending on data availability. Differences in indicator coverage largely drive the observed differences across countries. Specifically, the indicators with the most heterogeneous coverage are the six indicators of diet quality sourced from the Global Diet Quality Project (currently available for only 41 mostly low- and lower-middle-income countries); the livelihood indicators of employment, social protection, child labour and landholdings; and the resilience indicators of genetic resources and coping strategies (available for countries with a high prevalence of food insecurity). Looking across countries within each indicator, countries with the indicator typically have time series of similar durations. Yield and emissions intensity for additional products are provided in Supplementary Appendix 1 and the baseline dataset. FIES, Food Insecurity Experience Scale.

By region (Supplementary Figs. 1.31.11), Oceania has the greatest scarcity in data overall, with very few diet quality indicators collected in that region and only for children in 4 of the 14 countries. Dietary data are collected in fewer countries of North America and Europe, northern Africa and western Asia, and Latin America and the Caribbean than in other regions. Countries with the fewest indicator and year observations are small island nations (for example, Caribbean and Pacific islands), very small high-income countries (HICs) (for example, Brunei, Monaco and Singapore), several countries in the Middle East (such as Saudi Arabia and Qatar) and countries (recently) experiencing conflict (for example, Eritrea and Syria).

Global baseline

Table 1 presents the selected indicators and their global distributions in the most recent year for which data are available (the definitions, data sources, rationale for inclusion, key limitations and desirable direction of change are provided in Extended Data Table 1). In Table 1 (and Fig. 3), the best ranking and worst ranking reflect the ranking of all countries per indicator relative to the desirable direction of change, where the best (first ranking) is the highest value for indicators where higher is more desirable and the lowest for indicators where lower is more desirable. A lower rank indicates a better ranked position. The characterization is meant to be descriptive of the relative baseline starting points and is not intended as a performance assessment, which is a subsequent research agenda of the FSCI in the next two years. The baseline dataset includes the latest available data point per country–indicator, which differs given data availability. Most data (92.5%) are from 2017–2022, 6.5% are from 2010–2016 and only 1% are from 2000–2009. The specific year per country–indicator pair is reported in Supplementary Data 1, and the complete country-level dataset is in Supplementary Data 2. Several general patterns emerge from this global view, supported by descriptive analyses by region and income group for all indicators in Supplementary Figs.

Fig. 3: Average country ranking per theme, by country income level.
figure 3

The mean ranking of indicators within each theme shows that no country is performing in alignment with desirable outcomes for all themes. The bottom ranking indicates scores farthest from the minimum or maximum value observed across all countries, depending on whether lower or higher values are aligned to the desirable outcome. Countries are ranked per indicator relative to all other countries, and the average rank for all indicators within a theme is shown per country. Countries are grouped by income level and presented in order of increasing income from left to right and top to bottom. The horizontal black lines indicate the global median rank pooling all indicators. The results are from our calculations based on the data sources listed in Extended Data Table 1 and from the latest data point per country–indicator pair, of which the majority come from 2017–2021. Supplementary Data 1 contains the specific year for each country–indicator data point. Binary and categorical indicators are not ranked and are therefore excluded from the governance theme average. Country ranking per indicator is averaged at the theme level. Not all countries have data for every indicator. Missing data do not bias the total ranking visualized, but they do result in implicit weighting of the thematic mean rank by the present indicators. Country abbreviations shown as ISO alpha-3 country codes.

All food environment indicators suggest inequalities across countries: the availability of fruits and vegetables is generally a challenge in low- and middle-income countries, while HICs generally have widespread availability of ultra-processed foods. The cost of healthy diets is similar across most countries, but given wide differences in purchasing power, that cost is largely unaffordable across low- and middle-income countries.

Despite some improving trajectories, total food system emissions are increasing and remain high in HICs. Northern Africa and western and southern Asia remain at the greatest risk of exhausting available water resources. Only 88% of agricultural lands have the minimum of 10% functional integrity needed to support food production, meaning over one tenth of the world’s agricultural lands lack foundational ecosystem services such as crop pollination, pest regulation and soil protection, and other research suggests that the 10% threshold may be insufficient39.

The available data provide only a partial view of food-system-based livelihoods, but even the incomplete picture suggests deep inequalities. Important differences in unemployment and underemployment between rural and urban areas show that unemployment is prevalent in urban areas while underemployment is more prevalent in rural areas. Other evidence shows a larger gender gap in labour force participation in rural areas40. Even where there is adequate coverage of social protection programmes, the level of benefits provided may be insufficient to produce meaningful impacts, and informal and seasonal workers are often excluded41,42,43,44. Finally, access to land shows a stark gender disparity with no country approaching gender equality in landholdings.

