Give Aid to Poor People, Not Poor Places?

Prioritising poor countries to receive our foreign aid might prevent us from getting it to poor people. As my colleagues Owen Barder and Matt Juden discussed, the latest data shows that although 2014 was a record year for aid spending, there was a significant fall in the share going to least developed countries, which face the most urgent development challenges.

This trend is worrying. It also highlights deeper issues with the way aid is spent.

Imagine a country with national income of $100 and 10 people. If they earned $10 each and the poverty line were $1, then Success! Poverty free! But if there were a second country in which one person earned $100 and her nine compatriots earned nothing, the country would have a poverty headcount of nine out of ten, or 90 percent. Not so great. Yet because both places have identical incomes per person, neither might qualify as “poor countries”. If, instead, we gave more money to places where they were poor people, then our aid budget would (rightly) go to the country with nine destitute people in it.

Comparing the distribution of aid in 2014 to the distribution of the global poor give us a rough cut at what this would look. This means asserting a poverty line and finding out what share of the population fall below it. Though the World Bank makes this data available, actually using its PovCal database quickly causes one to lose the will to live — it only shows a single data point for a single country at a time. Instead, I’ve used a version of that data which my colleagues Justin Sandefur and Sarah Dkystra painstakingly generated to calculate the number of poor people under three poverty lines: the $1.25 a day standard and two arbitrary ones of $5/day and $10/day. (The international poverty line is now $1.90 a day; I’m using the previous one which applied to prices in 2005, the same units as Justin and Sarah’s data).

This really is a rough cut: we don’t have poverty statistics for all the countries that get aid, nor are all the headcount figures equally current. Because a lot of aid goes to multicountry facilities or to unspecified developing countries, the analysis of what’s left covers $80 billion of the $160 billion in aid spent in 2014.

With those caveats in mind, there are big gaps between where aid gets spent and where the poor live. In the graph above, countries to the right of the diagonal line receive a share of aid that is less than their share of people living in poverty. Asia is the standout case of this mismatch. Surprisingly, sub-Saharan Africa is the opposite case: though people are very poor, this region has a smaller share of the global poor simply because Asian populations are an order of magnitude larger.

The distance from the diagonal line summarises which places would “win” or “lose” if we moved from how aid was spent in 2014 to spending aid based on where people live below the poverty line. At the regional level, allocating aid to tackle per-person poverty means a realignment in development support, mainly away from Africa and towards Asia. (Several years ago, Andy Sumner made a closely related point in a great CGD working paper).

What’s driving the reallocation? The chart below shows the top ten and bottom ten changers — those places where aid would go up the most or down the most if we shifted to an poverty-headcount rule for aid allocation. (I’ve manually excluded Syria from this list; it would appear to lose out, but only because the headcount data precede the civil war). The reason is simple: the combination of vast populations and double-digit poverty headcounts in India and China mean that these countries dominate the global distribution of poverty.

This kind of data-driven counterfactual thinking is interesting, and it helps us put some structure on what a different aid regime might look like. But is it good policy? There are two arguments against using the distribution of poverty to more closely guide the distribution aid.

The first is that we don’t program aid to target the poorest. The World Bank, for example, gives dramatically greater weight to measures of aid recipients’ governance and institutional quality in its lending decisions. (More perniciously, we’ve always given aid for less honourable ends: everything from securing our allies’ goodwill to helping friendly governments win elections to subsidising our own exports).

The second is that we shouldn’t program aid purely to target the poor. As my colleague Owen Barder put it, if we can educate two girls in country X for the price of educating one girl in country Y, then we should spend more in country X. In other words: allocate based on impact. An extension of this is that we shouldn’t spend aid in places where the poor are likely to be lifted out of poverty; since China has a space program, the thinking goes, the Chinese poor are likely to be lifted out of poverty faster than those in, say, the Democratic Republic of Congo. (Adrian Wood makes this point well, showing how accounting for economic growth in aid allocations implies spending less today where people are less likely to be poor tomorrow).

I’m sympathetic to both arguments. But I’m not persuaded. For one thing, resource allocations based on “policy scores” are famously problematic. Even if we could measure what it means to have good policy, allocating aid where it’s most potentially effective leaves the neediest communities behind because they happen to be in places that are badly run and unlikely to improve.

More importantly, thinking about impact actually bolsters the case for taking the distribution of poverty more seriously in our decisions about how to spend aid. The incidence of poverty is calculated in purchasing power parity terms, which, roughly speaking, means that we’re considering consumption, not money. Moving towards an allocation of aid that is more closely based on differences in consumption makes sense: shouldn’t improving the consumption of the most destitute be a major part of the impact we want our aid dollars to deliver?

Domestic capacity has to play a role in the conversation. But if individual poverty is the outcome we really care about, then individual poverty should be an important determinant of how we spend aid. That could imply a shift towards a new model of how to prioritise our spending. And that, in turn, may mean spending less in (some) countries that happen to be labelled as poor.

Thanks to Matt Collin and Lee Crawfurd for helpful comments and to John Osterman for his work on the graphs.

You can download the Stata code that I used to generate the data for these graphs here. It uses data from Sarah Dykstra, Benjamin Dykstra, and Justin Sandefur, 2014. "We Just Ran Twenty-Three Million Queries of the World Bank's Website - Working Paper 362," Working Papers 362, Center for Global Development. Their data are available here. It also uses data on aid flows from the OECD (DAC Table 2a), available here.