The world's dominant systems for assessing food insecurity are missing roughly one in five people facing acute hunger. Recent research reveals what had long been assumed to be true—that global measurements overcount the scale of food crises—is entirely backwards.
The systems designed to identify and quantify hunger are instead operating in a systematically conservative manner, consistently underestimating the actual number of people in urgent need of assistance.
This discovery carries significant implications for international humanitarian response, as these measurements determine the allocation of billions in emergency aid annually.
In 2024, more than $6 billion in humanitarian assistance was allocated based on analyses from the Integrated Food Security Phase Classification system, the primary mechanism through which international organizations target food aid to vulnerable populations. If the underlying assessments are missing one-fifth of those in crisis, the actual scale of need far exceeds what relief agencies are being asked to address.
The Nature of Global Food Insecurity Assessment
The Integrated Food Security Phase Classification, or IPC, serves as the central pillar of international food security monitoring. Established in 2004 as a consortium of 21 partner organizations, the system analyzes food security conditions in approximately 30 countries across the world, primarily in regions prone to acute food crises.
The system categorizes food insecurity into five phases, ranging from phase 1 (none/minimal) to phase 5 (catastrophe/famine), with the critical threshold residing at phase 3, which designates an area as "in crisis" when more than 20 percent of the population lacks sufficient food.
This threshold is not arbitrary. The 20 percent marker determines whether a region triggers humanitarian intervention protocols and emergency funding releases.
Analysts working within the IPC framework gather multifaceted evidence—from food prices and weather patterns to dietary diversity and consumption data—and convene to assess information according to established protocols. They then assign classifications for each subnational zone based on their collective analysis.
The challenge lies not in the theoretical framework but in its practical application. Working groups conducting these assessments operate in regions where reliable data is often scarce and conditions are rapidly deteriorating.
They must synthesize multiple information streams that frequently contradict one another, creating situations where analysis teams face genuine uncertainty about how to classify areas falling near critical decision thresholds.
Evidence of Systematic Underestimation
A comprehensive evaluation published in Nature Food in December 2025 provides the first empirical evidence of this systemic bias. Researchers analyzed nearly 10,000 food security assessments conducted between 2017 and 2023, covering 917 million individuals across 33 countries.
When accounting for multiple assessments of the same populations over time, the dataset encompassed 2.8 billion person-observations.
The research team examined the distribution of food insecurity percentages near the critical 20 percent threshold separating phase 2 from phase 3. What they discovered was striking: a pronounced clustering of assessments just below the phase 3 threshold.
This pattern appeared consistently across multiple countries experiencing varying overall levels of food insecurity, suggesting it reflected a systematic tendency rather than random variation or genuine differences in underlying conditions.
The researchers then conducted their own independent analysis using the same underlying data available to IPC working groups. Their reassessment identified 293.1 million people in phase 3 or higher (requiring emergency assistance), compared to the IPC's official count of 226.9 million people.
The difference amounts to 66.2 million individuals—approximately one in five of those experiencing acute food insecurity—going unaccounted for in global estimates.
The Psychology of Conservative Bias
The tendency toward underestimation appears rooted in legitimate professional caution. Research revealed that stakeholders and users of the IPC system—humanitarian agencies, governments, and international organizations—tend to assume the system overstates the severity of crises.
This widespread perception that the IPC exaggerates need creates institutional pressure on analysis teams to be more conservative in their assessments.
The bias intensifies precisely where it matters most. When food security data available to working groups provide conflicting signals about the situation on the ground, committees are significantly more likely to classify areas as just below the critical threshold.
In other words, when uncertainty is highest, the tendency to err on the side of caution—and thus undercount hunger—becomes most pronounced.
Evidence suggests this reflects a genuine cognitive phenomenon among analysts. The more contradictory or noisy the underlying indicators, the greater the underestimation appears to be.
Rather than applying formal protocols to navigate conflicting evidence, working groups tend to adopt what researchers describe as a "convergence of evidence" approach, which in practice means deferring to lower classifications when data conflict.
Multiple Measurement Frameworks, Persistent Limitations
The IPC operates alongside other global measurement systems, each approaching food insecurity from different angles and each carrying its own limitations.