The data show that indicators of overall governance track country income, while those more closely related to food systems show more heterogeneity across regions and income groups. For example, only 29 countries explicitly recognize the right to food, while the United States, Canada, the United Kingdom and Australia notably have no degree of legal recognition. In addition, health-related food taxes exist in 38 countries spread across all continents.

Looking across resilience indicators for a sub-group of countries (Supplementary Fig. 1.5.17), the data show that some countries (for example, the Philippines, Nicaragua and Indonesia) demonstrate relatively higher food price volatility or food supply variability than others (for example, the Netherlands, Thailand and India). These are countries facing higher exposure and/or lower resilience capacities (such as Nicaragua and Ecuador), showing that they also fare worse in their food system outcomes than those less exposed to shocks and/or characterized by higher social capital and dietary sourcing flexibility (such as Thailand and the Netherlands). However, this trend displays important variability, reflecting the specificity in how shocks propagate through a country’s food system, and calls for more in-depth analyses such as the future work planned to focus on interactions and dynamics of change across food systems.

Many aspects of food systems are associated with country income level25, raising questions of which indicators evade income trends and whether there are inflection points by income that might help countries set priorities. Figure 3 presents the country-level mean ranking per theme, grouped by country income level (for grouping by region, see Supplementary Figs. 1.121.20). The results illustrate that within every income group, there are some countries performing better than others. Even among the lowest-income countries, Uganda and Mozambique rank near the global median across all indicators, while on the other end of the spectrum, despite their high-income status, several countries rank worse on average than countries with many fewer resources. This analysis begins to suggest which countries might have useful lessons for others, especially those non-HICs outranking their income-group peers such as Uganda, Mozambique, Sri Lanka, the Philippines, Nigeria and Kazakhstan.

Looking at each indicator by country groupings, Fig. 4 shows the distribution of country income groups relative to the global average. The figure is aligned to the desirable direction of change (defined in Extended Data Table 1) such that to the right of the global mean is better. Additional analyses present the values displayed and test for statistically significant differences by country income group and by region (Supplementary Tables 1.1 and 1.4) and provide weighted means and medians by region and income group (Supplementary Tables 1.2, 1.3, 1.5 and 1.6). These analyses show that only the presence of a national food system pathway is not statistically significantly different by region, while numerous variables do not differ by country income group, including the cost of a healthy diet, the availability of fruits and vegetables, the minimum dietary diversity for women, food system emissions, cereal yields, cropland change, agricultural water withdrawals, functional integrity, rural underemployment, women’s share of landholdings, the presence of a food system transformation pathway, the Open Budget Index and mobile subscriptions.

Fig. 4: Average country ranking per theme, by country income level.
figure 4

Normalized difference between each income group mean value per indicator and the global mean for that indicator (represented by the black vertical lines). Differences are aligned to the desirable direction of change such that points to the left of the global mean indicate that the indicator mean level is less desirable than the global mean and points to the right indicate values more desirable than the global mean. The results are from our calculations based on the data sources listed in Extended Data Table 1 and from the latest data point per country–indicator pair, of which the majority come from 2017–2021. Supplementary Data 1 contains the specific year for each country–indicator data point. The normalized distance to the global mean (weighted means following the weights defined in Table 1) is calculated relative to the global mean and scaled to the minimum and maximum of the income group mean, per indicator (the global mean is centred at 0). The sign of the normalized distance has been reversed for all indicators where a lower value is more desirable, such that −1 can be interpreted as ‘worse than’ and 1 can be interpreted as ‘better than’ the global mean. The number of people who cannot afford a healthy diet and the degree of legal recognition of the right to food are not shown. The product mixes in aggregate categories of emissions intensities (cereals) and yields (cereals, citrus, fruit, pulses, roots and tubers, and vegetables) differ across countries. Yield and emissions intensity for additional products are included in Supplementary Appendix 1 and the baseline dataset.

Beyond country income level, understanding each indicator’s relationship to GDP per capita is useful for hypothesis generation. Supplementary Figs. 1.211.26 show the relationship between each (continuous) indicator and GDP per capita. Several indicators exhibit less obvious relationships to GDP, including the cost of a healthy diet, pesticide use and sustainable nitrogen management, yields for vegetables and roots and tubers (potentially reflecting different crop mixes), female landholdings, food price volatility, food supply variability and mobile phone subscriptions. These findings underscore the potential for policymakers and other actors to influence more desirable outcomes on at least some indicators of food systems even in low-income countries, and to identify where income seems to be a necessary driver (though alone probably insufficient) of more desirable outcomes.


