On this page I report instances in which antipoverty policies are being advocated in ways that largely ignore the economics of poverty.
In a recent blog post, Mikail Rutowski calls for expanded public spending on safety nets, based on some estimates he quotes of the impacts on poverty of existing safety nets. Mikail heads up the World Bank’s Social Protection (SP) group, which advocates SP programs in the developing world. Mikail claims that 69 million people in the developing world have been lifted out of poverty by such programs (relative to the Bank’s absolute line; 97 million are claimed to have been lifted out of the poorest 20%). What is this based on? Is it believable?
The calculation comes from a database maintained by the Bank’s SP team, called ASPIRE. I am a fan of ASPIRE for the easily-accessible data it provides on coverage of SP programs. (I use these data in Chapter 10 of EOP.) But a big cautionary note needs to be attached to their estimates of the impacts of social safety nets on poverty. (And that note is currently missing from the relevant tables.) While they are not very clear on the website, as I understand it their calculations simply subtract public transfers received from total income at the household level. One then compares the poverty count for total income with that for total income less receipts from the SP programs.
There are a number of reasons for caution here. Of course there are measurement errors in survey data. (Are these programs targeting the right people? Are transfers being reported accurately?) But maybe the biggest concern is that the ASPIRE calculation ignores all behavioral responses to public transfers; nothing else changes relevant to recipient incomes. For example, if private transfers (from family or friends) fall in response then this is ignored. So too are the forgone wage earnings of teens who would otherwise be working without the conditional-cash transfer. Indeed, any changes in work effort within the family are ignored. These responses will mean that there is less impact on poverty than the 69 million number. Against these effects, cash transfers can help families get over economic constraints they face, such as in access to credit. So the impacts on poverty might be higher than 69 million. On top of all this, the calculation ignores effects on prices and wages, though I can be more forgiving on that score.
In short, this is a “Non-Economics of Poverty”—indeed, no economics at all! The “69 million” is almost certainly off the mark, and maybe by a wide margin. Figuring out plausible bounds on the true number would be useful. Researchers at the Bank and elsewhere have long recognized that there are behavioral responses to transfers, and have done quite a lot of research on those responses that could help in establishing plausible bounds.
In fairness to ASPIRE, their Non-Economics calculation is still common. (I admit I have also made a non-behavioral assumption at times, to make a quick calculation.) But we can all do better, and we can certainly flag caution for users. At a minimum, some sensitivity tests should be provided in the (otherwise) excellent ASPIRE database.
This can be such a deceptive number. It is very easy: one simply adds up all the monetary gaps below the poverty line for all poor people. (You can calculate it easily as the Poverty Gap Index times the poverty line times population size.) One widely-cited calculation in a 2016 Brookings article gives the aggregate poverty gap for the $1.90 a day poverty line at $80 billion in 2015.
This $80 billion figure is evaluated at market exchange rates. This was presumably done by the Brookings authors because they are imagining that the gap would be covered by foreign aid, or taxes on the global rich. Then market exchange rates (MER) make sense in representing the aggregate cost. Given that the cost-of-living is lower in poorer countries (due to the presence of goods and services that are not internationally traded, and so are cheaper in poorer countries where wage rates are lower) the reasoning here is that $80 billion in the form of external aid paid out at MER will buy enough to cover the aggregate poverty gap when calculated at purchasing power parity (PPP) exchange rates which is roughly double the $80 billion figure.
However, it is likely that the bulk of the cost of this transfer policy will be financed from domestic resources in developing countries rather than external aid. (Even if aid donors agree to pay out directly and explicitly for such transfers, the money is fungible.) Then the $80 billion number is potentially deceptive. In thinking about the cost of this policy we should focus on the opportunity cost to developing countries. Then one would want to know what the public expenditure on transfers would buy within those economies given the prices that prevail there. For this purpose, the cost evaluated at PPP exchange rates is more relevant. And then the bill would double. So the $80 billion number is potentially deceptive.
But lets put this point aside, as there is a deeper issue that worries me more. The conclusion drawn from the $80 billion calculation is that it should be easy to eliminate poverty. The trouble is that this entirely ignores all the information and incentive issues that confront targeted transfers as I explain above under the heading “How much do current safety nets reduce poverty?” The types of “poverty gap” calculations one often sees are examples of the non-economics of poverty.
Even ignoring all that, there are some confusing usages of the idea of the aggregate poverty gap. For example, any less-than-careful readers of this (otherwise interesting) blog post by Peter Diamandis on Universal Basic Income might be be drawn to conclude that $80 billion would suffice to provide a UBI sufficient to eliminate poverty (though he does not actually say that). But a moment’s reflection tells us that the global cost of a Basic Income of $1.90 a day for everyone in the developing world is close to $4,200 billion at PPP. Even if one halved that number to convert to market exchange rates, the aid bill would be 25 times greater than the $80 billion figure. (Of course, if Diamandis had said that then his point would have lost some of its punch!)
While I do think the Basic Income idea should stay on the table of policy options (see my VOX-EU blog), I also think advocates should be careful with their numbers and how they are presented to avoid misunderstandings that could easily undermine their case with closer scrutiny.