The Food Insecurity Experience Scale, developed by the United Nations Food and Agriculture Organization and based on the Voices of the Hungry initiative, measures food insecurity through household surveys that ask about experiences of food deprivation. The Prevalence of Undernourishment, the FAO's traditional indicator, estimates the proportion of populations unable to meet dietary energy requirements.
These approaches differ fundamentally in what they capture. The FIES focuses on the experience of food scarcity at the household level, gathered through questions about food access concerns and reduced consumption.
The Prevalence of Undernourishment relies on aggregate national data regarding food availability and distribution, modeling the proportion of populations with inadequate caloric intake.
The differences matter. Research comparing experience-based measures with nutrition-based measures in Bangladesh, Ethiopia, and India found substantial disagreement about which households are most food-insecure.
When experience-based tools identify individuals as food-insecure, they often differ from those identified through caloric consumption measures. More critically, the two approaches frequently disagree on which households face the most severe food insecurity—precisely the subset whose consumption patterns most urgently require tracking and intervention.
A broader systematic review examining tools available in developed countries found that most focus exclusively on the food access dimension of food insecurity, typically measured through financial constraints.
This narrow focus means these tools likely underestimate true prevalence by failing to assess three of the four core dimensions of food security: food availability, stability over time, and food utilization.
Data Scarcity and Methodological Challenges
Underlying these measurement challenges lie fundamental data limitations. The FAO's traditional methodology for estimating undernourishment relies heavily on food balance sheets—aggregate national accounting of food supplies—supplemented where possible by household consumption and expenditure surveys.
Yet food balance sheets carry documented accuracy issues. They struggle to account for non-commercial food production, properly quantify losses, track unreported trade flows, and manage statistical uncertainty around stocks and distribution.
Household surveys, while providing more granular information, introduce different problems. Data collection over short reference periods generates significant measurement error.
Respondents may misremember quantities consumed, struggle with unfamiliar measurement units, or interpret survey questions inconsistently across cultures and different livelihood contexts. Where surveys do exist, they often cannot be disaggregated into sufficiently small geographic units to enable precise targeting of assistance..pdf)
For countries experiencing acute crises—exactly those most requiring accurate assessment—reliable data often becomes increasingly unavailable.
Insecurity, displacement, and deteriorating infrastructure make systematic data collection nearly impossible in regions where food insecurity peaks. Thus, the assessments meant to trigger emergency response occur precisely where information becomes least reliable.
The Funding Implications
The measurement gap translates directly into humanitarian impact. If 66 million fewer people are being identified as requiring emergency assistance than actually need it, humanitarian agencies receive requests for aid that underestimate actual necessity by hundreds of millions of dollars annually.
This mismatch intensifies at a moment when international humanitarian funding is contracting sharply.
In 2025, the World Food Programme faces a 40 percent reduction in funding compared to 2024, creating what the organization describes as an "unprecedented crisis for tens of millions across the globe reliant on food aid." With more than 58 million people at immediate risk of losing life-saving assistance across the WFP's 28 most critical crisis response operations, further funding cuts would mean cuts to program coverage affecting people whose conditions fall just below official crisis classifications.
This creates a cascading problem. If official counts already underestimate the actual number of acutely food-insecure people, and funding must now be allocated based on these already-conservative estimates while total resources shrink, the practical coverage gap widens dramatically.
People identified as experiencing phase 2 (stressed) conditions rather than phase 3 (crisis) conditions receive lower priority in emergency response planning and resource allocation.
The World Food Programme's experience reflects this dynamic across multiple regions. In Nigeria, 33.1 million people are expected to face severe food shortages during the 2025 hunger season, yet WFP estimates it requires $620 million to continue even current levels of food assistance across the Central Sahel region.
In Bolivia, climate-driven food insecurity among rural and Indigenous populations has intensified, but emergency response funding depends on classifications that may already undercount the affected population by significant margins.
Toward Improved Assessment
Recent research suggests machine learning and advanced analytical approaches could improve classification accuracy.
