 This presentation is slightly different, or you could say very different. One difference is it's not based on a paper, rather it's reflections about some specific challenges we will face in taking research on aid effectiveness forward, and I'm particularly interested in taking it forward through country studies to try and give greater understanding of how aid has worked or operated in particular countries. So I view that as a compliment to the kind of cross country studies. The basic principle I'm going to have here is that if you want to do aid effectiveness studies, there's two fundamental pieces of data you need. You need a measure of GDP or GNI or whatever you want to call it, you need a measure of national income, and you need a measure of aid. So what I'm focusing on here is, well, we know both of these are measured with error. So the data we have available is, it's subject to measurement error, and the issue I'm concerned about is, in particular, for GDP, you have different sources of data that all purport to be measuring GDP in the same country over the same period of time. Are they the same? Do they tell you the same story? And I'll illustrate that one. And then secondly, then going on to the aid side, there's a particular concerns about measuring aid, and especially because most of the data we have and all of the data that's used for the cross country analysis is the donor data. It's what the donors say. They've allocated or dispersed. And of course, there's a difference between what they say they've committed and what they disperse. But that's not the same as what the recipient countries receive. And one of the main limitations in current data is that we have very poor data on what the recipients receive. And I'll talk about that. I just want to emphasize that maybe to the audience here it goes without saying, but I just want to emphasize I think aid effectiveness is an important issue. It's important as a research question, pure intellectual curiosity. It's an interesting topic. It's politically an important topic in donor and recipient countries. And of course, the research can have significant policy implications. It can encourage changes in the design of aid instruments or the delivery of aid, or it can be used selectively to support particular changes. So it is an important issue, so we should work on it. Now I'm going to, one could characterize a lot of the cross-country regression work as asking, does aid work? Usually defining that question in a very precise way about is a particular coefficient positive and significant. That's informative and it's troublesome as Sam and Henrik in different ways have over viewed. But often it doesn't tell you very much about how aid works. And I choose this phrase how aid works because in English it's quite a nuanced phrase. On the one hand it can mean that you're going to assess the mechanisms of the way in which aid has effects, whether they're good or bad. And that would be one interpretation of how aid works. You're going to talk about the how. Some of the things will be good and some of the things will be bad. And in that way the inflection or the emphasis is on how. But you can also put the emphasis or the inflection on works. So that how aid works is actually an introduction to a narrative of how it's been effective. I make that point because, you know, fundamental to that there are, people do come often with ideological preconceptions and that can affect the language they use and the way in which they interpret phrases. And I just want to be upfront about that. And my interpretation is the first one. It's an assessment of the mechanisms. They may be good, they may be bad. And I think to get a lot of understanding of that requires case analysis and in my case I mean country studies. Now country studies are not a panacea. They don't give you a simple solution. The obvious problem is it's very difficult to generalize from country studies. I don't find that a problem. I would actually personally be far more interested in actually gaining some understanding of the way in which aid has had effects in Tanzania or Ghana than getting an average coefficient across a whole range of countries. So I think it's useful. One way of addressing the generalizability issue is, well, you can take a common methodology and apply that to a range of countries. And a very good example of that which I'll come back to is the paper by Giselia Smaller and Tarp in the Oxford Bulletin where they use the co-integrated VAR methodology to address the question of what's the effectiveness of aid in sub-Saharan African countries. So they apply the same methodology to 36 countries and reach inferences. Now I quite like the co-integrated VAR methodology for time series analysis. Katrina Giselia describes it very nicely as letting the data speak for itself. And the methodology that they've developed, which Henrik has contributed, tells you how to frame and answer the question. So how to ask the question of the data. But you need something more to interpret the answer. You should have some theory to guide you into how you interpret the answers. And fundamentally, you need a narrative. It's not the econometric analysis itself that is really important in a country study. It's how it quantifies and supports the story, the narrative of what was going on. But there's another aspect to see VAR analysis which is what I'll focus on is, yes, it's a good idea