Impacts of Humanitarian Aid Delay and Cut on Welfare of Internally Displaced Persons (IDPs): Evidence from Sudan

Randomizing the timing of the survey
To study the impact of humanitarian assistance cuts and delays on the welfare of IDPs, we randomize the interview dates and create an exogenous variation in the timing of aid delays (the number of days passed between the last aid transfer and the interview date). To this end, we randomly assigned the 149 distribution centers into three distinct survey waves: Wave 1 includes 47 clusters, Wave 2 includes 52 clusters, and Wave 3 includes 50 clusters. The survey will be rolled out sequentially across these waves, with households in Wave 1 interviewed first and those in Wave 3 interviewed last.

The core idea behind this design is to leverage variation in the timing between the receipt of humanitarian assistance and the timing of the survey. All households are dependent on humanitarian aid (cash or in-kind), but the length of the gap between the last distribution and the survey interview varies across waves. Households in Wave 1 will be surveyed relatively soon after receiving assistance, while those in Wave 3 will be surveyed after a relatively longer interval. This time gap introduces experimental variation that allows us to assess how the delay in receiving aid—or the prolonged absence of it—affects key outcomes, such as food security, subjective well-being, mental health, and willingness to pay to avoid such delays. By comparing outcomes across these waves, the study can disentangle the effects of humanitarian delays and funding shortfalls in real time, offering critical insights into the welfare consequences of disrupted aid in fragile and high-need settings, such as Sudan.

Randomizing information intervention
In addition to leveraging exogenous variation in the timing of the interview dates to assess the impact of humanitarian assistance delays, this study includes a second randomized intervention designed to evaluate the effects of an information treatment on internally displaced persons’ (IDPs) willingness to pay (WTP) for a hypothetical peace and stability restoration initiative. A randomly selected subset of respondents will be provided with factual, relatively high-salience information on the ongoing humanitarian crisis in Sudan, covering actual statistics, such as the scale of humanitarian needs, the number of internally displaced persons in their respective state, and cross-border migration flows. The key outcome of interest is whether exposure to this information influences respondents’ stated WTP to support peacebuilding efforts. This follows the empirical approach proposed by Domínguez and Scartascini (2024), which was used to elicit willingness to pay for price reduction in Latin America. In a context where global donor funding for peacebuilding and humanitarian efforts is shrinking, understanding local capacities and willingness to invest in collective action and public goods and services is crucial. Building on these experiences, we elicited WTP through a series of structured questions beginning with: “Considering your current financial situation, would you be willing to contribute 20,000 SDG per month to support the peace restoration initiative?” If the respondent answers “yes,” the contribution amount is progressively increased, up to a maximum of 40,000 SDG, to gauge their upper bound. Conversely, if the initial response is “no,” the amount is gradually decreased to a minimum of 1,000 SDG to measure their lower bound. This iterative approach allows for a more precise estimation of respondents’ WTP. The randomized information intervention is offered towards the end of the interview immediately before the WTP questions, the last module in the survey. Thus, we do not expect the information treatment to affect any other outcomes other than the IDP’s WTP.

We will estimate the causal effect of the information treatment on WTP and will explore heterogeneity in responses based on the length of delay in humanitarian assistance (which effectively introduces scarcity), using the randomly assigned survey waves as a proxy for the time since last aid receipt. This approach allows us to assess whether the impact of the information intervention varies by the level of hardship experienced due to aid delays, shedding light on how scarcity may moderate support for peace initiatives. Additional heterogeneity analyses will consider factors such as displacement history and demographic characteristics such as gender.

Specifically, in addition to the scenario description and the willingness-to-pay (WTP) prompt related to the ongoing conflict in Sudan, respondents in the information treatment group will be presented with the following factual statement: “Nearly 24.8 million people in Sudan are in urgent need of humanitarian assistance, while 8.8 million people have been internally displaced, marking the world’s largest displacement crisis. Additionally, approximately 3 million people, primarily women and children, have sought refuge in neighboring countries.”

