Since March, we have been asked a lot: how do we get emergency assistance to people worst affected by the pandemic? I am not going to bore you with the standard answers the experts have been giving. For our readership, they are well known by now: make the existing programs more generous; expand them using existing databases or allowing new applications; use all-digital/mobile platforms; don’t forget supporting small businesses; etc. For an excellent place for resources, please see my colleague Ugo Gentilini’s page.
But, having spent an earlier part of my career studying the targeting of regular safety net or anti-poverty programs, one idea kept gnawing at me, although I was hesitant to propose it. What if you got money to communities fast, hopefully using existing structures, and then let the communities take care of the targeting themselves? This would allow geographic targeting that can be fairly progressive, and then rely on idiosyncratic information at the local level to target households as needed. It has the advantage of being potentially much faster than having new households apply, be vetted, and onboarded to your system and, in many cases, would get the money to the ground way faster. In very poor areas, you could skip within-community targeting altogether and provide universal assistance. The downside is that within-village inequalities in power and corruption might make this second phase of targeting within communities regressive, with the resulting overall targeting effectiveness uncertain.
I’ll get to the downsides further below, but the possibility that this would not be feasible or attractive to governments in many places prevented me from suggesting this as a legitimate strategy to any policymaker. Governments are risk averse and like to avoid scandals: nothing worse than stories of some village elites skimming off the cash transfers meant for the poor and vulnerable in their communities. Especially, when tensions are high and many of us are on edge to begin with. So, I shelved it…
Well, more than six months into the pandemic, comes a story from Indonesia that is doing this. With apologies to our readership here, the story comes not from a new academic paper, but from a newspaper article in Indonesia, told to us by a colleague and friend, Scott Guggenheim. Scott is a social anthropologist, who not only worked at the World Bank for many years, but spent decades in Indonesia working on community-based development, among many other things. So, between his communication and Google Translate, I can vouch for the veracity of the account.
Well, what happened in Indonesia then? As we alluded to above, there is a tradeoff between speed of payments and targeting. This sometimes requires innovative approaches to targeting that are nimbler and might be feasible under the right conditions. Indonesia has both a social safety net program providing conditional cash transfers for poor households, but also a community-based development (CBD) program – primarily used for local public goods creation. While the CCT program was running smoothly prior to the pandemic, trying to expand the rolls turned out to pose a real challenge to the government: the database was outdated and had errors and updating it was going to take some time – time the government did not have during this emergency. So, what did they do? They expanded their support of the community-development fund and allowed 35% of the funds to go to locally-identified cash transfers to households. Simplifying the rules for the allocation of CBD funds towards emergency cash transfers was possible without having to worry about huge targeting errors, because the Indonesian government had earlier run an RCT comparing the effectiveness of using proxy means-tested targeting vs. community-based targeting in reducing poverty, and concluded that the latter was almost as effective as the former, with the latter also having the benefit of having higher community satisfaction (Alatas et al. AER 2012). The result is fast disbursement of cash into communities (using geographic targeting) and combining it with targeting at the local level to reach poor and vulnerable people currently not reached by the existing safety net programs. Nice example of cross-sectoral collaboration (social protection and social development) to boot…
But, hang on. Could other countries do the same and obtain the same results? My view is that some probably can and many not. See, there are decades of experience with community-driven development in Indonesia, with huge amounts of social capital and institutional infrastructure to make such a scheme work. Even in countries, where you had your poverty maps and an easy means of sending money to the bank accounts of villages for public goods projects (say, through previous Social Action Funds types of programs), could you rely on them distribute emergency money fairly and efficiently across households? Are the systems, the checks and balances, in place to make sure that massive corruption and stealing does not take place? To be clear, this is not a developing country problem – this tension exists anywhere.
I don’t have the answer; every country would have to consider this problem for themselves. But, I have a couple of papers that hint at some considerations. First, you’d have to worry about power structures in the communities: in a paper in the Journal of Public Economics (2008), we showed that villages were less likely to select social action funds projects that provided private/excludable goods for the poor (such as latrines) in Ecuador, as the share of income that accrued to the top X% of the income distribution increased. So, elite capture is something to consider: who will make the decisions? How will deliberations take place? How gets to be at the table?
Second, if there are any demand-driven aspects of being eligible for assistance, the most vulnerable will often lose out. In another paper in the Journal of Public Economics (2013), we showed that while the centrally dictated features of Tanzania’s Social Action Funds program (TASAF), namely the predetermined funding allocations to areas based on poverty maps, along with eligibility criteria, were pro-poor, because people within the selected areas had to come forward to be eligible, the within-village processes were regressive: wealth, education, access to media and phones, and political connectedness were all positively correlated with knowledge of the program. Even living near the village center was a big predictor of beneficiary status, with people living in the outer, remote hamlets more likely to miss out.
To be sure, these are not insurmountable obstacles: some governments may simply be willing to have some of these things happen, to be able to get assistance to people as quickly as possible, even if there are more targeting errors than normal times. Combining geographic targeting with categorical targeting within villages (all children and elderly, all disabled, unemployed, etc.) for all people who are not already safety net beneficiaries, could cover most of the needy households. But, one can also see why governments may be reluctant to go down such a route. Even if the fiscal space is there, the infrastructure to transfer the money to communities may not be. Worries about conflict in some communities may be so high, so as to forgo choosing this route.
But, at least, good for Indonesia: not only are they taking great advantage of all the investments they had in their social protection and community development systems to respond effectively during a crisis, but they are also using at-scale RCTs, which they very much helped happen, to inform their response a decade later. I’ll take a feel-good story like that right now and hope to see other innovative approaches to help people to get through this difficult time.
Join the Conversation