to let the data speak for itself, but you should also take in mind, are you asking the right data? Because if there are different sources of data, it's worth considering would they give you a different answer. And that's one issue. And then in the aid context, the issue I'll talk about towards the end is when, or under what circumstances, might we be more interested in getting recipient data. So data from the recipient's perspective of what they've received in aid. And the fundamental issue here is that what they, in so far as one can measure it, it's quite clear that what they receive is not the same as what the donors allocate. One caveat I should attach for any earlier career researchers here with publication and progression ambitions don't start off with country case studies. Because they won't help your career. It's only people who've reached a stage where they don't have to care so much about journal publications that they can then think, right now I actually want to learn something about what really happens. And they can afford to do the country studies. Economics journals in particular don't seem to lighten them very much. Now, so the Sub-Saharan Africa, this is kind of reiterating. I'll acknowledge, yes, there is this debate by Morton German and others on the reliability of African GDP data. I acknowledge that. And, but my view is yes, that's true, but I'm not going to, I'm not, I don't agree of going as far as to say this data is worthless. It tells you nothing. I think the data, the data has information. It's what's measured. It will have influenced policy decisions. If they're using that data, that data has influenced decisions that they're making, it influences analysis. So that's the data we have. It's, it may be, it does have measurement error. So we need to be recognized that and address it. And then when you're looking at the aid aspect, you need to go a bit further and say, okay, what's the aid that the recipient has received? So let's consider the GDP data. And I'll start here with Togo. So I'm just going to pick a couple of countries to illustrate. And what we've done here is from 1965 to, I think it's about 2007, the different lines are Pen World Tables 6, Pen World Tables 7, Pen World Tables 8, and World Development Indicators. And all of them are GDP series, calculated on an internationally comparative basis. Now the first thing that's pretty obvious is, well, they're not the same. And I picked Togo because it's one illustration of a case where not only are there level differences, which you can often explain by the way in which a change is in how it's constructed. But there are also some trend differences, not major ones. But, you know, so you might have a suspicion, I want to say that you're using that whichever of these series you use in your analysis, it may affect the inference. Here's Uganda, I'll put that here because I'll come back to it later. Here you actually might be more confident. You see, there's not that much difference between them. Pen World Table 6 for some reason seems to give a higher measure than all of the others, but aside from that wedge, the series look pretty much the same. And here's Tanzania. We've redone these because of the availability of the World Development Indicators. This one's indexed at 100 at the end so that you're comparing them similarly. And again for Tanzania, certainly since the mid-1980s, the series have been quite similar. But once you go back earlier, there are differences. So then this begs the question, OK, do these differences matter? Oh, and just to add, investment GDP is another important variable that you can derive from these. This has been in, again, whichever series you use, it makes a difference. They're not the same. They move more or less the same, but again, Pen World Table 6 seems to think it was a lot lower than the others. So does this affect your inference? Well, this is illustrating. This is an exercise that a PhD student of mine has done, Lionel Roger. And what it does here is it takes the Giselius, Muller and Tarp paper, their sample, which they estimated using Pen World Table 6, applies for the exact same sample. In other words, insofar as possible, the same countries for the same years, applying the same methodology, same like lens, everything else. So the methodology is not the specification. The methodology is the same. The sample is the same. The only thing that changes is the measure of GDP. So when he does it with Pen World Table 6, he gets there are some small differences, but he gets the same inferences that for 26 of the countries, you can conclude that aid has been effective. And there's 10 countries where maybe it hasn't, or it's been maybe even damaging, but certainly not effective. The only point I want to make here is that as you move through the series, the results start to change a bit more. There's a fair difference when you move to Pen World Table 7. The differences in Pen World Table 8, the number of countries for effectiveness has fallen, but the number of countries for harmfulness has fallen as well. The WDI is the most different, but the sample is a lot smaller. It's more difficult. You can't get the WDI. It doesn't have the same length of data for all of the countries. So the bottom line of this is that the source of data can matter for inference. In about 2 thirds of the cases, so let's say in about 20 of the countries, it doesn't have an effect. The