Global Hunger Research Project

Understanding Global Hunger: A Comprehensive Research Initiative

Why this dashboard exists

Hunger is rarely caused by one thing. This dashboard brings together economic, health, climate, and crisis indicators so you can explore how they relate to food insecurity across the world. It’s built for quick exploration and for deeper, country-by-country investigation.

Core research question:

What factors drive hunger and hunger outbreaks, and how will these factors change in the future?

Where to start

  • Interactive Map: Explore global patterns and click countries for details.
  • Overview: See global distributions, relationships, and top-risk countries.
  • Country Details: Review the score breakdown and the latest indicators for a selected country.
  • Time Series: Explore how key indicators change over time.
  • Data Explorer: Browse the integrated dataset and download views.
  • Statistical Analysis: Run correlation and clustering-style summaries.

Hunger Vulnerability Score (0–100)

The Hunger Vulnerability Score is a 0–100 composite that summarizes multiple risk dimensions (food security, poverty, economic capacity, health, climate vulnerability, and crisis exposure). For a full component-by-component breakdown, see Country Details .

Data Sources

All dataset citations and short descriptions are listed on the Data Sources page.

Research Context

Hunger and food insecurity remain among the most significant challenges facing humanity. Despite global progress in reducing poverty and improving food production, millions of people worldwide still experience chronic hunger, and acute food crises continue to emerge in various regions.

This research project aims to contribute to the global effort to understand and address hunger by:

  • Providing a comprehensive, data-driven assessment of hunger vulnerability
  • Identifying countries and regions at highest risk
  • Highlighting the multi-faceted nature of hunger (economic, health, climate, conflict)
  • Enabling evidence-based policy decisions and resource allocation

A quick note on missing data

Note: Some indicators may be missing or lagged for certain countries. Interpret scores as relative risk signals and use local context when making decisions.


Author: Garrett Zhou

Project: Global Hunger Research - 2024

Last Updated: June 2026

Risk summary

Distribution of countries across vulnerability tiers (left) and the five most vulnerable countries with scores above 50 (right). Interactive map filters do not apply on this tab.

Countries by vulnerability tier
Not affected by vulnerability or statistic filters on the map.

Top 5 most vulnerable countries
Highest scores among countries above 50; not affected by map filters.

Cross-country relationships

Each point is a country. Choose the outcome (Y), predictor (X), and axis scale; the scatter, least-squares line, and equation update automatically. Interactive map filters do not apply.

Indicator scatter plot

Score distribution (full dataset)

Histogram uses all countries in the integrated dataset (not only filtered rows) for a stable global reference.

Vulnerability score — global histogram

Data Explorer
Explore all integrated datasets


Complete Dataset

Scroll horizontally to view all columns. Use filters to search and sort. Download as CSV or Excel.


Statistical visualizations

Distribution histograms for key indicators. Hover a bar for bin details; use the camera icon to download a chart.

Distribution overview


Additional indicators


Summary Statistics Table

How the vulnerability score is calculated (same rules as country profiles)

Read this while you explore the map. The score has twelve parts . Each part adds points toward a total from 0 to 100. The final score is the sum of all parts, capped at 100 (Formula: min(100, sum of all pillars) ). Part multipliers (below the map) change how much each part counts on this map only : multiplier 1 = published weights, 0 = that part turned off, 2 = double that part’s points.