Researchers exploring semi-supervised machine learning techniques that combine labeled IPC assessments with unlabeled data from satellite imagery, market prices, and other indicators found significant improvements in predictive performance compared to models relying solely on labeled training data.
The Nature Food study itself does not recommend replacing human analytical judgment with purely algorithmic approaches.
Rather, researchers suggest that machine learning could enhance data collection and modeling within the existing IPC framework, helping to identify patterns that human committees might overlook and providing additional evidence during periods of uncertainty.
Improving data collection represents another avenue. The FAO notes that more precise household survey modules, particularly regarding food consumption, could reduce measurement error that artificially inflates the variance in population-level estimates.
Greater standardization in how working groups apply convergence-of-evidence protocols and explicit documentation of decision-making processes could increase consistency and reduce the bias toward conservative classifications.
Global Impact and Regional Variation
The underestimation appears to affect multiple regions differently. In Africa, the proportion of population facing hunger surpassed 20 percent in 2024, with 307 million people experiencing hunger—more than one-fifth of the continent's population.
In Western Asia, hunger rates have similarly increased. Yet recent progress in Latin America and South Asia demonstrates that food insecurity is not inevitable; regions with policy interventions and adequate resources have managed reductions.
Brazil's removal from the global hunger map in 2025, based on reducing undernourishment to below 2.5 percent of population, illustrates what sustained policy commitment and resource allocation can accomplish.
The country achieved this transition over a three-year period (2022-2024) despite economic headwinds and global commodity price volatility. The accomplishment required not just political will but adequate resources to implement food assistance, social protection, and agricultural development programs.
The Scale of Need and Humanitarian Response
Global hunger statistics, even in their current form, indicate a crisis of staggering proportions. In 2023, approximately 765 million people lacked sufficient food to meet basic needs, with nearly one-third—roughly 250 million—experiencing acute food insecurity severe enough to threaten their lives.
By 2024, hunger affected 307 million people in Africa, 323 million in Asia, and 34 million in Latin America and the Caribbean.
If measurement systems are underestimating by one-fifth, the actual numbers are even more severe. The effective global population in acute food insecurity would exceed 300 million rather than official figures around 250 million.
The gap between assessed humanitarian needs and available funding, already described as substantial, becomes genuinely catastrophic when actual need exceeds official assessment by this magnitude.
The disconnect between official measurements and actual need matters because humanitarian response planning and resource allocation flow from assessment data. When committees classify an area as phase 2 rather than phase 3, it receives fundamentally different emergency response priorities.
When global IPC analyses undercount by 66 million people, international pledges of humanitarian support are made against estimates that systematically misrepresent reality.
Understanding the Problem as First Step
The finding that global food insecurity measurement systematically underestimates rather than overstates the problem represents both a methodological correction and a call to action.
It acknowledges that the professionals operating the IPC system, working within genuine constraints of data scarcity and analytical uncertainty, have made rational choices under difficult circumstances. Yet these rational individual decisions accumulated into a systemic bias that obscures the true magnitude of global hunger.
Addressing this requires simultaneous work on multiple fronts. Improving underlying data collection, particularly household-level surveys that capture granular information about food access and consumption, would reduce the analytical uncertainty that currently drives conservative classifications.
Refining analytical protocols to more transparently handle conflicting evidence could reduce the tendency toward lower classifications when data conflict. Deploying machine learning tools to identify patterns in available data while maintaining human oversight could provide additional evidence during complex analytical situations.
Most fundamentally, however, the finding underscores that food insecurity assessments serve not abstract academic purposes but determine real humanitarian response and life-or-death resource allocation decisions.
Understanding that current figures likely underestimate the global population of food-insecure people clarifies that the scale of humanitarian need is even more severe than official declarations suggest. This recognition should inform both the urgency of humanitarian response and the level of resources required to address global food insecurity adequately.
The world's most vulnerable populations cannot be served by measurement systems that, in the face of uncertainty, systematically undercount their numbers.
As international humanitarian support shrinks while hunger spreads, ensuring that assessment methods capture actual need rather than conservative estimates becomes not merely a technical concern but a fundamental question of humanitarian response adequacy.