Eliciting preferences for transfer modalities from both spouses
Another innovation we introduce in this survey is that we elicit the preference for humanitarian aid delivery modality from both spouses in the household. Specifically, this entails choice across cash, in-kind, digital, and value voucher transfers. Understanding IDPs’ preference for humanitarian aid delivery mechanisms is crucial for improving the effectiveness of humanitarian services. Building on the mixed evidence on whether men and women have different preferences for modality of assistance, we delve into these dynamics in an active conflict setting, whereby women represent the largest share of IDPs across Sudan. We hypothesize that preference for aid delivery modalities could differ across gender due to several factors, including but not limited to income level, access to markets and financial inclusion, storage facilities, conflict-induced lack of security, and ability to travel to market, and intrahousehold decision-making. For this reason, we administer the module on modality preferences for both spouses in the household.

Randomizing the timing of the survey
To study the impact of humanitarian assistance cuts and delays on the welfare of IDPs, we randomize the interview dates and create an exogenous variation in the timing of aid delays (the number of days passed between the last aid transfer and the interview date). To this end, we randomly assigned the 149 distribution centers into three distinct survey waves: Wave 1 includes 47 clusters, Wave 2 includes 52 clusters, and Wave 3 includes 50 clusters. The survey will be rolled out sequentially across these waves, with households in Wave 1 interviewed first and those in Wave 3 interviewed last.

The core idea behind this design is to leverage variation in the timing between the receipt of humanitarian assistance and the timing of the survey. All households are dependent on humanitarian aid (cash or in-kind), but the length of the gap between the last distribution and the survey interview varies across waves. Households in Wave 1 will be surveyed relatively soon after receiving assistance, while those in Wave 3 will be surveyed after a relatively longer interval. This time gap introduces experimental variation that allows us to assess how the delay in receiving aid—or the prolonged absence of it—affects key outcomes, such as food security, subjective well-being, mental health, and willingness to pay to avoid such delays. By comparing outcomes across these waves, the study can disentangle the effects of humanitarian delays and funding shortfalls in real time, offering critical insights into the welfare consequences of disrupted aid in fragile and high-need settings, such as Sudan.

Randomizing information intervention
In addition to leveraging exogenous variation in the timing of the interview dates to assess the impact of humanitarian assistance delays, this study includes a second randomized intervention designed to evaluate the effects of an information treatment on internally displaced persons’ (IDPs) willingness to pay (WTP) for a hypothetical peace and stability restoration initiative. A randomly selected subset of respondents will be provided with factual, relatively high-salience information on the ongoing humanitarian crisis in Sudan, covering actual statistics, such as the scale of humanitarian needs, the number of internally displaced persons in their respective state, and cross-border migration flows. The key outcome of interest is whether exposure to this information influences respondents’ stated WTP to support peacebuilding efforts. This follows the empirical approach proposed by Domínguez and Scartascini (2024), which was used to elicit willingness to pay for price reduction in Latin America. In a context where global donor funding for peacebuilding and humanitarian efforts is shrinking, understanding local capacities and willingness to invest in collective action and public goods and services is crucial. Building on these experiences, we elicited WTP through a series of structured questions beginning with: “Considering your current financial situation, would you be willing to contribute 20,000 SDG per month to support the peace restoration initiative?” If the respondent answers “yes,” the contribution amount is progressively increased, up to a maximum of 40,000 SDG, to gauge their upper bound. Conversely, if the initial response is “no,” the amount is gradually decreased to a minimum of 1,000 SDG to measure their lower bound. This iterative approach allows for a more precise estimation of respondents’ WTP. The randomized information intervention is offered towards the end of the interview immediately before the WTP questions, the last module in the survey. Thus, we do not expect the information treatment to affect any other outcomes other than the IDP’s WTP.

We will estimate the causal effect of the information treatment on WTP and will explore heterogeneity in responses based on the length of delay in humanitarian assistance (which effectively introduces scarcity), using the randomly assigned survey waves as a proxy for the time since last aid receipt. This approach allows us to assess whether the impact of the information intervention varies by the level of hardship experienced due to aid delays, shedding light on how scarcity may moderate support for peace initiatives. Additional heterogeneity analyses will consider factors such as displacement history and demographic characteristics such as gender.

