 Rwy'n meddwl, mae'r bobl yn ymddangos. A'r gweithio'r gweithio, a'r amser, yn ymddangos, yn ymddangos. Mae'r gweithio'r gweithio, yma, mae'n gweithio'r gweithio'r gweithio. Felly, mae'n gweithio'r gweithio'r gweithio sy'n dwi'n helpu y bod yn ymddangos yn gweithio gweithio. A'r gweithio'r gweithio yn ymddangos y gafod dŵr y gweithio ac y gweithio'r gweithio, mae hyn ymddangos, mae hyn yn ymddangos. Mae hyn yn ymddangos, a gwond hanes i ddweud efallai Rwy'n gweithio'r gweithio, mae hyn yn ymddangos. Yna, mae'r amser y t-r-a'r methu yw Carl yw'r cyfnod, yw'r cyffredin, mae'n gwybod i'r 80% ar y dyfodol. Yn ymdegwch, y dyfodol ddiadau ymdegwch, yw'r cyffredin, yw'r cyffredin, yw'r cyffredin, yw'r cyffredin, yw'r cyffredin, yw'r cyffredin, yw'r cyffredin, ac yw'r cyffredin, ac mae'r rhaid o'r gweithio ychydig yma yn ddigon. Mae'n fawr mewn gwirionedd ac yn ymgyrch yn ymdegwch, mae'n gweithio'n gwybod, yw'r gweithio yw'r gweithbeth, yw'r gweithio'r gweithio'r gweithbeth, ond yw'r residul yw'r gweithbeth yw'r model yw'r gweithbeth yn ymdegwch yw'r gweithbeth. A'r gweithbeth yw'r gweithbeth yn ymdegwch ar dda, felly mae'r gweithbethau yw'r gweithbeth o'i gwybod i'r buswch y bwysig ar gyfer'u gweithio. Ond o'n gweithio'r problemu, rydyn ni'n ei gweithio'r cyffredin o'r ysgolwch o'r cwmwyntau yma yng nghymru o'i gweithio'r bwyswch. A'i gweithio'r gweithio'r gweithio. Felly, mae'n ei gweithio'r cyffredinorion yma. Mae'r gweithio'r gweithio'r gweithio'r gweithio'r hwn. Mae'r gweithio'r gweithio'r gweithio, mae'r gweithio'r gweithio, gwahanol yn cydweithio cydweithio, yw'r cydweithio? Dwi wedi bod, ysgrifon sydd gennym ar yr hyn o ddweud y dda, dda i'r gwahanol yn omdegol, yw ddweud dechrau, hyd yn dyma yn dweud, roeddon yn gwybod yn cydweithio ar gyfer gwaith, ac mae'r gwahanol yn rhan, ac mae'n rhan o gyfragoedd ychydig. Rwyf ar� gyda'r 19 dioled, a llaw wladach chi, yn meddwl fel ynddo i ddweud o gwahanol, mae'n bydd ar gyfer gwaith sy'n dweud, it's damaging your economy you have to reduce tariffs and they've done it. So why is trade still significant? Because trade taxes have been dramatically reduced. If you include trade... as a measure of tax potential it's not because you're talking about trade taxes per se because they've also been told to reduce then it's because you think well it's a measure of economic activity and that's what you're trying to proxy. But then you've got to ask, yeah, but is it or what is it capturing? And then the other commonly used measure that people have is the share of agriculture in the economy. Why is that included? Well, the large audience, it's very difficult to tax agriculture. And you know, countries collect low formal taxes from the agricultural sector, from informal labour. felly mae'n ddweud y sector informaidd o'r sector cyfnodol. Ond y gallwn y gallwn i'r cyfnodol yn ymgyrchol, a'r cyfnodol yn ymgyrchol yw'r dystor sy'n cyfnodol, ychydig i'r cyfnodol yma o'r ffordd o ffordd o ffordd o ffordd, o'r cyfnodol, o'r cyfnodol, o'r cyfnodol, ac mae'n ffordd o'r cyfnodol, ac yn ymgyrchol, y cyfnodol yn fawr yn ddweud. Yn ymgyrchol, ymgyrchol i'n ddweud, yn fawr i'w dda. Mae'n fwygoch yn ddweud caelmiyor i ddalun. Felly, dweud ymdoedd o'r clasifol yma yn ddweud, maen nhw dyfynu'n ddweud. Felly, rydych chi'n mynd i'w ddweud y ffordd o'i ni'n cyfnodol ymdamiol, ywed o'r cyfnodol yma y bydd yn ddweud, yr eifyd yn ddweud y bydd y ffordd o'i ddweud o'r cyfnodol, o'r cyfnodol, o'r cyfnodol. That's fundamental. So again, when you're including that variable, it's not working the way you want it. I'm using that to motivate what Gregor touched on and what Amelia was touching on. How do we want to interpret effort and how do we want to use it? Who's effort? I would make a big distinction between political effort, which is the process by which the state or the government decides how much tax revenue do we want to collect, or how much revenue do we want to collect, where tax is one component of the revenue. And you can do two general ways of looking at that. They've got expenditure needs at desired level of public spending, and given that, it needs to be funded, and then they make a decision about, okay, what tax revenue do we need, given other sources of revenue, which might be aid in the old days, or increasingly it's not, or it might be resource rents. But there's a decision being made, and you have to understand, well, what's that political decision? Some countries want high levels of public services, they'll tolerate higher levels of tax. Other countries don't. So Korea, who identified it, often comes out as low tax effort. It's not. They're collecting the amount of tax they want to collect to deliver the level of public services they want to deliver. They're not complaining about that. It's a political decision. Similarly, another approach could be you're thinking of a median voter taxpayer. What are they willing to pay in return for whatever they're getting from the state? The other element of tax effort is the administrative effort. The collection efficiency. The challenge in any of the cross-country analysis is you can't really distinguish them. Sometimes you can try and include political or institutional variables, but that's problematic. They're rarely