Finding Fungibility: Loans, Grants, and the Fiscal Response to Aid

In 2000, donors around the world withheld over $300 million of development aid that had previously been promised to Kenya. Faced with a drastic budget shortfall, the Kenyan government hiked up taxes and made severe spending cuts. These cuts hit even strategic sectors like the army, the police, and the judiciary. Kenya is far from the only country to raise taxes and cut spending across the board in response to an aid freeze. However, the aid withheld from Kenya was never meant to be spent across the board; it was, after all, development aid. It was meant for infrastructure projects, for education programs, for health initiatives. Yet somehow, the aid freeze still resulted in cuts to essential services in all sectors. A drastic drop in the amount of aid received impacted Kenya’s entire budget.

This case and many others point to the fact that development aid is like any other source of government income: it makes governments rethink how they collect and spend money. This occurs despite aid donors’ wishes. In a donor's ideal world, each recipient government would collect and spend money in exactly the same way with or without foreign aid. Aid would enable the government to provide extra public services in addition to what it already provides. Essentially, aid would be the icing atop the cake of the government's budget. But it seems as though this ideal case is far from reality.

Donors worry specifically about the possibility that aid is fungible. Aid fungibility occurs when recipient governments can treat aid as a substitute for their other sources of income. When aid is fungible, recipient governments can spend aid money as they please regardless of what the donor wants. As a former World Bank economist Paul Rosenstein-Rodan put it, “When the World Bank thinks it is financing an electric power station, it is really financing a brothel.”

There are two important types of aid fungibility. The first occurs when aid recipients collect fewer tax revenues in response to more aid. If recipient governments can provide public services using aid money instead of tax revenues, then there’s no incentive to collect taxes – after all, taxes aren’t exactly popular. This situation can be described as fungibility in revenues. Alternatively, recipients may move money from pro-development sectors to other sectors in response to aid. If they can use aid money to cover their costs in sectors like education and health, then they can spend other revenues on things like defense or, as Rosenstein-Rodan would have it, brothels. Education or health spending might not end up increasing at all. This situation can be described as fungibility in expenditures. Between revenues and expenditures, a government might end up adjusting the entirety of its budget in response to aid if that aid flow is fungible.

If aid fungibility is so important, then we should care about what causes it. But despite extensive efforts to measure aid fungibility and its effects (see Morrissey’s recent review), there has been little research into its determinants. So what causes aid fungibility? One possibility is that the form of an aid flow – that is, the decision to send aid as a loan or as a grant – determines the fungibility of aid. Specifically, loans should be less fungible than grants.

There is fairly strong empirical evidence to support this hypothesis. If loans are less fungible than grants, then an increase in aid loans (as compared to an increase in aid grants) should be accompanied by less of a decrease in tax revenues and more of an increase in development expenditures. Regressing tax revenues on aid loans and aid grants tests this prediction for fungibility in revenues, and regressing development spending on aid loans and aid grants tests it for fungibility in expenditures. Using this model with data from AidData and the IMF, the coefficient of loans is significantly higher than the coefficient of grants across both revenues and expenditures, which is exactly what we’d expect to see if loans were indeed less fungible than grants. This finding holds across multiple sets of controls, and it stands up to several robustness checks.

So why exactly do loans seem to be less fungible than grants? It’s difficult to demonstrate a causal relationship, but there’s some evidence that two key factors are involved. First, loans have to be repaid. When aid recipients get a loan, they know that they will owe a large sum of money in the future. There is thus no reason for recipient governments to collect fewer tax revenues in the present – if they stop collecting money this year, they’ll have to collect more money in the future. That makes donors less likely to use loans as a substitute for tax revenues. Second, there are usually more conditions attached to loans. Aid often comes along with conditions that give donors influence over the recipient’s governmental budget, and these conditions can make it very hard for aid recipients spend money in ways that the donor does not prefer. Loans have more of these conditions than grants do, so it’s harder for recipients to reallocate aid against the donors’ wishes when it’s given as a loan. In short, because loans (1) must be repaid and (2) come with more conditions, they are less fungible than grants when examining both revenues and expenditures.

The insight that loans are less fungible than grants is a useful first step towards understanding the roots of aid fungibility. However, there is still much work to be done. As a World Bank publication puts it, “aid's true effect depends on the crucial (but difficult to assess) question: what would have happened in the absence of donor financing?” We have to understand the fiscal response to aid if we want to make aid more effective – and understanding aid fungibility is a crucial part of that process.