An Anticipatory Cash Transfer ‘Superfund’ Should Be a Key Part of the Global Humanitarian System

On the CGD blog today, Ashley Pople and Stefan Dercon summarise the enormous benefits of a cash transfer paid to vulnerable Bangladeshis in anticipation of a flood that was predicted to occur based on modelling. Households that received the transfer were far less likely to go a day without eating, reported higher child food consumption and wellbeing after the floods hit, lost fewer assets, and had better earnings potential post-flooding. And these benefits accrued well before conventional humanitarian aid would have been delivered. These results aren’t outliers. In Northern Nigeria, anticipatory transfers supported better adaptation and resilience investments. Famine early warning systems have long been used to fundraise for remedial work, and in Ethiopia and Kenya, existing cash transfer programmes were used to avert its worst effects on the basis of expected famine conditions.

In a world where climate change causes more frequent and more extreme natural disasters, anticipatory cash can and should be an important part of the global humanitarian system. Northern India has just been hit by a ferocious heatwave: a cash transfer that reached the poor before it hit would have meant that many who are reliant on manual labour would not have to risk their health or lives working through the heat. Yet, though UNOCHA, WFP, the Red Cross and others already operate anticipatory programmes, the  collective scale of these programmes—whether measured in people covered (around 13 million) or financing (just $200 million or so)—remains far too small; and each one supporting a specific country-risk pair.

There is a solution, though—one that would help the donor community meet four pressing challenges. A single, much larger anticipatory cash transfers scheme, covering multiple geographies and multiple kinds of forecastable disasters, and implemented through existing, nationally-administered cash transfer schemes, would be a dramatic improvement over the disparate and atomized schemes, run through a plethora of implementing partners and individually funded, that are currently used. Such a scheme would have four big benefits:

• It would be faster and more effective than a system that relies primarily on ex-post fundraising and transfers. By mainstreaming the pre-agreed model, it reduces the ‘begging bowl’ of humanitarian funding, and we know from multiple studies that, where feasible, anticipatory funding has benefits over ex-post funding. It would not replace all ex-post funding, since not all risks or manifestations of risk can be anticipated, but it would be an improvement over the existing balance.

• It mitigates funding issues and simplifies governance arrangements for anticipatory cash transfers, because by aggregating risk of payout over multiple geographies and different categories of disaster it both makes it less likely that the money will sit unused (with high opportunity cost) for a long period, and means that the overall level of funding needed to cover multiple risks and geographies is lower.

• It would be an unusually effective use of climate adaptation finance—and with an increasing share of donor budgets being set aside for climate financing, using it effectively for adaptation and poverty-reducing activities is a priority.

• Implemented through existing national transfer and social protection systems, it would be a quick and impactful method of localizing development aid, both in decision-making (who gets grants) and in administration (national, not donor-implemented, systems).

This kind of scheme is already feasible. In what follows, I set out how it could work. Making it happen would require bringing together and applying a simplifying rule to existing prediction and forecasting technologies; working with and extending national cash transfer schemes; and solving some administrative and financing hurdles that are already within the control of funding agencies.

Creating a single ‘risk-to-livelihoods’ threshold

Anticipatory finance requires defining a trigger above which we expect a disaster to have unacceptable human costs. But because different kinds of natural disaster are more or less amenable to accurate forecasting (earthquakes, for example, are more difficult to predict than drought or flooding; and indeed, our predictive models do not perform equally well in all places) and because different developing countries typically face different kinds of shock, this is more complex than simply monitoring, for example, growth rates or unemployment rates. Indeed, even the level at which support should be triggered for a single class of disaster will likely vary by country. But is a surmountable problem; even if it only covers a subset of disasters in a subset of regions and needs to be supplemented by post-hoc support in some cases, an automatic anticipatory mechanism would still be a substantial improvement and save lives. The approach should be to group together a few simple triggers, and communicate by using a common language across disasters: a risk-to-livelihoods threshold. For example:

• Flood support could be triggered when forecasts of river levels breach certain, model-defined points (as has already been done in Bangladesh), or when rainfall is forecast to breach specific pre-defined levels (though flash flooding from rainfall is much harder to predict, and so will be more likely to require post-hoc funding as well)

• Early warning systems for famine already exist, and have an extremely high ‘hit rate’

• Hurricane seasonal probabilities could be used to trigger payments to support pre-emptive adaption action; specific path modelling could trigger payouts to more narrowly defined geographic areas for near-term adaptation (including movement, food purchases, and the like), though improvements to forecasts would substantially improve targeting and increase affordability

