Effects of anticipatory humanitarian cash assistance to households forecasted to experience extreme flooding: evidence from Bangladesh

The concept of anticipatory humanitarian action has gained broad-based acceptance in recent years. In 2022, 35 countries had anticipatory action frameworks covering 7.6 million people (Weingärtner et al. 2020; Pople et al. 2021; Anticipation Hub 2023). Instead of waiting for a disaster to occur and providing response assistance to the affected population, governments and humanitarian organisations now increasingly use weather forecast and risk data to act before the shock materialises. When a pre-defined, impact-based trigger threshold is reached, forecast-based financing (FbF) releases pre-arranged funding to assist the at-risk population before an extreme weather event, like flooding or drought, to help the most vulnerable avoid or mitigate the impacts of the hazard. Building on an extensive body of Early Warning Early Action work within the humanitarian sector, the anticipatory action approach was piloted with limited funding in a handful of countries starting in 2015 (Coughlan de Perez et al. 2015) and has since seen significant scale-up.

Today, around 60 countries implement anticipatory humanitarian action (Anticipation Hub 2022). Major humanitarian actors are mainstreaming anticipation: the International Federation of Red Cross and Red Crescent Societies (IFRC) has established Forecast-based Action by the Disaster Response Emergency Fund (FbA by the DREF), a funding mechanism dedicated to help National Societies take early action before disasters strike. The United Nations Office for the Coordination of Humanitarian Affairs (OCHA) has committed US$ 140 million from the Central Emergency Response Fund (CERF) to support anticipatory action in a diverse range of contexts and for different types of shocks.

Despite the notable proliferation of anticipatory humanitarian action across countries and actors, the number of impact evaluations to assess whether the approach is effective in helping disaster-affected households avoid or mitigate negative impacts is growing but remains limited overall (Weingärtner et al. 2020).

This study seeks to contribute to the evidence base on anticipatory action by examining the effects of a forecast-based unconditional cash transfer provided to households about to experience severe flooding. The Bangladesh Red Crescent Society (BDRCS), in collaboration with other partners,¹ activated their early action protocol (EAP) on 25 June 2020, ahead of the 2020 monsoon floods that were among the most extreme and protracted in decades. BDRCS reached 3,789 households across three districts along the Jamuna River with a multi-purpose cash grant of BDT 4,500 (about US$ 53) between 48 and 24 h before the first flood peak which occurred between 30 June and 3 July 2020, and 2 weeks before a second flood peak started on 14 July 2020. The cash transfer value is equivalent to approximately 2 weeks' food expenditure (Hill & Genoni 2019). To assess whether the anticipatory cash transfer was effective in helping households to evacuate in a timely manner, protect their health and assets, avoid negative coping strategies, and recover more quickly from the shock, we follow a quasi-experimental approach and compare households that received the forecast-based transfer to otherwise comparable households that did not.

A large body of literature exists on recurring cash transfers to assist vulnerable populations in resource-poor settings (Bastagli et al. 2016). The evidence suggests that regular monetary support can effectively buffer the adverse income effects of exogenous shocks and reduce negative coping behaviours. However, there is very limited rigorous evidence on the effectiveness of one-off cash transfers, despite humanitarian agencies making increasing use of them for disaster relief since the early 2000s (Bailey & Harvey 2015). A systematic review specifically of the evidence for anticipatory action (Weingärtner et al. 2020) found little rigorous evidence of their effectiveness in improving outcomes for the populations covered, with several notable exceptions and recent additions: an early assessment of anticipatory cash transfers in Bangladesh, provided by BDRCS ahead of severe flooding in 2017, showed that the cash grants contributed to improving households' access to food, a reduction in high-interest debt accrual of vulnerable households, and reduced psychosocial stress during and after the flood period, compared to a control group of similarly vulnerable and flood-affected communities that did not receive the forecast-based cash assistance (Gros et al. 2019). A similar study of an anticipatory cash transfer by WFP in 2020, drawing on a larger sample size, found comparable results, indicating that households who received money before the flood peak were less likely to go a day without eating during the flood, reported significantly higher child and adult food consumption and well-being 3 months after the flood, and experienced lower asset loss, engaged in less costly borrowing after the flood, and reported higher earning potential (Pople et al. 2021).

Our paper seeks to answer two overarching research questions:

• 1.

  Did the anticipatory humanitarian cash transfer help to reduce the direct impact of the flood on beneficiary households vis-a-vis the counterfactual of comparable, disaster-affected households that did not receive the intervention, considering their health and well-being, and the loss of personal and productive assets such as agricultural tools and livestock?

• 2.