  1. Undernourishment (maximum 25 points): Multiply the percentage of people who are undernourished by 0.25, then cap at 25. If undernourishment data are missing, use the poverty percentage the same way.
    Formula: min(PoU% × 0.25, 25); if PoU missing, min(poverty% × 0.25, 25)
  2. Poverty (maximum 8 points): Multiply the poverty percentage by 0.16, then cap at 8. Poverty uses the World Bank $1.90-per-day line or, when needed, the Our World in Data $3-per-day line.
    Formula: min(poverty% × 0.16, 8) (WB $1.90 or OWID $3 line)
  3. Income per person (maximum 7 points): Points depend on gross domestic product per person in United States dollars: below $1,000 → 7 points; below $3,000 → 5 points; below $10,000 → 3 points; below $20,000 → 1 point.
    Formula: GDP per-capita USD buckets — under 1k→7, under 3k→5, under 10k→3, under 20k→1
  4. Life expectancy (maximum 5 points): If life expectancy is under 50 years, add 5 points; under 60 years, add 4 points; under 70 years, add 2 points.
    Formula: under 50y→5, under 60→4, under 70→2
  5. Child stunting (maximum 5 points): Multiply the child stunting percentage by 0.05, then cap at 5.
    Formula: min(stunting% × 0.05, 5)
  6. Climate vulnerability (maximum 10 points): If the climate vulnerability index is at least 60, add 3 points; at least 70, add 7 points; at least 80, add 10 points.
    Formula: vulnerability index 60/70/80 thresholds → 3 / 7 / 10 pts
  7. Conflict intensity (maximum 10 points): If there is active conflict: very high intensity adds 10 points, high adds 7, medium adds 4, low adds 2.
    Formula: if active — Very High→10, High→7, Medium→4, Low→2
  8. Major hunger crises and food security phase (maximum 15 points): Add 15 points if the country had a major hunger crisis in the 21st century or is in food security phase 4 or higher. Add 8 points if it is in food security phase 3.
    Formula: major 21st-c. hunger crisis OR IPC phase 4+→15; phase 3→8
  9. Food import dependency (maximum 5 points): Points increase when average food import share reaches 50%, 30%, or 20% thresholds.
    Formula: avg food import share tiers (50% / 30% / 20%)
  10. Food supply (maximum 5 points): Points increase when daily food supply per person falls below 2,400, below 2,200, or below 2,000 calories.
    Formula: kcal/cap/day under 2000 / 2200 / 2400
  11. Water stress (maximum 5 points): Points increase when renewable water per person falls below 1,700, below 1,000, or below 500 cubic meters per year.
    Formula: renewable m³/cap under 500 / 1000 / 1700
  12. Forced displacement (maximum 5 points): Points are based on refugees and internally displaced people, using either their share of the population or absolute numbers.
    Formula: refugees/IDPs as share of population or absolute size tiers

Global Hunger Vulnerability Map (0-100 Scale)

Score part multipliers

Map score shown = sum of (multiplier × points for each part), capped at 100 (Formula: min(100, Σ (multiplier × pillar points)) ). Multiplier 1 on every slider matches the published score. 0 turns a part off on this map; 2 doubles that part. Country profile pages always use the published score.

Map Filters

Limit which countries appear on the map. Countries missing a selected statistic may be hidden when that filter is enabled.

Map Legend

Vulnerability Score (0-100):

🟢 0-25: Low Vulnerability (Green)

🟡 25-50: Moderate Vulnerability (Yellow)

🟠 50-75: High Vulnerability (Orange)

🔴 75-100: Critical Vulnerability (Red)


Note: Map colors use the multipliers under the map; hover shows the vulnerability score and each formula pillar. Countries excluded by filters remain as outlines only (no fill or hover).

Instructions

How to use this map:

1. Use the formula box at the top for coefficients; adjust multipliers under the map to explore.

2. Hover over countries for stats and the map score.

3. Click a country for the full country profile.

4. Use the sidebar to pick countries; open Map Filters below the multipliers for year, score, and statistic filters.

5. Green → red = lower → higher displayed vulnerability.

Your scenario — country landscape

Imaginary country — inputs

Set raw statistics as if this were a single country. Pillars use the same rules and coefficients as the dashboard. Multipliers (right) scale each pillar’s points: ×1 = published index, ×0 = off, ×2 = double.

Conflict points apply only when this is checked.

Pillar multipliers & result

Displayed score = min(100, Σ multiplier × pillar) . Default ×1 matches the published formula (same coefficients as the Map tab reference box).