The indicator framework presented in this paper allows progress across global food systems to be meaningfully tracked, complementing the SDGs and other indicator frameworks with a curated set of existing indicators to monitor food systems, selected through a consultative process. It provides the foundation for future research to better understand how and where change comes about, and importantly how to identify where improvements in any one domain do not necessarily translate into improvements in others45,46. Looking across this baseline, the indicators included offer a trove of information that provides transparency and specificity to the important constructs but does not prescribe obvious or uniform actions. Three clear messages emerge. First, no country, region or income group exhibits desirable status across all indicators. Second, not all food system indicators are aligned to country income level; there are a diversity of food system trajectories. And third, there are some critical data gaps to monitor the world’s food systems that must be filled in the near term to guide action in service of food system transformation, meeting the SDGs and ensuring that food systems positively contribute to the many global goals linked to food systems.

The FSCI effort is intended to complement other global goal setting and monitoring efforts such as the SDGs, through the lens of food systems, which have been only partially captured in existing goals, indicators and monitoring efforts. We aim for synergies with these internationally recognized goals, but the very small overlap between the SDGs and the FSCI framework reflects the fact that food systems were not yet considered a mainstream framing approach when the SDGs were developed. As food systems become more widely understood from a systems perspective, the large set of FSCI indicators that are not in the SDGs provides some guidance as to indicators that could be considered for the next set of global goals.

The process of indicator selection identified key data gaps—the specific information that needs to be collected at scale to achieve the ambitious goal of tracking and informing food system transformation. The gaps span all themes—for example, livelihood indicators beyond agriculture, food loss and waste, and governance of food systems. Many ongoing initiatives are working to fill some gaps (Supplementary Table 4.2), with notable achievements already in bringing data together (for example, the Food Systems Dashboard47). The baseline dataset provides a starting point for tracking, and the framework of indicators can be used by policymakers and other food system actors to diagnose their food systems and formulate appropriate responses, including transformation plans, and monitor advances in their countries. The baseline description demonstrates that no country shows positive outcomes across all dimensions. In addition, given that some food system outcomes are independent of national income levels, dedicated monitoring and transformation agendas specific to food systems are needed. Ongoing expansion of the FAOSTAT database and the Global Diet Quality Project will also help fill these gaps18,48. Other advances are dramatically reducing costs and increasing the quality and granularity of new data collection (for example, the 50×2030 Initiative)49,50,51,52.

This indicator framework was developed with usefulness to countries and other food system decision-makers as a driving purpose, but country-level testing and adaptation is warranted. Following the UNFSS process, at the time of latest analysis by the Food and Agriculture Organization (FAO), 123 governments had developed national food system transformation pathways12. The five domains of the FSCI architecture map closely to these pathways and will allow them to be well monitored with the indicators selected and presented here (Supplementary Table D.3). There is utility in tracking national progress relative to goals as well as relative progress within a region, by income peer group or in the world overall. In addition to meeting the information needs at a country level, the indicator framework is useful in addressing the supranational and transboundary issues within food systems that require alignment, coordination and goals at higher jurisdictional levels. Decision-makers can use the framework as a starting point to consider what changes in indicators are achievable at different scales and can forge coalitions to drive change. Furthermore, different actors may find certain indicators more useful for guiding action than others. For example, donors may be more concerned with cross-country comparisons when deciding how to allocate resources. National policymakers may be more interested in understanding how their country is doing over time on indicators under more direct national influence or control.

This baseline sets the stage, but future work is needed to close data gaps, assess status relative to benchmarks aligned to transformation, understand how food systems evolve over time (including interactions across different indicators that affect the sustainability of food systems overall) and better understand and take action to support the needs of national and global data users. The FSCI will undertake this research and action agenda in the coming years alongside regularly updated assessments tracking progress from this baseline forward, including the addition of new indicators or the refinement of the current set of indicators as food systems science progresses. By doing so, the FSCI aims to facilitate and accelerate food system transformation to deliver a healthier, more equitable, sustainable and resilient future for all.


This paper has used the term ‘food systems’ throughout, in line with the UNFSS language. However, the indicator framework presented takes an expanded concept of agri-food systems given that many indicators cannot distinguish between food and non-food components of production and value addition, although such non-food components greatly influence the environment, social outcomes and the food people ultimately eat. Hence, food systems as used here encompass activities and processes around non-food agricultural products (for example, forestry, fibres and biofuels) that are interconnected with food for human consumption1.

