https://orcid.org/0000-0002-4697-640X
![Sample our Economics, Finance,Business & Industry journals, sign in here to start your access, latest two full volumes FREE to you for 14 days]({"type":"image" , "src" : "/sda/5931/TOC-Subject-Sample-2021-Economics-Finance-Business-Industry.jpg"})
Abstract
Large, unexpected shocks are becoming more frequent, making the design of robust social transfer programs more vital than ever. We evaluate the performance of the Food Friendly Program (FFP), the largest in-kind social transfer program in Bangladesh, before and during the nation-wide COVID-19 lockdown. Using two-rounds of nationally representative household surveys combined with administrative data, we document that high leakages and large welfare losses are related to corruption. This contrasts with the performance of the pre-lockdown FFP, when leakage was low and coverage high. We then compare the performance of the FFP with two initiatives launched following the pandemic: an in-kind and cash transfer program, respectively. These programs have markedly higher levels of leakage than the FFP. Our findings are relevant to other large shocks, such as those caused by climate change, and have important policy implications for the design and delivery of transfer programs in developing countries characterized by institutional weaknesses.
Acknowledgements
We acknowledge helpful comments by Mushfiq Mobarak, Aaron Nicholas, and Timothy Richards. We are thankful to the United States Agency for International Development (USAID) for supporting the phone survey; Pathways Consulting Services Limited (PCSL) for administering the survey in 2020; and officials at the Directorate General of Food (DG Food) Bangladesh for administrative information on the Food Friendly Program (FFP). The usual disclaimer applies. IRB approval (IFPRI-IRB) was granted under study number: 00007490. The authors have no competing interests to declare. Data for replication of these results are available upon request.
Disclosure statement
No potential conflict of interest was reported by the author(s).
Notes
1 Although poverty is more prevalent in rural areas and in the agricultural sector (Castañeda et al., Citation2018), self-employed women and workers in the non-agricultural sector experienced some of the largest drops in employment during the early months of the pandemic in 2020 (Bundervoet, Dávalos, & Garcia, Citation2022).
2 See Bowen et al. (Citation2020) for a lucid discussion on adaptive social response.
3 Alfaro, Becerra, and Eslava (Citation2020) also show that in economies with larger informal sectors, a greater share of the workforce faced unemployment risk during lockdowns.
4 In Bangladesh, the informal sector employs 85.1 percent of the labor force and the share of self-employment in total employment is 44.3 percent (Table S34 and Chart 6.6, respectively from Bangladesh Bureau of Statistics (BBS) (Citation2018).
5 In a qualitative study conducted by Dutta (Citation2020), the author documents instances of exclusion for qualified beneficiaries to the PDS, which is an important aspect not covered in our study.
6 At the risk of over-simplification, this literature can be broadly divided into the following categories: (i) the mortality rate of COVID-19 and how it can vary depending on age, income, and race in the case of rich countries; (ii) the effectiveness of public health measures, including restrictions such as social distancing measures, lockdowns, border restrictions and closures, masking and mass vaccination; (iii) the impacts on income and employment, including heterogeneous effects along the lines of gender, occupation, and socio-economic conditions; (iv) the effects on education and learning due to school closures, the mental health of children and adults, and the various ramifications on families including differential impacts on women and other disadvantaged groups; and (v) public policy responses that are not directly related to health behaviour and outcomes.
7 These are the two pre-harvest lean seasons in rural Bangladesh when the price of rice increases, and job opportunities and agricultural wages decrease. This leaves poor individuals, especially agricultural day laborers, vulnerable to seasonal hunger and occasional famine. See Bryan, Chowdhury, and Mobarak (Citation2014) for further details.
8 The Union Parishad is the lowest local government unit in Bangladesh. Multiple unions form an upazila (UPZ); multiple UPZs form a district; and multiple districts form a division. In total, there are eight divisions, 64 districts and 492 UPZs in Bangladesh.