robust. The odd time they might be significant, corruption often turns out to be significant, but what's that telling you? Is it telling you that there's a lack of political effort? Or is it telling you, no, the nature of the politics in that country means there's a preference for low taxes? You don't quite know. That's the difficulty with the general approach, is that you can't distinguish what is a political decision and what is administrative efficiency. Administrative efficiency is the one that we are best equipped to deal with. We've got a lot of increasing ammunition and toolkit for helping tax administrators to improve collection efficiency, but ultimately they're doing that within a constraint of a political environment that makes the decisions about how much tax they want to collect and how much or what extent of tax expenditures broadly defined they want to tolerate. I think the starting point of these effort measures is useful, but it has to be used then to try and say, okay, and this will be my recommendation to the World Bank if you're going to use these to inform any discussions with the country, you should start by going to them, present them with these range of estimates and say, look, this study thinks you're 80%, this study thinks you're 25%. What do you think? Where would you want to be? What do you think your tax revenue to GDP ratio should be? Why? Why do you think it should be that? Is that what you think taxpayers will tolerate? Is that what you think you need to fund your spending? And once you've agreed that and started to understand what the political calculus is for deciding on their tax revenue, then you address the challenge that we're better able to comment on, which is okay, you decided what you want, what type of taxes you want, these are the recommendations for how you can do it better. And we can make recommendations about how you should structure VAT, what you should do for any trade taxes you want to contain, what you need to do on income taxes, what can you do about the hard-to-tax individuals or hard-to-tax companies that is a problem in every country, not just low-income countries, and then what can you do to improve collection efficiency with administrative reforms? So I think as a starting point, it's very useful, but the best way to use it is to present the range of estimates to the policymakers in the country and say, which do you think is right and which do you think you want, and how can we get there? Thank you, and thanks for all the presentations. Okay, we've come to the end of an illuminating discussion, so we'll open the floor now to questions. One question is, would it be possible to have a very simple approach? It's a very unacademic question, actually, where we look at a tax-to-GDP ratio, look at neighbouring countries, perhaps, and then we look at what countries determine, what governments determine as their medium-term collection aspirations. I'm thinking about Ethiopia, which has a revenue collection of roughly 8% of GDP, but based on recommendations by the IMF, actually, in the medium-term revenue strategy, they are talking about 18% of tax-to-GDP ratio. So in that case, we could say the tax effort would have a gap of 10% of GDP. This is not totally serious. I have a couple of more serious questions. Cai, you said that you found that your TRE model is better in estimating or in distinguishing between the efficiency gap and then the residual and then the efficiency gap. What made you think that it is effectively better? That would be a question to you because you didn't elaborate on that. You just said that you found that it would be better. I would like to know how changing variables in your model changes the scores. Is it very sensible to changing variables? My assumption is that it is. Then, of course, the choice of variables that Oliver mentioned as a key issue and also Emilia mentioned as a key issue is key. Emilia, you mentioned the importance of governance indicators, but you also mentioned the endogeneity issue that comes with governance indicators. It would be interesting to see how you approach that. Also, you said that it should not matter if a country is resource rich or not. I'm not sure about this because the literature on the resource course will always tell you, Dutch disease will always tell you that you have to take an inflows of revenues into account in order to avoid Dutch disease issues and to improve diversification. I think we can take one or two more questions before we give the people the opportunity to answer. Rose raised her hand first. Let me first appreciate all the presenters. Very good food for thoughts. My first issue is with this stochastic frontier method. I like the fact that I can see LICs. They seem to be having a higher efficiency score even compared to middle income countries. But the question is, how do you translate that efficiency to higher tax revenue? What is it that we are trying to measure so you are efficient but your tax revenue is low? What is it that we can actually pull out of that efficiency to ensure that LICs are able to enhance their tax revenue? My second question is related to the residential tax base. I like what came out that you look at the element, VAT or