• Heatwaves can be predicted up to a week to ten days in advance, though model accuracy is currently lower in the tropics; transfers made even a couple of days in advance, though, might have substantial benefits if they reduce the necessity to engage in physical labour through a heatwave

Some of the technical details, including specific levels to trigger action remain to be determined—a task beyond this blog, but eminently achievable. Many of the remaining challenges can be solved at an accelerated pace if we use and learn about existing anticipatory methods and approaches more. Specifically, forecasting capabilities are not equally strong everywhere, and investments in improving them in some places will be necessary: an audit of needs and plan to meet them may be a necessary first step, and would be a starting point for setting up a single anticipatory fund as proposed here anyway. This would also define a starting point and plan for scale up for the scheme.

Extending and using national cash transfer schemes

Cash transfers are now a well-established and logistically simple approach to providing emergency support in developing countries, though in the poorest countries especially there is untapped potential to increase coverage and use faster, more efficient and more flexible delivery mechanisms (notably mobile money). But poor countries can’t implement a system like this on their own. Fiscal space is a problem, with cash transfer programmes during Covid being smaller in coverage and often less generous in poorer places (see, for example, figure 11 here). Though donors can provide additional funding to existing schemes if moved to do so, too often funding gaps arise, and cost lives. A global or much extended scheme that operates only through national systems would provide both the finance to extend schemes and an incentive for national authorities to extend and improve their own systems to qualify for it.

This would be a dramatic improvement on existing systems, which typically use donor-designed systems, in parallel to national social protection systems. And we’ve never been in a better position to do it. Thanks to the work of Ugo Gentilini and colleagues at the World Bank, we know more about the coverage, functioning and capacity of social protection and cash transfer systems around the world than ever before; most of these systems either already have or can acquire information on additional potential recipients beyond those receiving the ‘core’ support. The model of providing surge payments to go to people in areas at risk of famine or other disasters through existing cash transfer delivery systems has already been successfully implemented. Though the existing systems are not perfect, the experience of Covid (when many countries constructed pre-identified lists of ‘vulnerable households’) has shown we can nevertheless use them to get support out, and it’s only by use that they will be improved.

This isn’t, however, to dismiss real difficulties that require resolution. We should be particularly concerned at ‘false exclusion errors’, where ex-ante assessments of the right beneficiaries exclude people who need support when the disaster strikes. Exclusion errors are not trivial to solve (and exist with almost all forms of support, anticipatory or otherwise); careful monitoring and learning is required to minimize them over time (and there is evidence that non-traditional mechanisms can reduce, but not eliminate, them).

Overcoming donor funding and administrative constraints

The final, related, problem is how to fund the whole system. At present, the largest anticipatory finance schemes are run through the Central Emergency Relief Fund (CERF), but as noted above, few run through domestic social protection schemes, and each scheme is separate. This requires a number of earmarked funds, that can only disburse under closely specified conditions. But donors balk at setting aside money for spending that may never be triggered, as I set out here.

The single super-fund I propose here, financed by multiple donors, would have a number of advantages over the existing set up. Firstly, by diversifying geographic and disaster coverage it increases efficiency by pooling payout risks: rather than fully financing 20 separate schemes, with each requiring accessible financing in the event of triggers being met, a pooled scheme can be financed to a lower level (on average) on a year-by-year basis, given that in any given year, most triggers in most countries will not be met. This achieves wider coverage for a smaller overall financial outlay. By running all of these schemes through national social protection schemes, there is also a clear mechanism through which unspent balances can be used for a developmentally-useful purpose, so unused balances are never ‘wasted’. Alternatives are also possible: for example, using insurance (or funder/group of funders of last resort) to back up a small ‘standing’ pot maintained and topped up by donors as needed when trigger thresholds are crossed; insurance can cover the overall fund, not each disaster—it would mean a more difficult initial negotiation, but much smoother payouts once in operation. Yet another alternative is to reform humanitarian funding in toto: switching to assessed contributions from UN member states rather than voluntary contributions, as proposed by Mark Lowcock, and funding anticipatory response from the ensuing budget.

Humanitarian funding and climate adaptation finance need improvement. While cash cannot do everything—and we should continue to fund public goods and collective adaptations—we should do all we can to make an anticipatory superfund happen. We live in a world where extreme weather events are becoming routine. Let’s make minimizing their human cost routine as well.

This piece benefited from excellent comments from Siddh Haria, Charles Kenny, Mark Lowcock, Ashley Pople, Giulio Schinaia, Hannah Timmis, and Rocco Zizzamia. Errors, omissions, and the like are my own.