After toiling away for decades in relative obscurity, development finance institutions (DFIs) have found themselves thrust into the limelight and told to transform “billions to trillions,” to fill the yawning SDG financing gap.
This turn towards mobilizing private finance for international development raises a whole host of questions: about the relative efficiency of public and private provision; about DFIs’ ability to deploy capital in the poorest countries; and about the development impact of private provision in social sectors, such as health and education.
But beneath those questions lies a more fundamental one: are DFIs adding to the quantity of investment in developing countries, or merely doing things that private investors would have done anyway? If DFIs are merely substituting for private finance, then their achievements—including the all-important indirect multiplier effects of “making markets” and catalysing follow-on investments necessary to get to trillions—will be illusory. Given the magnitude of hopes that are pinned on it, this question—whether DFIs are additional—is one of the most important unanswered questions in development economics.
And here’s the kicker: we don’t know the answer, and probably never will. In a new paper, The Elusive Quest for Additionality, written with Nic van de Sijpe and Raphael Calel, we show that obtaining a reliable estimate of additionality from investment data is all but impossible. The problem is the “demand led” nature of DFIs’ business model: they respond to investment opportunities, and researchers do not have access to all the information DFIs have when taking investment decisions. In the paper, we describe how DFIs go about investing, and translate that into a data-generating process which we can then use to test possible estimation strategies. We show that depending on the criteria that DFIs use to select investments, we might observe a positive correlation between DFIs’ investments and the overall level of investment, when DFIs are actually having zero impact, or a negative correlation when DFIs are actually having a positive impact.
That won’t surprise members of the Sacred Order of Social Scientists (whose initiation rites involve chanting “correlation is not causation”) but we hope our explanation of how DFIs go about investing will help sharpen the debate. Of more technical interest is our investigation into the performance of techniques that researchers use to overcome “endogeneity” problems: a supply-push instrumental variable, and the “system” GMM estimator. We show both of those generate false positives in this context.
An alternative to estimating the impact of DFIs on the quantity of investment is to gather information on the characteristics of the projects that they back and try to estimate the probability that private investors would have done the same deal. Unfortunately, because all investors base their decisions on their expectation of financial returns, which cannot be reliably inferred from observable project characteristics, that approach does not work either. It could be, for example, that DFIs snap up all projects with certain characteristics—e.g., those that look more developmental—which would leave no observations of private investors doing such deals in the data. Leaping to the conclusion that private investors would not have done those deals would be a mistake, but it’s what a naïve attempt to estimate the probability of additionality would conclude.
So, if we cannot “measure” additionality, what should we do? Our answer is that we should look for circumstantial evidence that would make us more inclined to believe additionality is present, even if it does not prove it, and ask DFIs to demonstrate that they are taking decisions in a way that means their investments are likely to be additional. An important point here is that we do not want DFIs only to invest when they are certain of additionality—if a DFI turned away 10 deals that only had a 50 percent chance of being additional, for example, that would mean not doing five deals that would have brought additional investment to capital-starved countries. The costs are asymmetric—failing to act is worse than acting unnecessarily.
There is also a point about incentives within DFIs. The people who are in the best position to know if a deal is additional are the transaction team working on it, but these are the same people who can have an incentive to pretend additionality is present when it is not. A successful DFI needs to do deals, but it must ensure pressure to “get money out the door” does not override voices inside the organization when they question an investment’s additionality.
Before concluding, an important caveat is needed: increasing the quantity of investment is only one way in which DFIs have development impact. They also seek to raise the quality of investment, placing requirements on their investees that private financiers would not—that can mean anything from requiring adherence to higher environmental and social standards, to pursuing a radically different business model with greater developmental benefits. While this “value additionality” typically does not help get us from billions to trillions, it’s an increasingly important route for DFIs to have development impact.
The idea that we ought to allocate billions of dollars from aid budgets to institutions that cannot definitively demonstrate their impact is uncomfortable, but the idea that obsessing over measurable results is bad for development should be familiar to everyone in the field. Traditional aid spending also faces risks, including the risk that money won’t be used as intended, which can be almost impossible to know for sure, because money is fungible. “Incredible certitude” is a siren song; better to acknowledge uncertainty and take sensible decisions in the face of it.