Credit Access is Associated with Improved Food Security in the Horn of Africa
Authors: Jordi Cerdà-Bautista, Vasileios Sitokonstantinou, José Manuel Veiga López-Peña, Duccio Piovani, José María Tárraga, Gustau Camps-Valls
Organizations: Image Processing Laboratory, Universitat de València, Spain · Artificial Intelligence Group, Wageningen University & Research, Wageningen, The Netherlands · Food Security Unit, Joint Research Centre (JRC), European Commission, Ispra, Italy · Early Warning & Forecasting Unit, World Food Programme (WFP), United Nations, Rome, Italy · Internal Displacement Monitoring Centre (IDMC), Geneva, Switzerland
The intensification of climate change poses a growing threat to food security, especially in vulnerable communities. This study employs an observational machine-learning framework to estimate the causal association between access to credit and acute food insecurity in Somalia and across the Horn of Africa, drawing on a harmonized dataset spanning key environmental, socioeconomic, and conflict-related factors from 2015 to 2022. Results indicate that greater credit access is associated with a 2% reduction in acute food insecurity at the population level over the study period. Given that, on average, 16% of the population is in crisis, this effect represents a meaningful shift within the at-risk group. We interpret these estimates under explicit identification assumptions and complement them with robustness and refutation tests. The results provide context-specific evidence on how financial access correlates with food security outcomes in data-scarce, crisis-affected settings, and offer a transparent framework for integrating heterogeneous data sources when randomized evaluations are infeasible.
Figures & tables
Figure 1 : Study area and data sources. (A) Map of Somalia showing districts with available data, highlighting Baidoa, an important city in the Bay area (shown in the map), as an example. (B) Standardized time series and box plots displaying key environmental and socioeconomic variables for Baidoa, illustrating temporal trends and variability. (C) The annual distribution of IPC phases (percentages) for Baidoa shows shifts in food security over time.
Table 1: Variables and sources used in the study, spatial and temporal resolutions.
Figure 3 : Treatment Effect of Credit Access on IPC 3+ populations across different data aggregations: yearly, seasonal, IPC-based, and monthly. Each panel represents a distinct aggregation level, showing ATE, ATT, and ATC estimates across treatment thresholds. The top section of each panel displays the distribution of treated- and control-group samples, while the bottom section shows treatment estimates with confidence intervals. The density plots on the right depict the distribution of ATE, ATT, and ATC values. These results highlight how treatment effects vary depending on data aggregation and treatment definition.
Figure 4 : Sensitivity analysis of confounders in causal effect estimation. The left column illustrates causal diagrams excluding a key confounder, while the right column shows the corresponding ATE estimates.
Figure 5 : Heterogeneity analysis of the causal effect of credit access on IPC 3+ populations under two crisis scenarios using the IPC-based data aggregation. The top panel shows the distribution of treated and control group samples across different treatment thresholds, distinguishing between crisis and non-crisis conditions. The bottom panel presents the estimated Average Treatment Effect (ATE) for each threshold and crisis scenario, along with the distribution of all ATE estimations.
Appendix figures & tables8 assets
Supplementary material from the paper’s appendix.
Appendix
Figure 6 : Comparison of food insecurity levels in included vs. excluded districts. IPC 3+ values are normalized by district population. Blue: included districts (n=56). Orange: excluded districts (n=16).
Figure 7 : Kernel density estimates of propensity scores across four data aggregations. Rows correspond to data aggregation (yearly, seasonal, monthly, IPC-based). Columns correspond to treatment thresholds (30th/45th/60th/75th percentiles), with increasing strictness from left to right.
Humanitarian targeting and household responses are triggered by distress signals.
IPC Manual (2021) FAO (2023) iDMC (2020)
Appendix
Table 2: Justification of causal links in the DAG based on literature.
Confounder removed
Mean Δ ATE
95% CI of Δ ATE
p-value
Food prices
-0.002
[-0.003, -0.001]
0.549
Livestock prices
0.001
[0.000, 0.002]
0.561
Sorghum Production
0.001
[0.000, 0.002]
0.527
Drought IDPs
-0.001
[-0.002, -0.001]
0.525
Appendix
Table 3: Average difference in ATE between the full and LOO models across thresholds.
Cause Effect Estimation
Refutation Tests
Placebo
RCC
RSR
Th
Method
ATE
CI
p-value
Effect*
p-value
Effect*
p-value
Effect*
p-value
30
LR
-0.038
(-0.068, -0.010)
0.014
0.000
0.008
-0.038
0.988
-0.038
0.947
30
M
-0.009
(-0.041, 0.010)
0.518
0.006
0.563
-0.019
0.052
-0.013
0.630
30
IPS W
-0.024
(-0.087, 0.036)
0.447
0.001
0.125
-0.029
0.030
-0.029
0.481
30
T-L
-0.038
(-0.058, -0.016)
0.019
0.001
0.220
-0.039
0.318
-0.039
0.866
Appendix
Table 4 : Treatment threshold, method, ATE estimates, 95% confidence intervals and p-values. Refutation tests fail if their p-value is less than 0.05. Yearly dataset.
Cause Effect Estimation
Refutation Tests
Placebo
RCC
RSR
Th
Method
ATE
CI
p-value
Effect*
p-value
Effect*
p-value
Effect*
p-value
30
LR
-0.036
(-0.059, -0.015)
0.001
0.000
0.980
-0.036
0.992
-0.036
0.986
30
M
-0.033
(-0.056, -0.011)
0.004
0.004
0.920
-0.032
0.802
-0.033
0.982
30
IPS W
-0.032
(-0.075, 0.007)
0.133
0.001
0.420
-0.032
0.787
-0.032
0.938
30
T-L
-0.050
(-0.073, -0.024)
0.001
0.000
0.480
-0.047
0.009
-0.047
0.614
Appendix
Table 5 : Treatment threshold, method, ATE estimates, 95% confidence intervals and p-values. Refutation tests fail if their p-value is less than 0.05. Seasonal dataset.