Select Country


Click Clear to delete the search and type a new country name.

Vulnerability score breakdown

Key insights

Vulnerability over time

Based on indicators available by year. Not every score component has a full time series.

Main score drivers

Share of each factor in the total score. Zero-weight factors are omitted.

Current year summary

Historical time series

Bangladesh, Climate Change & Food Security

This page summarizes the author’s North Carolina Youth Institute / World Food Prize research paper on how climate stressors interact with food security in Bangladesh — a densely populated delta highly exposed to floods, cyclones, and sea-level rise.

Goal

  • Explain how climate change undermines food security in Bangladesh, using both narrative evidence and quantitative analysis.
  • Estimate how climate vulnerability relates to undernourishment across countries, then interpret what that implies for Bangladesh.
  • Evaluate policy options — especially nutrition programs and climate-resilient agriculture — and outline a data-informed recommendation for investment priorities.

Background

Bangladesh sits on the Ganges–Brahmaputra–Meghna delta: extreme monsoon rainfall, river flooding, cyclones, and salinity intrusion threaten crops, drinking water, and livelihoods. Despite progress (e.g. lower undernourishment than a decade ago), the country remains highly exposed; vulnerability can rise quickly after major shocks, with hunger metrics following. The paper connects this geography to a typical family narrative, national statistics, and international context (aid, delta planning, and political transitions).

Methodology

  1. Cross-country regression: Relates climate vulnerability (e.g. ND-GAIN-style index) to national undernourishment rates to quantify the association (paper reports roughly 48% of variation explained in the sample used).
  2. Bangladesh time series: Pairs historical climate vulnerability with FAO undernourishment for Bangladesh (2002–2022) to describe periods of stability, shock (e.g. post-Sidr/Aila), and recovery.
  3. Policy optimization (Model 2): A constrained nonlinear allocation model over ~13 interventions (school feeding, saline- and flood-tolerant rice, AWD irrigation, storage, insurance, early warning, etc.) with diminishing returns, synergy terms, budget caps, and rules such as a BNP funding floor — calibrated to Bangladesh’s approximate caloric deficit and undernourished population.

Results (high level)

  • Regression: A fitted relationship implies that at Bangladesh’s vulnerability level, undernourishment would be higher without strong adaptation and aid; actual undernourishment has remained below that “prediction,” illustrating an adaptation buffer that can shrink after shocks.
  • Time series: Vulnerability spiked again in the early 2020s after a long recovery — a warning that headline hunger rates can lag vulnerability.
  • Optimization: Under documented assumptions (~18M undernourished, ~700 kcal/day gap, ~4.6 trillion kcal annual deficit, $8B budget, 10-year horizon, synergies), the model allocates heavily to direct nutrition (fortified feeding), saline-tolerant rice, storage, and irrigation-style interventions while respecting policy constraints; BNP-scenario budgets show similar patterns with modest reallocation.

Full narrative, citations, and tables are in the project paper and supporting files (e.g. Garrett Zhou WFP Bangladesh Paper D3.txt , model2 wfp.md , model2_results.txt ).

In this dashboard

Open Country Details and select Bangladesh to see live indicators (undernourishment, GRFC/IPC, disasters, climate vulnerability, poverty, and more) that complement the paper.

About the Global Hunger Index (GHI)

What is the Global Hunger Index?

The Global Hunger Index (GHI) is a peer-reviewed annual report that comprehensively measures and tracks hunger at the global, regional, and national levels. The GHI is calculated annually by Welthungerhilfe (WHH) and Concern Worldwide , with data support from the International Food Policy Research Institute (IFPRI).

The GHI was first published in 2006 and has since become one of the most widely recognized tools for measuring hunger worldwide. It provides a standardized way to compare hunger levels across countries and track progress over time.