A rigorous set of prerequisite criteria were established that all indicators had to meet to be considered at all for this work, which included feasibility (having recent data and being planned to be updated within the next eight years), coverage (at least 70 countries across regions and income levels) and transparency (no modelled indicators with undisclosed or untraceable methodologies). A comprehensive multi-stage, multi-stakeholder process was then conducted to select the list of indicators analysed in this paper (described in further detail below). Using a quantitative survey, dozens of experts were asked to rate each candidate indicator on its relevance, the quality of the data and methods, and its interpretability for policy purposes. Indicators assessed to be relevant, high quality and interpretable were considered to be useful, and a usefulness criterion was applied to the suite of indicators selected to monitor each domain to ensure sufficient but not redundant information. Finally, crucial input on regional priorities and policy utility provided by policy stakeholders was incorporated. Several indicators come from common sources such as FAOSTAT, the Gallup World Poll and the World Bank, but data from many other academic and non-governmental organization sources are also included. This replicable protocol including the survey and consultation processes culminated in our final selection of the indicators presented in this paper. All data and replication code are publicly available.


The data used in this paper were sourced from many global, publicly available data sources. Extended Data Table 1 provides the data source, description, rationale for inclusion and coverage metadata for each indicator as well as any notable limitations and mitigation actions to address them. Supplementary Data 1 provides an Excel spreadsheet containing the complete metadata, a codebook, country and year coverage, and the year of the latest data point per country–indicator that comprises the baseline. Supplementary Data 2 contains the complete baseline dataset of the latest data point per country per indicator used in the baseline analysis presented herein.

Indicator selection

We employed a multi-stage, multi-stakeholder process to select the list of indicators analysed in this paper. A preliminary set of criteria was previously published in Fanzo et al.1. In the first stage of indicator selection, we refined these criteria by deeming three attributes to be essential: feasibility, coverage and transparency. Next, we refined the four criteria established previously: relevant, high quality, interpretable and useful. Extended Data Table 2 details the requirements, criteria definitions and sub-criteria.

Working group members compiled a list of candidate indicators for each domain that met the prerequisite requirements for potential inclusion. Supplementary Appendix 2 contains the indicator catalogue of all candidates, indicator options excluded for failure to meet the prerequisites and all relevant information that was provided to assess the indicators. This list of candidate indicators was assessed against the first three criteria (relevance, quality and interpretability) using an online survey by all the collaborators and an additional group of over two dozen external experts who were volunteer respondents based on a list of experts generated by all the authors with additional research to reach relevant people unknown to the author group. Everyone assessed indicators in the domain(s) aligned with their expertise. The respondents were asked to choose their level of agreement (from 1 to 5) with the statement that the candidate indicator met each sub-criterion, the elements in the bulleted lists in Extended Data Table 2. All respondents were also asked to state their agreement that the indicator is important for tracking food system transformation and to share their interpretation of both importance and transformation in that context, providing complementary qualitative data. Finally, the external experts were also asked to suggest additional data sources for candidate indicators and to describe any observed gaps in the domains and indicators and how they recommend filling those gaps. For those who assessed governance indicators, an additional question asked what new indicators the respondent deemed necessary and asked for recommendations for their construction. Supplementary Appendix 3 contains the full report of the survey procedures and outcomes, including all the scoring results. Figure 1 summarizes the flow of indicators through the process.

In parallel, the FAO convened five regional policy stakeholder consultations in Latin America and the Caribbean, sub-Saharan Africa, North Africa and the Middle East, Asia and the Pacific, and Europe. Over 500 people participated, averaging 75–100 per region. The consultations included a short overview presentation and breakout discussions of each thematic area. The participants were asked to assess the local pertinence of the architecture and indicator framework and to solicit regional priorities, interests and needs. Supplementary Appendix 3 contains the reports for each regional consultation. The consultation asked experts and stakeholders to suggest alternative indicators and data sources and to identify gaps, which resulted in the addition of several indicators to the initial list of candidates.

