For years, Abhijit Banerjee has been a leading advocate of a rigorous approach to foreign aid. In particular, his call for randomized trials has helped shed light on the fact that the World Bank and other aid agencies do surprisingly little to properly evaluate the impact of their projects.

For example, the former head of the U.S. Agency for International Development, Andrew Natsios, recently acknowledged that “we don’t get an objective analysis of what is really going on, whether the programs are working or not.” In 2000 the bipartisan Meltzer Commission of the U.S. Congress found that the World Bank reviewed only five to ten percent of its programs within three to five years of the funds’ distribution. The Bank’s evaluations are done internally, and its measurements of sustainability are largely based on expectations of future success.

Adam Lerrick of Carnegie Mellon University notes that outside verification of World Bank self-evaluations is not possible because the necessary data is not made available to the public. He further notes that the reported success rate of World Bank projects shot up to 72 percent a year after the Meltzer Commission pointed out that, according to the Bank’s own evaluations, only 50 percent of their projects were sustainable. That is hardly the kind of transparent and accountable management that the Bank regularly demands of its client states in the developing world. Other than changing the name of the Operations Evaluation Department to the Independent Evaluation Group and using more favorable assessment criteria, not much has changed in the way the Bank evaluates its own projects.

The need for independent regular evaluations of impact is now recognized by a broad spectrum of observers. As Nancy Birdsall, the president of the Center for Global Development, states, “Without impact evaluations that are rigorous, independent, and thus credible, we cannot know what programs work. We cannot even argue convincingly that foreign aid itself works.” Ideally, such work would be done by private firms reporting to the donor governments. In many cases, randomized trials of the kind advocated by Banerjee would be the preferred approach. In other cases, other types of analysis that emphasize success in the long term and evaluation of actual output or accomplishment would have to be done.

But we should not expect the emerging consensus on the need for independent audits of aid performance to include aid officials. For example, Branko Milanovic, an economist who works on poverty issues at the World Bank, objects to randomized trials because they treat people like “guinea pigs.” Beyond recommending the creation of a utopian global-welfare agency, he does not offer a way to measure or improve the effectiveness of aid. Other aid officials have also objected or remained silent, knowing full well the political difficulties of promoting accountability within bureaucracies, especially international ones.

So are there good reasons be optimistic about aid? Or should we instead focus on other factors that cause growth? As much as I support Banerjee’s push for randomized trials, I am less convinced about his case for aid optimism. Banerjee notes that major aid agencies have real power they can exercise in convincing recipient governments to do the right thing, that aid in the past has done good, and that aid thinking has evolved so that success is no longer measured by the amount of money given.

I take issue with all of those points. Indeed, although the thinking on aid among development experts has evolved in the past two decades, there is a huge gap between what might be called the expert consensus and the political push for massive increases in aid. Most of the literature in the past decade or so has been characterized by aid skepticism, has suggested a modest role for aid in promoting growth and poverty reduction, or has concluded that aid will work once recipient or donor behavior changes in some way. Banerjee’s proposals fall into this last category.

Yet the UN’s campaign to double worldwide aid rests largely on the idea that the supposedly low level of aid funding—rather than government policies or behavior—has been a central reason for the failure of the poorest countries to grow. Proponents of this view therefore argue that the funding increases should be made quickly through the established aid agencies and should promote a long list of broad goals. The UN adviser Jeffrey Sachs’s Millennium Villages Project, set up in selected low-income country villages to show what big aid increases can accomplish, is seeing significant funding without any apparent scientific attempts to test for spending effectiveness.

Are we really to believe that large increases in aid disbursed by a multitude of donors are going to improve the incentives of lenders and borrowers alike? Is more money really going to improve the discouraging record of aid conditioned on policy change when a major problem (well recognized by borrowers) continues to be the aid agencies’ institutional urge to lend? The politics of official assistance do not support the case for optimism. The fact that some aid has been successful in the past is also not encouraging—it would be difficult to spend trillions of dollars over a half century without doing some things right—but it does underscore the need for testing the impact of aid.

Thus, while the push for more aid will surely weaken the push for more accountability and effectiveness, I fully support efforts to audit the performance of aid, and I admire Banerjee’s optimism. In the messy world of aid, we may well end up with more independently evaluated projects. The trials and evaluations will help determine whether aid can be central to development or whether it has a limited role, a possibility that we must keep an open mind to.