Purpose and Background

The GHI was created to:

  • Raise awareness and understanding of the problem of hunger
  • Provide a way to compare hunger levels across countries and regions
  • Track progress in reducing hunger over time
  • Encourage increased attention to and action against hunger
  • Provide policymakers with data to inform decision-making

How GHI Works

The GHI score is calculated using a formula that combines four equally weighted indicators:

  • Undernourishment (33.3%): The proportion of the population that is undernourished (lacking sufficient caloric intake). This is the primary indicator of hunger.
  • Child Wasting (16.7%): The proportion of children under five years old who are wasted (low weight for their height), indicating acute malnutrition.
  • Child Stunting (33.3%): The proportion of children under five years old who are stunted (low height for their age), indicating chronic malnutrition.
  • Child Mortality (16.7%): The mortality rate of children under five years old, which often reflects the fatal combination of inadequate nutrition and unhealthy environments.

The GHI score ranges from 0 to 100, where:

  • 0-9.9: Low hunger
  • 10-19.9: Moderate hunger
  • 20-34.9: Serious hunger
  • 35-49.9: Alarming hunger
  • 50+: Extremely alarming hunger

Official GHI Website: https://www.globalhungerindex.org/

Our Vulnerability Score vs. GHI: A Detailed Comparison

Methodology Comparison

🌍 Our Vulnerability Score

Components (12 factors; max points sum to 100):

  • Undernourishment: 25 points (25% of total score)
  • Poverty and income per person: 15 points (15%)
  • Life expectancy: 5 points (5%)
  • Child stunting: 5 points (5%)
  • Climate vulnerability: 10 points (10%)
  • Conflict intensity: 10 points (10%)
  • Historical hunger crises and food security phase: 15 points (15%)
  • Food import dependency: 5 points (5%)
  • Natural resources (food supply and water stress): 10 points (10%)

Scale: 0-100

Focus: Comprehensive multi-factor assessment including economic, social, environmental, and crisis indicators

📊 Global Hunger Index (GHI)

Components (4 factors):

  • Undernourishment (33.3%)
  • Child wasting (16.7%)
  • Child stunting (33.3%)
  • Child mortality (16.7%)

Scale: 0-100

Focus: Core nutrition and child health indicators


Advantages and Disadvantages

✅ Advantages of Our Vulnerability Score
  • Comprehensive: Incorporates 12 factors; max points sum to 100, with percentages shown for clarity
  • Predictive: Includes economic indicators (GDP, poverty) that can predict future vulnerability
  • Context-aware: Considers conflict and disasters that directly impact food security
  • Forward-looking: Can identify countries at risk before acute hunger crises occur
  • Multi-dimensional: Captures economic, social, environmental, and crisis dimensions of hunger
⚠️ Limitations of Our Vulnerability Score
  • Complexity: More factors can make interpretation less straightforward
  • Data dependency: Requires data from multiple sources, some of which may be incomplete
  • Weighting: Subjective decisions about point allocations for different factors
  • Less established: Not as widely recognized or validated as GHI
  • Update frequency: Depends on multiple data sources with different update schedules
✅ Advantages of GHI
  • Established: Widely recognized and trusted by policymakers and researchers
  • Focused: Clear focus on core nutrition indicators
  • Standardized: Consistent methodology since 2006, allowing for reliable trend analysis
  • Child-focused: Emphasizes child nutrition, which is critical for long-term development
  • Peer-reviewed: Annual publication with rigorous methodology review
⚠️ Limitations of GHI
  • Limited scope: Only four indicators, may miss important contextual factors
  • Reactive: Primarily measures current hunger rather than predicting future risk
  • Economic blind spot: Does not directly incorporate economic indicators like GDP or poverty
  • Crisis factors: Does not account for conflict, disasters, or trade disruptions
  • Data lag: Annual publication means data may be 1-2 years old

💡 Key Insight

Both measures serve important but complementary purposes. The GHI provides a focused, standardized measure of current hunger levels, while our vulnerability score offers a more comprehensive assessment that includes predictive factors and contextual risks. Using both together provides the most complete picture of a country's food security situation.