other things, the EPSU and only maybe those who are employed. You can keep on going up the ladder to see the tax and who is paying. But I still want to ask the same question I asked in the morning. When you are doing that, how do you determine the progressiveness of tax that you are not necessary of abandoning the same residential tax base by taxing them from various angles? I may just first share some of the comments by the discussant. I think that the very good summary of where we stand with these types of papers and what we can and can't do with them. I like the conclusion that there is a very good entry point to this discussion but we should probably look a bit deeper at what we can do. One thing you picked up on things like including variables like 3.8% of GDP as a good measure of economic activity. To the best of my knowledge, almost none of these models, including the one that we worked on, is able to control for the tax policy environment across countries and across time. A very high trade to GDP figure in an EU country where most of that trade is not subject to customs tax doesn't really tell us much about the power of that for enhancing tax revenue. It's incredibly difficult to bring together all of the information on tax policy structures across countries and over time. Perhaps at some point in time a very sophisticated model might try to do that but we might be going further down a rabbit hole in trying to do that. That is one weakness that exists across most of these sorts of models and it's very pertinent and similar to what you mentioned about not being able to control administrative capacity. Again, that changes over time in across countries but very important, especially in many low income contexts for the ability to actually tax a given tax base. We seem to be reasonably okay at controlling for what that tax base might look like but not great at describing how either from an admin or a policy side we actually go about collecting tax from those bases. Moving on to the two questions. I think Christian's question verges on being an illegal question given Abram's dictation that we shouldn't ask questions about econometrics. But I will do my best. Again with a caveat that our absent co-author was the brains behind the implementation of the true random effects methodology. What makes us think that it performs better? The best way that I can describe it and again apologies for my perhaps simplistic understanding of what we've done is that firstly we did that sort of exercise of understanding what tax effort scores looked like under different models compared to what's going into the models. I only had a very brief time to discuss that example of Slovakia but we saw time and again where countries had extreme values on some of the inputs we were seeing very high or very low scores coming out of the model. For me there's not a more scientific way to describe it other than saying it was a bit suspect. But essentially through various other diagnostic tests within the paper we were able to determine that or we at least suspect that those other models were attributing a lot of what was a random error to tax collection inefficiency. And we think that the true random effects model it's better at disentangling those. It might be that it goes the other way and it's too extreme and actually the scores are too conservative as in too high. But that's the sort of best explanation I can give you and apologies that I can't go in much more depth than that. The second question was I think from Rose saying that well in many low income countries you see that on average they have fairly high tax effort scores and how do we sort of relate that to the fact that they also collect on average quite low levels of tax as a percentage of GDP. My sort of best attempt at explaining that is essentially that you have to think about what goes into the model. And so let's say in the model you have the share of agriculture and GDP and GDP per capita. Let's say there's two variables go into the model. Most of these studies have put a lot more variables in. But when you put in a fairly low GDP per capita and a highly agricultural society you expect that you expect given those inputs that probably your tax ratio it won't be predicted at being very high. So when you come up with quite a good tax effort score that means that given the economic constraints that already exist and have gone in as the inputs you're actually doing not bad at collecting tax from all tax base. That's my sort of most simple way of explaining it to you. Thanks. I'll pass it. Thanks. So thank you so much. I think on Oliver suggestion for the War Bank is actually what we do. We take a range of estimates and then we go deeper in the conversation. And also we look tax by tax and compared to aspiration. So we have this I would say a core diagnostic that's called probably finance review and one is a tax chapter. And this is how we engage in this tax efforts is just a first step to engage with authorities. Which also maybe relates a little bit to the broader question. I think to all of us what you know tax effort should not be calculated as something like tax to GDP vis-à-vis the medium term revenue strategy and plans. Yes. If we think