  Did the intervention improve beneficiaries' conditions for recovery in the aftermath of the flood event, particularly regarding avoiding negative coping strategies, their financial situation and earning potential?

Bangladesh is among the countries most vulnerable to natural disasters and climate change-related extreme weather in the world. Flood events occur annually in the monsoon season (June to September). The areas most exposed to flood hazards are the Jamuna, Padma, and Meghna River systems, and especially poor and vulnerable households living on char lands² and along the riverbanks.

The 2020 monsoon floods in Bangladesh were the second highest in the past 35 years. More than one million households were inundated and over 5 million people were directly affected by flooding. Flood waters disrupted agricultural production, food markets, schools, health services, and damaged infrastructure. The Ministry of Agriculture estimates that 110,000 hectares of cropland were damaged, while 257 people lost their lives due to the floods (United Nations Resident Coordinator and Office 2020).

At the household level, the eligibility to be included as beneficiaries in the anticipatory cash distribution was assessed based on a vulnerability score calculated for each household using five criteria: the quality of the housing structure; the level of inundation during previous floods; the number of household members who are children, elderly, or disabled; family structure (being female-headed, widowed, or divorced); and the primary livelihood strategy. Based on the vulnerability score and available resources, BDRCS transferred 4,500 BDT (about $53) to 3,789 households in the ten unions across three districts along the Jamuna River.

This study is based on a random sample household survey among anticipatory cash transfer recipients and comparably vulnerable, flood-affected households not reached by the forecast-based assistance because of resource limitations. Comparison group households were identified using the same targeting criteria that were applied for identifying cash transfer beneficiaries. To ensure comparability of all households in the survey, only those within 1.5 standard deviations (SD) of the average vulnerability score across all screened households were considered for inclusion in the sample.³

BDRCS volunteers collected the data in January 2021⁴ in the area affected by the flood. In addition to questions assessing the flood impact and potential effects of the anticipatory action cash transfer, the survey gathered demographic, socio-economic, and geographical information from each respondent household to enable propensity score matching (PSM) for a balanced sample. The sampling was stratified by districts and unions where the anticipatory actions were implemented. The sample was distributed proportional to the size of the beneficiary group in each geographic area. The final sample comprised 444 observations, with 222 anticipatory cash transfer beneficiaries and 222 comparison households. Supplementary Table A1.1 in Appendix 1 describes the geographical distribution of the two groups by unions.

The main variables for the analysis were divided into confounding characteristics used for the matching method and outcome variables used for measuring the effect of the anticipatory actions on beneficiaries. The final dataset contains 444 observations on 187 variables. Supplementary Table A1.2 in Appendix 1 presents the main socio-economic, demographic, and geographical characteristics of the sample, while Supplementary Table A1.3 in Appendix 1 shows the main outcome variables used in the analysis to assess the effectiveness of the interventions.

The study follows a quasi-experimental design, a widely used approach in the literature for the evaluation of development and humanitarian interventions and unconditional cash transfers to cope with climate hazards (Bastagli et al. 2019; Gros et al. 2019, 2022). Quasi-experimental methods simulate the randomisation of the intervention assignment and apply balancing methods between the two groups under study – the programme beneficiaries and the comparison group representing the counterfactual. With this empirical framework, any differences observed in the outcome variables can be attributed to the intervention. To this end, a matching method is used to associate observations in the intervention group (i.e., the households that received the anticipatory cash transfer) with similar observations in the comparison group, factoring in potentially confounding variables to reduce any bias due to structural differences between the two groups. Only households that are comparable in terms of demographic, socio-economic, and geographical characteristics are retained in the sample.

We employ a PSM approach, using the two nearest neighbours algorithm (2NN) for comparing the two groups (Angrist & Pischke 2009; Cerulli 2015). The use of PSM is a well-established, non-parametric technique often adopted in policy evaluations and econometric analysis (Card & Krueger 1993; Angrist & Pischke 2009; Imbens & Wooldridge 2009). It does not rely on any prior distributional assumptions and allows for the identification and correction of the selection process and imbalances in the data, thereby ascertaining the comparability between the two groups (Baser 2006; Cerulli 2015). We use PSM to select subsamples of beneficiaries and non-beneficiaries for comparison according to their background characteristics (e.g., socio-economic, geographical, or demographic variables).⁵ We then calculate the area of common support and the main differences in each variable of interest. In this way, the final difference between the two groups for each target variable can be considered as the average treatment effect on the treated (ATT). In this study, the ATT represents the effect of the anticipatory cash transfer on the beneficiaries.

The main findings of the study are reported in Table 1 which shows only statistically significant results related to the key outcome variables of interest. The data indicate that anticipatory financial assistance played an important role in helping households mitigate the flood impacts and recovering in the aftermath of the event. We discuss each finding in detail in the following sections.