Score Comparison: Our Vulnerability Score vs. GHI

Score Distribution Comparison

Score Correlation Analysis

Time Series Analysis
Track trends and forecast future values

Interactive Time Series

Select a variable to view global (World) trends from the World Bank World aggregate (ISO code WLD)—not a sum or unweighted average of countries. Enable forecasting to see projected values.

Toolbar Guide:

  • 📷 Camera: Download graph as JPG
  • 🔍 Zoom: Click and drag to zoom, double-click to reset
  • 📊 Pan: Click and drag to pan around
  • 📐 Select: Select data points
  • ↔️ Auto-scale: Reset axes to fit data
  • 📋 Reset axes: Reset to default view

Variable Selection


Tip: Select different variables to compare trends across key indicators.

Forecast Options ? How it's calculated: Trends are fit on the most recent 15 years (minimum 5 points). Population and total GDP use log-linear (compound) growth; other variables use linear change per year. Each forecast step extends from the latest observed value using that recent trend—not from the full-history regression line—so the dashed line connects to the last data point. Illustrative only, not predictions.

Note: Projections are illustrative, not predictions.

Statistical Analysis
Advanced statistical methods for hunger research

Background & Purpose

Statistical analysis helps us understand the relationships between different factors that contribute to global hunger. By applying rigorous statistical methods, we can identify which variables are most strongly associated with hunger vulnerability, test hypotheses about differences between groups, and quantify the strength of relationships between indicators.

What we analyze: We examine correlations between economic, social, and health indicators; compare high-risk (vulnerability score 50–75) and low-risk (< 25) countries using statistical tests; and report OLS regressions of FAO undernourishment (%) on the same raw inputs used in the vulnerability formula (see Regression tab below).

Correlation Analysis

What is a Correlation Matrix?

A correlation matrix is a table showing correlation coefficients between multiple variables. Each cell shows how strongly two variables are related:

  • Values range from -1 to +1: +1 means perfect positive correlation (as one increases, so does the other), -1 means perfect negative correlation (as one increases, the other decreases), and 0 means no relationship.
  • Color coding: Red indicates positive correlations, blue indicates negative correlations. Darker colors mean stronger relationships.
  • Why it matters: Understanding correlations helps identify which factors move together, which can inform policy and intervention strategies.

Regression: FAO undernourishment vs. vulnerability inputs

Outcome: FAO prevalence of undernourishment (%) — not the composite 0–100 dashboard vulnerability score. Method: ordinary least squares (OLS) on a cross-section of countries; predictors are the same raw inputs used in the vulnerability formula.

  • Univariate: each predictor is regressed on undernourishment one at a time (simple linear model).
  • Multivariate: all predictors enter a single model jointly; coefficients are conditional on the others in that specification.
  • Interpretation: results describe statistical associations, not proven causes; use them to explore relationships, not to justify dropping formula variables.
Reading the tables
Estimate
Expected change in undernourishment (%) per one-unit increase in the predictor.
Standard error
Uncertainty around the estimate; smaller values mean more precision.
p value
How surprising the estimate would be if there were no linear association; values below 0.05 are often treated as statistically significant.
R squared / Adjusted R squared
Share of variation in undernourishment explained by the model (0–1). Adjusted R squared accounts for the number of predictors.
Standardized coefficient
Coefficient on a common scale (standard deviations), for comparing strength across predictors (univariate table).
t statistic
Estimate divided by standard error (multivariate tables).
~0
Very small coefficients are shown as ~0 ; click to reveal the exact value.

Statistical Tests

About Statistical Tests

We perform statistical tests to rigorously examine differences and relationships in our data:

  • T-tests : Compare means between two groups (e.g., high-risk 50–75 vs. low-risk < 25) to determine if differences are statistically significant.
  • Correlation Tests : Measure the strength and direction of relationships between two variables (e.g., GDP and poverty).