Cause Effect Estimation
Refutation Tests
Placebo
RCC
RSR
Th
Method
ATE
CI
p-value
Effect*
p-value
Effect*
p-value
Effect*
p-value
30
LR
-0.023
(-0.044, -0.004)
0.018
0.000
0.002
-0.023
0.995
-0.023
0.989
30
M
-0.029
(-0.051, -0.008)
0.007
0.002
0.020
-0.030
0.574
-0.028
0.913
30
IPS W
-0.015
(-0.056, 0.027)
0.470
0.000
0.201
-0.021
0.001
-0.021
0.003
30
T-L
-0.041
(-0.066, -0.018)
0.001
-0.000
0.430
-0.041
0.927
-0.041
0.974
Appendix
Table 6 : Treatment threshold, method, ATE estimates, 95% confidence intervals and p-values. Refutation tests fail if their p-value is less than 0.05. IPC-based dataset.
Cause Effect Estimation
Refutation Tests
Placebo
RCC
RSR
Th
Method
ATE
CI
p-value
Effect*
p-value
Effect*
p-value
Effect*
p-value
30
LR
-0.024
(-0.043, -0.003)
0.024
0.000
0.980
-0.025
0.996
-0.024
0.988
30
M
-0.026
(-0.049, -0.006)
0.012
0.001
0.980
-0.029
0.465
-0.026
0.876
30
IPS W
-0.012
(-0.050, 0.026)
0.541
-0.000
0.217
-0.017
0.001
-0.017
0.009
30
T-L
-0.021
(-0.038, -0.018)
0.001
-0.000
0.490
-0.021
0.942
-0.021
0.976
Appendix
Table 7 : Treatment threshold, method, ATE estimates, 95% confidence intervals and p-values. Refutation tests fail if their p-value is less than 0.05. Monthly dataset.
Climate variability influences whether a market disruption escalates into a food crisis, yet broad climate patterns like El Niño, tracked months before they alter hydro-climatic conditions, are still not incorporated as an early-warning component in food-security responses. We address this gap by introducing sensitivity regimes, a stratification of regions by the direction and strength of their vegetation response to the El Niño Southern Oscillation, and using them to estimate how food price spikes affect acute food insecurity across sub-Saharan Africa. Integrating remote sensing, socioeconomic data, and causal machine learning, we find that in regions where ENSO systematically suppresses vegetation, a price spike raises the share of the population at acute risk by 5.4 percentage points in the following month. In regions where vegetation is unaffected by or positively linked to ENSO, the estimated effect is smaller (around 2 percentage points) and statistically insignificant. These results demonstrate that climate context is critical for understanding food security vulnerabilities. Sensitivity regimes can be combined with operational price-spike triggers to stage anticipatory action: the ENSO state flags vulnerable regions months ahead, and a pre-positioned response in those regions to a price spike would avert the largest jump in acute food insecurity.
Jordi Cerdà-Bautista, Vasileios Sitokonstantinou, Homer Durand +3
Food access is multidimensional, but population-level measurement still relies heavily on geography because perceived dimensions of access are difficult to measure at scale. Here, we use 25,125 Google Maps reviews from 49 grocery stores in Raleigh, North Carolina, to measure five dimensions of food access: availability, accessibility, affordability, accommodation, and acceptability. We identify review topics with unsupervised topic modeling and assign them to access dimensions using zero-shot classification, with 85.4% agreement against manual coding. The resulting store-level measures capture distinct aspects of food access and reveal barriers that geographic proximity alone does not capture. Comparisons between nearby stores in the same chain further show that identical store policies can be perceived very differently across locations, consistent with food access reflecting the fit between residents and their food environment. Perceived food access also follows systematic socioeconomic and demographic patterns that broadly parallel, but do not replicate, those observed for geographic access. These results show that online grocery reviews can provide a scalable complement to geographic measures of food access.
Progress in inclusive household surveys has strengthened socioeconomic evidence for forcibly displaced populations, providing indispensable benchmarks on living conditions and welfare. However, these surveys remain resource-intensive and periodic, while conditions can change between rounds, particularly in settings affected by fragility, conflict, and violence. More frequently updated, spatially granular complementary evidence is therefore needed to identify where socioeconomic conditions may be changing between survey rounds and to inform operational prioritization. Earth observation and machine learning offer a scalable source of spatially explicit socioeconomic information. However, tools developed for general populations have not been systematically adapted and evaluated in forced displacement settings, where living conditions, settlement patterns, and displacement impacts may differ substantially. We address this gap by adapting a multimodal spatiotemporal vision transformer, pretrained on Demographic and Health Survey data from approximately 1.2 million households across 36 African countries, to forced displacement and host community settings in South Sudan, Cameroon, and Zambia. We develop and evaluate the updated, adapted model using socioeconomic indices derived from UNHCR FDS and RMS data. Our results show that satellite-derived geospatial covariates explain up to 66% of the variation in socioeconomic outcomes in camp-intersecting grids, with a mean absolute error (MAE) of 4.37 index points, and 41% in non-camp-intersecting areas, with an MAE of 5.41. The framework complements and adds value to periodic household surveys by filling critical spatial and temporal data gaps with regularly updated, model-based socioeconomic estimates. These estimates sustain insight between survey rounds and support timely humanitarian prioritization and field verification.
Steven Ndung'u, Adel Daoud, Ismael Yacoubou Djima +2