Specifically, in addition to the scenario description and the willingness-to-pay (WTP) prompt related to the ongoing conflict in Sudan, respondents in the information treatment group will be presented with the following factual statement: “Nearly 24.8 million people in Sudan are in urgent need of humanitarian assistance, while 8.8 million people have been internally displaced, marking the world’s largest displacement crisis. Additionally, approximately 3 million people, primarily women and children, have sought refuge in neighboring countries.”

Eliciting preferences for transfer modalities from both spouses
Another innovation we introduce in this survey is that we elicit the preference for humanitarian aid delivery modality from both spouses in the household. Specifically, this entails choice across cash, in-kind, digital, and value voucher transfers. Understanding IDPs’ preference for humanitarian aid delivery mechanisms is crucial for improving the effectiveness of humanitarian services. Building on the mixed evidence on whether men and women have different preferences for modality of assistance, we delve into these dynamics in an active conflict setting, whereby women represent the largest share of IDPs across Sudan. We hypothesize that preference for aid delivery modalities could differ across gender due to several factors, including but not limited to income level, access to markets and financial inclusion, storage facilities, conflict-induced lack of security, and ability to travel to market, and intrahousehold decision-making. For this reason, we administer the module on modality preferences for both spouses in the household.

2025-07-22

2025-10-04

1. Food Consumption Score (FCS)
2. Women’s Dietary Diversity Score (WDDS)
3. Food Insecurity Experience Scale (FIES)
4. Perceived Stress Scale (PSS) and Subjective Well-being
5. Preference for modality of transfer (in-kind, cash, digital, and voucher transfers)
6. Shocks and Coping Mechanisms
7. Willingness to pay to avoid delays in humanitarian assistance
8. Willingness to pay for peace restoration and the WASH program

1. FCS, developed by WFP, is based on a 7-day recall and captures dietary quantity and diversity. It weighs eight food groups according to nutritional value and reflects dietary adequacy among most household members.

2. WDDS is based on a 24-hour recall following WFP’s practice and measures women’s food security through dietary diversity. Respondents are the main women representative of reproductive age (15-49 years) in the Household. They report their own consumption across 14 food groups over the past day.

3. FIES, developed by FAO (2014; 2020), categorizes households as food secure, moderately insecure, or severely insecure. It uses an 8-item questionnaire on access to sufficient and nutritious food in the last 30 days. Responses are summed into a score from 0 to 8.

4. Stress is measured using Cohen’s Perceived Stress Scale (PSS), a 10-item module scored 0–40. We construct two binary indicators for moderate-to-severe stress (PSS ≥14) and severe stress (PSS ≥27).

5. Subjective well-being is measured using data on self-efficacy, satisfaction, and direct questions on subjective well-being. To measure self-efficacy, we use the “New General Self-Efficacy Scale”, in which respondents are asked about the degree of agreement towards 8 statements on whether they feel they will be able to achieve most of their goals, if they are certain they will achieve them if they face difficulties, etc. We measure satisfaction through a set of 3 direct questions on how they view their economic circumstances compared to other households and thinking about their own separately. Subjective well-being is captured using three questions on happiness, satisfaction, and economic outlook.

6. Shocks and Coping Mechanisms are measured through asking direct questions on whether the respondent faced any of 8 shocks, the frequency of exposure, the impact of each shock, and the coping strategies.

7. Willingness to pay, where we ask the respondents to choose between different options for receiving aid. Each option includes changes to how often they receive aid, how much you receive, and the method through which the aid is delivered.