that these medium term strategies actually reflect ambitions, aspirations and they are. Yes. So that's another way to look at it. I think all of us engaging on the tax issues for authorities are taking some of this into consideration. But this medium term strategies tend to change very often. Not so much as they change more than I would say tax collection. So that's another word of caution on the comment. I think I was misunderstood. Maybe I think it's it's it's it's I actually think that one have to control for resources. So actually I agree with that. Maybe I was not clear enough on institutional variables. That's a good question. As I can tell you people use different instruments to collect or to control for this endogeneity. I don't think these are very perfect ones. They are not a good one. What we do we actually I think most of the research that you have seen at this one of the slide do not take them into account for the reasons of this causality and not a good instrument to control to control for the endogeneity. So that's an open question to all the researchers that are here to maybe more advance more on that. On the roster question on the progressivity of the of the approach I presented. Look, I think actually this answer a little bit your concern because what this approach tries to do is to look. As you have in this charge is that for low income earners actually what they would pay is probably a VAT. So you know the progressivity would come we need to come from the VAT tax. But if you then go into the income distribution towards the end of the graph you will see that PAT CAT started to be on the top of it. Which means like if you you're a high income payer you know you will you will you will face the how much all these taxes give add to your progressivity overall tax system. And I agree with you that one can't look into the progressivity of particular tax and very often we don't calculate progressivity of the tax system because it's very complicated exercise. You can do progressivity of PAT or CAT which is also already very difficult for low income countries but very often we don't really do analysis of the progressivity or regressivity of the VAT. So sometimes we advise make your PAT more progressive but we don't know what happens with the VAT so overall we don't look at the tax system. But I think the approach that at least we try to sketch is going to control a little bit for that because it would basically be a system wide design with the progressivity of the overall tax system not a particular tax. But I think we agree on the concept that the details are into making so when we progress more I'm happy to discuss more. I have two points just on trade I agree with Oliver but don't forget that VAT replace tariffs in a lot of developing countries and VAT remains mainly collected at the border. So that may explain why trade remains an explanatory variable of tax efforts. And yeah Ethiopia 8 to 18 I don't know which paper do you look but don't forget I want I resign from the IMF to be free. But don't forget that a lot of IMF paper I have documents are signed by both parties so the authorities may be not serious to improve 8 to 18. But don't forget also that Ethiopia is also a federal country so maybe they are planning to repatriate some local tax or local government tax as a central level. And just finish France managed to reduce his debt from 120% of GDP to 60% in order to go in the match by delegating part of the central debt to SNCF, railway and a lot of. I think two things I would like to comment on. So how to go from high efficiency to higher revenues was one question right. And just to add to what Cale has already said for the DA scores at least it's trivial. It's really two sides of the same coin. So a country with a high efficiency has low untapped potential right. So efficiency of one actually means zero untapped potential although that should be taken with the current of thought because it's basically a low about estimate in terms of untapped potential. Of course even a country on the frontier in the sample there will likely be some scope to improve revenues. It's just there's no other country in the limited number of countries you observe that would push the frontier. So it's a conservative estimate but it's really two sides of the same coin. So for example you have Niger I think on the frontier with very low revenue to GDP ratio but also the weakest enabling conditions and it's found to be efficient right. That just means that there's upside and downside the countries are efficient but that also means it will be very hard for these countries to further increase revenue to GDP ratios even if they are below 10%. Just you know unless these fundamental economic conditions change. And then variable selection of course very important question especially for this nonparametric approach because it doesn't yield any feedback on which variables should be included right. As you get in a regression you get some significance levels etc. So basically we have to build on the insights from the literature that looks at more causal determinants right. But on the other hand then we don't have issues with dealing with endogeneity assuming that the variables we use are the ones