inferences are the same for those countries for all four data sets. So in about 2 thirds of the cases, it's the same. But there are a set of countries up to 10 where it does seem to matter. And it seems to matter quite a lot that you change, only change the source of the GDP, and your inference changes. The next stage in our work is to try and work out, well, what is it about these countries that makes them different? One type of thing, as you might expect, is they're often smaller countries that have had more volatility. They've had periods of conflict or periods of instability. And that would lead you to believe, well, yes, actually, it's quite sensible that they may have been periods where the data is less, where the measurement error in the data is greater. So we want to pursue that. But I will emphasize that in 2 thirds of the cases, the inference doesn't change. So now if you're going on to country studies, you get an even extra problem. Because now, first of all, you can bring national GDP statistics to the table. And the reason you might want to do that is because a lot of the other variables you may be interested in are going to be collected in national currency at current or constant prices or whatever, like tax data or public expenditure. So you bring that in. And then you have to ask questions about, OK, well, how do we measure the aid? Should speed up a little. This is just taking an illustration, data compiled by Thomas Poirier for Uganda. And again, the red line is World Development Indicators. And the dotted blue line is Uganda Bureau of Statistics. Up until about the mid-1970s, they're almost identical. And from about the late 1980s, they're pretty much the same. But there's a big difference in the 1980s. Now, of course, that's not surprising, because the 1980s, period from 1980 through to 87, was conflict. So you can explain why there are differences. You might have a suspicion that what the national statistics or what the Uganda Bureau of Statistics was doing was just a smoothing estimate. They had data from the late 1970s, so they were just smoothing it and adding an incremental increase each year. That would be consistent with what you see, whereas maybe the World Bank had better data or made different judgments. But I just want to emphasize that if you're interested in that period, it matters which set of data you use. And if you wanted to draw an inference say about the effects of the conflict on Uganda's growth performance, it would really matter which data set you looked at. This is, again, just giving the same story, looking at GDP per capita. And the next one is just to make one other point. Of course, even though there are periods when these series do look very similar, when you calculate the growth rates, and of course, the first difference is what you will have in the analysis, then there are quite different series, even where they look quite similar. Now kind of quickly to say, OK, how about the data, the aid data? So what I'm doing here is comparing three series. This is compiled by another PhD student, Amelia Timmis. And the green line is World Development Indicators. The red line is the DAC data. And the blue line is using Tanzanian data, compiled by Josephette Quaker. They're different. Now they should be different because they are, in particular, the Tanzanian data is measuring something different. It's measuring what the government records as it having received in the budget. But if, as I am, you're interested in the effect of aid on the behavior of the government, in particular on taxation and on public expenditure, that's the measure you want. You want to know what did the government think it was receiving because that's what will have influenced its fiscal behavior, not what the donor said they were giving. And they're different. This is illustrating it for Ethiopia. Again, they're different. That's just what I wanted to emphasize. The DAC and Ethiopian data are actually quite similar here. And it's the World Development Indicators that's the most variable from these. So it's just to make the point, the data matter. The data differ if you've got different sources or different ways of measuring it. And if you're doing country studies, I think if you're doing cross country studies, you should also be aware of this. But if you're doing country studies, you have to face this and address it. And it's a challenge. I'm not saying I've got the answers. That's what we're still working on. But I do want to emphasize the challenge because we have to recognize measurement error. We have to recognize alternative definitions and work out which is the definition and measure that matters, theoretically or for the narrative, for the effect you want to analyze. The inferences about the effects may depend on which source of data you use. So that should be taken into account and considered. And ultimately, if you want to understand how aid has worked or the effects it's had in a particular country, you need to begin analysis that looks at the issue from the perspective of that country. What did they receive? How did they perceive or record the aid they were getting? And indeed, how did they perceive and react to the conditions policy advice that was associated with this aid? Because it may not be the nominal monetary value of the aid that matters. It may be the associated conditions and reforms. But you need country analysis to understand that. Thank you.