To identify the final list of indicators, scores from the assessment of indicators against the six sub-criteria of relevance, quality and interpretability criteria were summed to the indicator level with equal weighting, providing a single score per indicator. Usefulness was assessed qualitatively at the level of indicator domains, with emphasis on meeting the needs illuminated by the policy stakeholder workshops. Twelve indicators were added and ultimately included in the final set after the survey and consultations because they address gaps that were widely identified. These indicators are safe drinking water, agri-food system emissions, yields, share of agriculture in GDP, underemployment, degree of legal recognition of the right to food, percentage of the urban population living in a signatory municipality to the MUFPP, food safety capacity, health-related food taxes, guarantees for public access to information, proportion of agricultural land with minimum species richness and the number of animal and plant genetic resources in conservation facilities. Some gaps identified in the consultations could not be filled and are instead described in the data gaps and research agenda discussion; in particular, the lack of food loss and waste data was a prominent theme of the consultations.

Analysis methods

Analyses were carried out in Stata v.17 and R v.4.2.2. The data were compiled into a dataset where all years of available data per country and indicator were included. In two instances (EM-DAT and Varieties of Democracy indices), data prior to 1960 were excluded because no other datasets provided data before that year. Initially, all territories classified in the UN Global Administrative Units List dataset53 and present in any datasets were included (94 areas in total). After compiling the complete dataset with all indicators, we investigated whether there was sufficient coverage across all indicators for any territories or areas that are not UN member states to remain in the dataset. A criterion was applied that the area must have at least 80% of all indicators. In practice, all territories were dropped at a much lower threshold, none having more than the median number of variables present for member states (40, where certain indicators are represented in the dataset by more than one variable). In sum, the dataset contains all the available data from 1960 to 2021 for all UN member states, and one indicator (the presence of a food system transformation pathway) defined only in 2022.

The focus of this manuscript is a baseline dataset comprising the latest data point per country per year. Overall, 92.5% of all data points are from 2017–2022, 6.5% are from 2010–2016 and only 1% are from 2000–2009. A small number of observations (N = 24 across all indicators) were dropped from the dataset because the latest data point for that country–indicator pair came from prior to 2000. The only indicator where this dropped more than a few observations is female share of landholdings, which has 13 countries whose data point in that cross-sectional dataset is from the 1990s or before. A new data source will become available through the SDG process (SDG 5.a.1) for this indicator in future years.

The Supplementary Information includes analysis of the data from 2000 forward wherever time series are available. Countries are grouped into regions based on modified groupings of the M49 classification system of the UN Statistical Commission, using a combination of continental and sub-regional groupings. Supplementary Fig. 1.2 depicts the alignment of countries to the modified M49 regional grouping used in this paper. Countries are identified by income group using the World Bank country income classification54.

The rankings of indicators (Fig. 3) are calculated by ordering every continuous indicator numerically and assigning each country a rank order for every indicator. The rank is reversed for all indicators where a higher value is more desirable (per Extended Data Table 1), such that a ranking of 1 is assigned to the country with the most extreme (highest or lowest) value, whichever direction is desirable for that indicator. We calculated the average rank for all indicators per theme, with the limitation that doing so implicitly weights the thematic average rank for any countries without data for any indicators within the theme. This is an unavoidable limitation and allowed for country-level visualization of data with great variation in their range and units of analysis.

The distributions of the indicators by region and income group relative to the global weighted mean (Fig. 4 and Supplementary Tables 1.1 and 1.4) are presented as the normalized difference from the global weighted mean. The global weighted mean is subtracted from the region (income group) weighted mean and normalized using min–max scaling, which divides the demeaned observation by the total range across all regions (income groups) (that is, it divides by the maximum observed minus the minimum observed). Deviations of region and income group weighted means from the global weighted mean (Supplementary Tables 1.1 and 1.4) are calculated using weighted least squares regression with heteroskedasticity robust standard errors regressing region (income group) on the demeaned observation. Demeaned observations are calculated by subtracting the global weighted mean from each observation. The sign of the demeaned observation is reversed for all indicators where the desirable direction of change is lower. Regression coefficients are the regional (income group) deviation from the global average with the sign indicating whether the region is performing worse (negative sign) or better (positive sign) than the global average. The signed deviation is then translated into a percentage deviation by dividing by the global average to harmonize the presentation of indicators given the different units and scales of their level measurements.

Finally, we emphasize that this exercise was based on a framework of food systems, and therefore we would expect that certain features of a country’s food system would be related to other features of a country’s food system. To explore this, we calculated a Spearman rank correlation matrix (Supplementary Fig. 1.26). However, we caution the interpretation of correlation as redundancy; we do not intend to create a single index, in which case high levels of correlation among the variables entering the model would be problematic. Instead, we put forward this matrix for the purpose of hypothesis generation regarding the key interactions among indicators that merit further investigation, which will be the focus of our research agenda over the next two years.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.