9 Data were collected through a short phone survey in which only some key binary questions were asked, such as whether a beneficiary received rice or not, and whether a beneficiary was asked to pay bribe. This type of information is not subject to the criticism raised by Abate, de Brauw, Hirvonen, and Wolle (Citation2023) that consumption data collected by phone surveys are not reliable.
10 All our regression results are robust both in terms of magnitude and statistical significance if we estimate an unbalanced panel consisting of 2,797 beneficiaries. However, using the unbalanced panel requires us to exclude dealers’ characteristics from the regressions. Since transfers to beneficiaries were made by dealers, we consider those characteristics as important controls in the regressions, and we therefore work with the balanced panel.
11 In fact, 2,797 households were reinterviewed. The working sample reduced to 2,602 because dealers’ information was not available for the remaining households.
12 Nonetheless, we cautiously interpret our results as causal effect even after controlling for a rich set of variables that include household and dealer characteristics and geographic variations.
13 We do not have information about any collaboration of the dealers with individuals in the upstream administrative or political layers.
14 We do so as UPZs are the primary decision-making unit in the current context (See Abadie, Athey, Imbens, & Wooldridge, Citation2022, for an in-depth discussion on clustering).
15 No dealer characteristics (not reported) are found to be significant in any of the specifications, which corroborates the pervasiveness of dealers demanding bribes from beneficiaries.
16 As mentioned earlier, the difference between the amount received and the allocated 30 kg is very small. Therefore, the main source of leakage is at the extensive margin (due to a beneficiary not receiving rice at all). If we assign 30 kg to amount received, then the leakage would be either 0 (for those who received rice) or 1 (for those who did not receive any rice), which would be the reciprocal of the dependent variable in our previous regression analyses. We use the actual amount of rice received by a beneficiary.
17 Since there is no restriction on what beneficiaries can do with the rice received, they can immediately sell the rice in the open market at a higher price. If they choose to do this, then their transportation costs will be lower since they no longer need to carry the rice back to their residence.
18 The total minimum market value of a 30-kg bag of rice is BDT 1,350 (45*30). A beneficiary pays only BDT 300 (30*10)) and incurs BDT 60 for transportation. Her benefit for receiving a bag of rice (or welfare loss if she does not receive the rice and the dealer sells it in the black market) is BDT 993 (1,350 – 300 – 60). Therefore, the welfare loss per MT of rice is BDT 33,000 (990*1000/30).
19 Note that we carefully checked the eligibility criteria of these two programs and verified with the government officials that the exclusion criteria of these two programs do not include beneficiary from other existing programs including the FFP.
20 Similar findings are reported in Nairobi, Kenya, where only a small fraction of vulnerable families benefited from cash transfer programs implemented during the COVID-19 because of cronyism, nepotism and outright favouritism (Miguel & Mobarak, Citation2022).
21 See the media reports published in The Daily Star on April 12, The Business Standard on April 15, and The Washington Post on April 26, 2020.
1 This number is calculated using a margin of error of 5 percent or less and 18 percent leakage; the latter is based on Olken (Citation2006) in the absence of any prior knowledge of leakage in the FFP or any such program in Bangladesh.
2 Data could not be collected from one union because no selected beneficiaries could be traced due to frequent migrations in the area.
3 For each household the PMT score is constructed by adding the constant with each of the covariate times the associated weight. In other words, , where “i” denotes household and “k” denotes covariate included in the regression. For further details, see Chowdhury et al. Citation2020.
4 This is perhaps because their phone numbers had changed. Frequent switching to a different mobile phone operator is a regular phenomenon in Bangladesh, depending on the strength of the network connection (especially in rural areas) and deals offered by competing operators. In Bangladesh, such switching requires a new SIM card with a new phone number. Ahmed, Islam, Pakrashi, Rahman, & Siddique (Citation2021) in their phone survey from a pre-existing sample in Bangladesh, also had more than 20% attrition for similar reasons as ours during the COVID.