 Thank you for coming this early morning. So I'm going to start with the aggregate relationship between aid and growth. And go start off with a basic question. Thinking about aid and growth, why are some countries poor? Many of them. And actually, there is a pretty simple answer. It's that they just don't produce very much. I started doing work with household surveys in Mozambique. And we did a household survey, and we did a counterfactual of a perfect transfer program. So every household has the same income. And in that case, in 1997, every household in Mozambique would have been poor. So it's the lack of goods that makes up the problem. Why do they produce so little? Well, they have rudimentary technology. They don't have much human capital. They don't have much physical capital. And even in cases where sometimes they add some physical capital or some human capital, they often lack the institutional structures to make this function well. Well, why is it that they lack this wherewithal? And the reason is they've failed to accumulate. There is this long process of accumulation. We stand here in this building, in this university, that's been built, buildings built probably before we were born. That's part of the accumulation. Many of you have children who are going to school this morning. Those children won't contribute to GDP for another 10, 15, 20 years. This is a long process of accumulation. So the role of aid is to facilitate this long run and often fragile process of accumulation. And that's what we're looking at today, have we managed to facilitate this long run process of accumulation? So and looking at the recent cross-country literature, there's been a piece published in 2008 that has been quite influential. And they look at lots of countries through time. And they look to see whether aid has contributed to growth across this set of countries. And they conclude that no, that there is no detectable impact of aid in this sample of countries through various time periods and using various methods. And so in essence, this is what's been enabled the micro-macro paradox. And it's been revived. Why is it called the micro-macro paradox? And this is fairly well accepted. By and large, through the aid literature, if you look at project evaluations, not every project succeeds. It's true. There are failures, as Finn was pointing out. But by and large, projects evaluated ex post receive positive evaluations. On average, we have positive project evaluations. We have positive what we call impact evaluations. Development economics has been at the forefront of what has been called the credibility revolution in econometrics. So they look in a very rigorous way at whether certain interventions with the project and without the project or with and without the intervention have actually had impact. And again, not every impact evaluation conducted comes to a positive conclusion. But most do. And then we often get positive sector evaluations. Aid does seem to have an impact on education, enrollments at least, on health. And yet, we find no detectable impact on growth. So there is some incoherence in this. And the macro literature is the outlier. There are a series of positive contributions of RS08. Rajan was the chief economist of the International Monetary Fund. Subramanian is not there as a charity case. They're smart guys. And there's a series of things that help really to advance the literature. And one is they establish a clear prior. So they look at modern growth theory. And they say, well, how much growth should we get if aid is being invested according to modern growth theory? And they come up that if aid is about 1% of GDP, then the increment to GDP growth should be about 0.1 percentage points. So if a country without aid was growing at 1% per capita per year with aid at 1% of GDP, it would grow at 1.1% per year. Second, they take a long-run perspective. Not throughout their entire article, but their principal conclusions are based on long-run cross sections, looking from 1960 to 2000 and 1970 to 2000. We agree with that because of this idea that growth is a long-run cumulative process. Finally, they set the standard for addressing the endogeneity issue. And this is what Finn was mentioning as the attribution problem. As Finn was mentioning, we are now, Denita is in the process of leaving Vietnam. All the monies that are being allocated to Vietnam will go somewhere else in a few years. Maybe that money will go to, hopefully that money, will be able to go to Somalia. They need it, but that money is going to Somalia, if it goes there, because they're poor. This is the attribution. And if we're not careful statistically, we can be concluding, we're in danger of concluding that something like mud causes rain rather than the other way around. So what have we done? We begin, and again, Rajan and Subramani kindly shared their data and their codes. So we're able to begin from exactly where they start. And this is the standard in this literature. So we want to know what exactly the differences are. So if we begin from exactly where they begin, then we know what the differences are. So we are able to reproduce their results exactly, and then we look at what's happening. And we make three sets of improvements. I mentioned the credibility revolution in econometrics program evaluation. This is the new perspective. And we develop what's called an estimator, an approach for looking at this that is based on this new approach in econometrics. This approach has implications for what we call the specification. These are our controls, what we're controlling for. So we make some changes for the way we control for other factors that are influencing growth aside from aid. This turns out to be one of the least having a smaller effect. Finally, we strengthen the instrument. And the instrument is what we're using to try to deal with this endogeneity or attribution issue. And here, as I said, Rajan and Subramani have set the standard. They do a smart thing, which is instead of trying to look at aid from the perspective of the recipient, they look at aid from the perspective of the donor. How much aid is being supplied rather than how much aid is coming in. And so there is a database at the OECD DAC that will tell you this kind of information. So you can go in there and you can hunt and see if how much aid did France give to Guinea-Bissau in 1978? The principle should be there. But in a lot of cases, it's not. You go and you look, France, 1978, give to Guinea-Bissau, there's no information at all. And what does it mean? Well, potentially two things. One, they didn't give any. Often, we don't report zeros. Or France gave something, but OECD doesn't know what it is. It's missing. And we worked on trying to figure out which one it is. Finally, I won't tell you how long it took us to come to the conclusion that perhaps we should ask somebody at the OECD DAC what we should do. And they came back very quickly and said, if it's not there, it's zero. Or it should be almost all of the time. And Rajan and Subramanian had treated it as missing, which creates quite a difference. There's an error in the instrument part. We do some other things to make the instrument stronger. But this turns out to be one of the more important things that we actually do. So what does it do at the end of the day? These are the sets of results that we have according to the various permutations. And the first thing to notice that their result, the result that they actually obtain, is exactly their pride. Point one, that's what we pointed out. It's exactly what they got. It's just not statistically significant. It's not estimated precisely enough. That's the first thing to point out from this slide. The second is, when we shift over to more this credibility revolution program evaluation perspective, the estimator, then in all cases, regardless of what else is done, the estimate is positive, significant, and in the range implied by growth theory. Alternatively, when we use an instrument that is missing the data error and has certain other additions, then in every case, regardless of if we use their specification, our estimator, their estimator, we get positive and significant implications for growth in the long run. Finally, the more of these improvements that we add, the more precisely the parameter is estimated. We end up with significance at the 99% level. So a quite precisely estimated coefficient, quite close to what we think is a reasonable prior. So what we get is for every percent of GDP and aid, you get 0.13 increment in per capita growth. So if it was without aid, 1%, with 1% of GDP, you would get 1.13% growth rate. What do we conclude? We conclude that on average and over time, aid contributes positively growth at levels predicted by growth theory. This is over periods of three decades or more. We also conclude that there is no micro-macro paradox, that in fact there's consistency across the project evaluations, the program evaluations, the sector level, and the macro level when taken over a period of time. And this we've published in a new journal edited by Joseph Stiglitz in their first issue. And this has been a very, very popular paper. One would think that an article by Joseph Stiglitz in his new journal on the financial crisis would be a popular paper. And it is a popular paper. But in fact, this paper has been downloaded more often than Joseph Stiglitz's paper since the papers have been published. It's generated a lot of interest. And many people think that it sets the standard for this particular set of analysis. So that's my contribution. And Sam is going to come up and talk about it, a different form of consistency.