1. Preference for digital vs cash vs in-kind by gender
2. Access to Markets and Financial Services
3. Market price of food items provided by WFP in-kind assistance, their complements, and their supplements
4. Empowerment: Trust, Control, Agency, and Autonomy in Decision Making
5. Willingness to pay for the peace restoration initiative
6. Willingness to pay for the community-based WASH program initiative

The experiment will follow a stratified random sampling approach where 2,500 IDP households into three groups to be interviewed at different times during data collection (approximately one month), based on their respective distribution center. Each group of distribution centers is randomly assigned to a certain week of data collection, considering the size of their respective state and the modality of transfer being offered. The sample covers 149 distribution centers across three states: Blue Nile (90), Gedaref (21), and Kassala (38). Of which, 56 centers used to distribute cash, and 93 are currently distributing in-kind. Our randomization involves stratification by state as well as modality of transfer (in-kind versus cash). The sampling will involve selection of a maximum 18 IDP households from each distribution center. Figure 1 illustrates the sampling frame, the study sample, and the random assignment.

The experiment will follow a stratified random sampling approach where 2,500 IDP households into three groups to be interviewed at different times during data collection (approximately one month), based on their respective distribution center. Each group of distribution centers is randomly assigned to a certain week of data collection, considering the size of their respective state and the modality of transfer being offered. The sample covers 149 distribution centers across three states: Blue Nile (90), Gedaref (21), and Kassala (38). Of which, 56 centers used to distribute cash, and 93 are currently distributing in-kind. Our randomization involves stratification by state as well as modality of transfer (in-kind versus cash). The sampling will involve selection of a maximum 18 IDP households from each distribution center. Figure 1 illustrates the sampling frame, the study sample, and the random assignment.

The randomization will be conducted at the distribution center level.

Distribution Center level

Yes

149 Distribution Centers

2,200 households

Wave 1: 49 distribution centers; Wave 2: 53 distribution centers; Wave 3: 50 distribution centers

We compute the minimum number of clusters needed for each arm, considering the primary outcomes described above. Although our total distribution centers and IDP camps are fixed, computing the number of clusters needed for each treatment arm to detect meaningful impact can inform us about the relative size of our sample compared to an ideal setting. Our power calculation aims to achieve 80 percent power at a significance level of 5 percent and an average of 18 households per distribution center. We compute these power calculations considering only the primary outcomes described above. As described above, our outcomes of interest include Food Consumption Score (FCS); Women’s Dietary Diversity Score (WDDS); Food Insecurity Experience Scale (FIES); Perceived Stress Scale (PSS); Subjective Wellbeing (satisfaction, happiness, and economic conditions). We compiled the mean and standard deviation of these primary outcomes as well as minimum detectable effects (MDEs) for each outcome using evidence from previous studies in Sudan and comparable contexts (IFPRI and UNDP, 2024a; IFPRI and UNDP, 2024b). For each outcome, we compiled means and standard deviations using the Sudan Urban Household data in 2024 (IFPRI and UNDP, 2024b) and applied an intra-cluster correlation coefficient (ICC) of 0.1. We then estimated the minimum detectable effects (MDEs) required to identify statistically significant differences under plausible impact scenarios. The mean from previous studies for FCS is 68.96, with a standard deviation of 21.64. To detect a 7.48% decrease (5.16-point drop), a total of 86 clusters (45 treatment and 45 control) are needed. Similarly, the DDS has a mean of 7.04, and detecting a 0.3-point (4.27%) decline requires 90 clusters in total. For FIES (mean 3.22), the study is powered to detect a 0.64-point reduction (about 19.75%) using 80 clusters (40 per arm). Detecting a 1.42-point increase in perceived stress (6.46% rise) requires 92 clusters. Satisfaction and happiness indicators have smaller standard deviations and detecting 10.42% changes in those outcomes (−0.68 and −0.15 points, respectively) requires 94 and 82 clusters. Our actual sample includes 47 clusters in Wave 1 and 50 clusters in Wave 3, totaling 97 clusters across both groups. This exceeds the minimum cluster requirements for all primary outcomes, suggesting that the study is sufficiently powered to detect policy-relevant effects in most domains of interest. Thus, effectively, the sample sizes reported in Table 1 suggest that our sample enables us to detect even slightly smaller impacts.