Interpreting Results: A p-value less than 0.05 typically indicates statistical significance, meaning the observed relationship or difference is unlikely due to chance alone.

Hunger Risk Prediction Model

Model Parameters

Model Performance


                  

Feature Importance

About Us
Meet the research team

Global Hunger Research Project

A collaborative research initiative at Duke University

Student Researcher

Garrett Zhou

Garrett Zhou

Class of 2027 High Schooler

Affiliation

Durham Academy
Durham, North Carolina

Background

Garrett is a high school student passionate about addressing global hunger and food insecurity. Through this research project, he aims to understand the complex factors driving hunger worldwide and develop data-driven solutions to help combat this critical issue.

Research Interests

  • Global hunger and food insecurity
  • Data-driven policy solutions
  • Food systems and sustainability
  • Public health and nutrition

Contact & Links

[email protected]
Blog: A Grain Of Change
YouTube Channel

Faculty Mentor

Professor Hannah Jacobs

Professor Hannah Jacobs

Professor of Duke Libraries

Affiliation

Duke University
Duke Libraries

Research Focus

Professor Jacobs specializes in Digital Humanities, bringing expertise in information science and digital methodologies to support innovative research projects. Her work bridges traditional humanities scholarship with modern digital tools and approaches.

Education & Expertise

  • MS in Information Science, UNC Chapel Hill
  • MA in Digital Humanities, King's College London
  • BA in English and Theatre, Warren Wilson College
  • Digital Humanities topics broadly
  • Information science and data management
  • Digital scholarship and research methodologies

Contact

[email protected]
Duke Libraries

Research Collaboration

About This Collaboration

This research project represents a unique collaboration between a high school student passionate about addressing global hunger and a Duke University professor specializing in Digital Humanities. Through this mentorship, Garrett is learning to apply data science and digital research methodologies to understand complex global challenges.

The project combines rigorous statistical analysis with interactive data visualization to make hunger research accessible to policymakers, researchers, and the public. By integrating multiple data sources and applying advanced analytical techniques, this work aims to provide actionable insights into the factors driving global hunger and food insecurity.

Mentorship Approach: Professor Jacobs provides guidance on research methodology, data analysis techniques, and digital scholarship practices. Her expertise in Digital Humanities helps bridge traditional research approaches with modern computational methods, enabling innovative approaches to understanding global hunger patterns.

Acknowledgments

We would like to express our gratitude to:

  • Duke University and Duke Libraries for providing the resources and support for this research
  • Professor Hannah Jacobs for her mentorship, guidance, and expertise in Digital Humanities
  • All data providers including the World Bank, FAO, WFP, WHO, and other organizations that make their data publicly available
  • The global research community working to address hunger and food insecurity

Data Sources
Chicago style + quick descriptions

Data Sources

This page lists every imported dataset currently integrated into the dashboard. Sources are formatted in Chicago style and each entry includes a short description of what we used it for.

Sources are listed in Chicago style with a short description of how each dataset is used in this dashboard. In data/raw badges show which files are present on this computer.

Data pipeline

Data refresh log: refreshed_at_utc=not_yet_run incoming_folder=none note=Run scripts/run_data_refresh_pipeline.sh from the project root to populate this file and refresh reports.

See docs/data_pipeline_implementation_plan.md , data/metadata/data_dictionary.csv , and scripts/run_data_refresh_pipeline.sh .

Data Coverage

Where data exist — and where they do not

This tab summarizes completeness across integrated countries (World Bank backbone, excluding aggregates). = value present; = missing. Use it to prioritize collection and to sanity-check merges.

Coverage by indicator — share of countries with a non-missing value (see Source column).

Indicator detail

Countries with fewest indicators

Lowest completeness among integrated countries (top 20).


Per-country matrix

Filter by country or region. Sort by % coverage to find sparse profiles quickly.