 I want to thank the organizers for inviting me. I should also say this work was born out of a World Bank IG evaluation study that I was working on together with Telco and Raghavan. And for that work, we were looking for indicators to assess countries' performance in domestic revenue mobilization. And we found that most existing indicators, in fact, focus on achieved outcomes. So classic example, just quoting revenue to GDP ratios. And most of these indicators suggest that low-income countries perform quite poorly in DIM. And of course, one of the main drivers behind that is that these indicators do not consider differences in the fundamental economic conditions that countries face. And therefore, sorry for that, the laptop still shows it. Yeah, it comes back. And therefore, have a sort of natural tendency to underestimate the performance of governments and countries that face less favorable conditions. So that is the motivation behind looking at estimates of tax efforts. And what we're doing in this paper is basically to add another tool to the toolbox with differences, pros and cons, I guess, to other approaches, such as the one just presented by Kyle. Now to illustrate that a bit further, on the left, you see box plots of revenue to GDP ratios by income group. And you can see that clear positive relation, right, where low-income countries just achieve much lower levels of revenue to GDP ratios than richer countries. And of course, part of that may be due to suboptimal tax structure, inefficient administration, et cetera. But the idea is that another part of that is also due just to differences in economic fundamentals. And on the right hand side, we try to capture this by a normalized composite index, accounting for some of the factors that are considered important in the literature. And the idea simply is that low-income countries also feature, for example, higher shares of agricultural GDP, which is often considered a hard-to-tax sector. They tend to have relatively less international trade, which is easier to tax sometimes. Or they tend to have higher age-dependency ratios, so corresponding to a narrow tax basis. Now what we do in this paper, we construct a new measure of D&M performance at the national level that does account for these differences in domestic economic conditions. We leave out institutional and political factors for a reason. They may be considered then in a secondary step. And what we argue for, what we use, are two tools from data and development analysis, which is a nonparametric approach. So I think, basically, at the other edge of the spectrum of tools, with stochastic frontier analysis being somewhere in between. So this is purely nonparametric. That means it's of a descriptive nature which comes with certain benefits and certain limitations. Now the output variable is just government revenue as a percentage of GDP. The input variable is intended to be a composite index that proxies for the enabling domestic economic conditions for D&M, such as the ones that you've seen on the previous slide. I want to give you a very brief intuition of the method what D&M does. So if this is the sample of observations, country year observations, Kyle showed you the stochastic frontier. D&M will look similar, but I think it's even more simpler to grasp. So in D&M, each unit on the frontier is considered to be efficient simply because there is no other country in the sample that is achieving a higher output with the same amount of input. So if you consider this country, for example, of course, there are many other countries in the sample with a higher revenue to GDP ratio, the output. But these countries also feature significantly more input or more favorable economic conditions. So each country on the frontier is said to be efficient. The countries inside, their efficiency is quantified based on the distance to the frontier, so the efficiency gap. And then the whole insights we derive in that paper are based eventually on these efficiency estimates. Now I said this is of a descriptive nature. So there are some advantages in terms of the technicalities that we're not going to talk about here with assumptions involved and so on, but we think the main advantage really for using this to inform policy making is that the outputs, these efficiency scores, are super easy to interpret. So you do not have to deal with T-values, P-values, issues of indigeneity, et cetera. It's purely descriptive. So based on if these measures, if the variables are selected in a way that you care about, if this is your measure of efficiency, then it's basically as simple as interpreting a genie coefficient. So you have a score that is normalized to range from zero inefficient to one efficient. And for example, an efficiency score of 0.6 would mean that a country is currently achieving 60% of the revenue to GDP ratio that it should in principle be able to achieve given of what other countries in the sample with similar enabling factors are already achieving. That's the intuition and the way to interpret this. Now I'm going to show you some results. These are based on a sample of 118 low and middle income countries. We exclude high income countries here. The output variable government to GDP ratio comes from the very nice univided data set, which we selected a few years ago, mainly because it had the best country coverage. So for this application, at least, even better than what we found than IMF data sets. The input variables come from the World Bank, but I should say we selected them based on the literature review, but they can be tailored or fined due to the particular application at hand. So this is merely a suggestion. The main contribution here is in the method, basically. So this is the main picture. You can see two frontiers here. The upper dashed line is the frontier for the whole sample. The lower is the one for low income countries only, the black dots. And what you can already see here is that the low income countries all tend to be located relatively closely to any of the frontiers. In fact, many of them are closer to the frontier compared to the red or the blue dots, the middle income countries, despite the fact that most of the low income countries have significantly lower revenue-to-GDP ratios, right? And the whole insights that we derive in that paper are basically based on the distance of these dots to the frontier, the efficiency scores. Okay, here are box plots of these efficiency scores by income group and region. And if you remember before I showed you the box plots of just revenue-to-GDP ratio by income group and there was this clear positive relationship. Now, looking at the efficiency scores, the relationship is basically flat. So meaning that a lot of the low income countries are in fact now very efficient once considering they are on average less favorable conditions and more efficient than many of the middle income countries, right? If you look at geographical regions, we find that the issue of high or low efficiency in DIM is not concentrated in any single region. It seems to be the case that every region has some countries with high efficiency, some countries with low efficiency. Here's some more number just to illustrate that once more. So revenue-to-GDP ratio, low income countries are about half of what upper middle income countries achieve but their efficiency score is just slightly lower than that of middle income countries. And again, the intuition is that these efficiency scores also take into account the low income countries have weaker enabling conditions that are shown here in the second column. Now we can also check changes in efficiency over time between 2012 and 2019 here. And you can see that in fact low income countries were catching up at least many of them and on average you have the biggest increase here. For regional averages, you can see this also applies to South Asia, Latin America and perhaps less Domina and ACA with negative changes. Now the global average is 62. So this suggests on average these low and middle income countries in the sample are achieving 62% of the revenue-to-GDP ratio that they should in principle be able to, which I guess is somewhere in the range of what other people find. So to conclude, obviously we are not suggesting this should replace existing industry but we are suggesting this should replace existing indicators but we think it might be a useful method to add to the toolbox, mainly because of its easy way to interpret. It also provides quite a rich set of insights that can be used for other types of analysis. So for example, informing country-level analysis. One other advantage or difference in regression-based approaches is that in a regression-based approach, the slope that you estimate is always affected by all the countries in the sample. Every observation has some impact on the slope. In DA, this is not the case, the efficiency of a given country is evaluated for a segment of the frontier and this segment is only driven by countries with similar enabling factors and these are called peers in the DA literature. So we can identify sort of the global peers of each country and they may not necessarily be the neighboring countries or even countries in the same regions, they may be located somewhere else on the globe but might be interesting to compare to them in a more detailed analysis. And what we do in the paper is we use the DRM efficiency scores then in a second step in a regression framework either as the dependent variable to see what might be driving differences in DRM efficiency across countries and there we then consider political or institutional factors as one possible explanation or you can also use the efficiency scores as the explanatory variable and then depending on what the left-hand side variable is, you may address different questions. For example, we are using World Bank Group support targeted especially for DRM as the dependent variable to see whether this World Bank Group support tends to target countries with high or low untapped potential in DRM. If you're interested in the results, they are all in the paper or feel free to come and talk to me. Thanks so much.