We first look at how the anticipatory cash transfer was used by the beneficiary households (see Supplementary Table A1.4 and Figure A1.1 in Appendix 1). Cash recipients spent the money principally on food: 91% of the respondents reported to have spent at least some of the funds received on foodstuff. Other relevant expenditure categories are purchasing of new livestock (34% of respondents), evacuation (24%), health expenses (29%), and reinforcing the housing structure in preparation for the flood (19%).⁶ These cash grant expenditure patterns are similar to the findings of previous anticipatory cash transfer assessments in Bangladesh (Gros et al. 2019; Pople et al. 2021).

The anticipatory action intervention was effective in helping households protect their livestock, with cash that can be used to construct floats or hire boats to take the animals out of the flooded area. A total of 42% of surveyed households reported to have at least one cow or calf before the floods. It is important to note that ownership of cows or calves among the highly vulnerable, poor, and landless households that were targeted by the anticipatory cash transfer is unlikely. Instead, most who reported taking care of this very valuable type of animal will have done so on behalf of more affluent families. Nevertheless, losing someone else's expensive cow in the flood can have significant negative repercussions for the household that is deemed responsible for the loss. Anticipatory action beneficiaries reported to have lost only 9% of their cows and calves (calculated as the number of dead animals over the total number of animals owned before the flood event occurred), while the households in the comparison group reported an animal mortality rate of 22%, a statistically significant difference at the 95% level (Figure 5).

Food-based coping strategies, such as reducing the number, frequency, or nutritional content of meals, did not show any statistically significant differences between beneficiary and non-beneficiary households. The average reduced coping strategy index (rCSI) (Maxwell & Caldwell 2008) score was 26.4 among beneficiaries and 25.4 among comparison households, on an index scale ranging from 0 to 56 (Supplementary Figure A.1.2 in Appendix 1). This is even though 91% of cash beneficiaries spent at least some of the transfer on food. While this raises questions about the effectiveness of anticipatory cash to avoid food-based coping strategies, the rCSI is a relatively crude measure that does not capture dietary quality. This could have been measured with a more granular metric, such as the food consumption score (FCS), which we propose to use in future research.

Our paper presents the results of a quasi-experimental evaluation of the effectiveness of an unconditional cash transfer to vulnerable households that were about to experience severe flooding in Bangladesh. We found robust statistical evidence to conclude that the anticipatory humanitarian intervention was effective in achieving its main objectives of helping households evacuate the flood-affected area effectively, reducing the flood impacts on personal health and well-being, and protecting their productive assets and livestock. The forecast-based cash transfer before the flood event was also effective in enabling beneficiaries to avoid negative coping strategies, such as taking on high-interest loans and selling valuable assets.

We did not find evidence of the effectiveness of the intervention in helping cash recipients avoid food-based coping mechanisms. Reducing the number or nutritional content of meals can have adverse effects on the health of household members, particularly women and children. We also did not detect significant effects of the intervention on the earning potential of beneficiary households after the flood. This is consistent with the earlier evaluation from Bangladesh (Gros et al. 2019) whereby intervention and comparison households were equally unable to go back to work for an extended period after the flood shock. A more recent evaluation of an anticipatory cash transfer of equal size in Bangladesh, albeit for a larger group of beneficiaries and drawing on a larger sample size, found that the intervention had a positive effect on the number of paid hours of work per adult over the previous week (Pople et al. 2021). While our metric – the number of days having been unable to go back to work after the flood – captures the immediate impact of the flood on the household's productive capacity in the aftermath of the flood, it is more likely to be subject to recall bias. Asking the respondent to remember the number of paid hours worked during the previous week is more likely to receive an accurate response, although this is a somewhat different measurement concept.

This analysis has confirmed the positive effects of providing unconditional cash assistance to vulnerable households in anticipation of severe weather events. This is particularly relevant given that cash transfers have been widely adopted as a fungible assistance modality in anticipatory action interventions (UNOCHA 2021; Asia-Pacific TWG on AA 2022; Anticipation Hub 2023). Further research is required to better understand the effect of forecast-based cash transfers on households' productive and recovery capacity in the aftermath of the flood.

C.G. designed the quasi-experimental study, developed the sampling frame, and guided data collection and analysis; he also led the write-up of findings. A.P. conducted the quantitative data analysis, prepared the results, tables, and charts, and contributed to the write-up. K.S., A.H., and M.S. led the study team in Bangladesh, oversaw the data collection and advised on the analysis.

Data cannot be made publicly available; readers should contact the corresponding author for details.

The authors declare there is no conflict.