we are interested in. It's a simple descriptive measurement exercise. Now in terms of robustness yes we do sensitivity analysis. So what we do is I mean as always with these composite indices you know the weighting scheme and the way to aggregate etc. matters. So what we do is we double the weight of every individual variable going into the input index and we drop each variable at a time. And of course the numbers change and we didn't check for individual countries. My guess is there will be some countries sort of with outliers that will change a lot in the ranking in terms of efficiency. But what we find very robust even when doing these exercises is the general insight aggregate picture that there are lots of low income countries that are very low on achieved levels of revenues but efficiency is actually high. So the overall sort of main results that I showed you are robust for that. And I guess the one reason is that the input variables we use are all highly correlated to a large extent. So even dropping a single one does not completely lead to a completely different picture. Okay we're kind of running out of time so we'll take just a handful of questions. So Peter had a question after which you would give it to the lady. It's lunch time Peter so please keep it short. Okay I'll try to be very quick and I'll ask a very pointed question and I'll make a quick comment. I'm Peter Chawla from UNDESA. Kyle, it's great. Thank you for your presentation. It was very interesting. Dominic presented some changes over time data. I wondered if you guys had looked at that and had any indication that things had changed over time. And I think that would be very interesting to know given the focus that has been on this issue for the last 10 years. Whether you've actually seen changes with your data. And I like the particularly want to ask you because I think the random effects. So I look at this data every year and to try and see where is progress happening in the world. And I can see so much randomness in entering into the data every year. I mean the policy environment doesn't change that much every year in most countries. Tax reforms are slow and yet the revenue data is all over the place if you go in a specific country. It goes up and down by a percentage point of GDP every year. So that's why I want to know when you've separated out the randomness, are you seeing some changes over time? Since I think you're implying that you're separating out more randomness. For Emilia, two quick questions. One, I like this idea that you've got there. I wondered how you treat corporate taxpayers or non-individual taxpayers. And particularly resident ones that are owned, ultimate ownership is outside the country. How would you treat those kind of taxpayers in this kind of model? Because I think that's a really interesting question which also gets to some of the things we talked about this morning in the plenary. And then the threshold for two ports of tax I think is a really interesting question. I would love to hear more about that. Because one of the points that we've made in our joint reports with the bank and others over the time is that you really need to think about that even as you're attacking shadow economy, how do you attack the parts of the shadow economy that are not two-ported tax? Which there are large parts of we think in most countries. So how would you think about incorporating some of that differential between the shadow economy that's two-ported tax and the shadow economy that is not two-ported tax into your model? Thanks. And the final question from the lady. Thank you. Thank you for the presentations. I think the tax effort is becoming more and more... I'm from a revenue administration perspective. And then I think we're all interested in what is the tax gap and what is our effort and can we actually improve on it? So I really have a keen interest on this. And perhaps just then also then linking to the prior speaker who commented on the annual volatility of tax revenues. We've seen it especially before and since COVID. And when we see a financial crisis, you will find this is that there's huge differentiation in your tax collected as a percentage of GDP. And perhaps that could be because we're focusing on cash collections. We're not looking at tax liabilities. So what we found is that you will have more variations if you look at tax collections. So where does that bring us? So if you have like a bottom-up approach rather than like a top-down approach, you will say this as well, this is what the economy has done this year. And this is what the economy has given me. So that will be your first entry point when you analyse how much revenue you've collected. And you will say this as were there any policy changes in that specific year and calculated how much will tax policy estimating to bring you in. So now you've got two legs of your total collection. And the third leg that's missing that Oliver was actually telling us about is your tax administration efficiency. So now you know what the economy has given you. That is a walk-in. It's coming in. You don't do anything. You do your calculations on how much you can expect the economy is going to give you. You've calculated on how much tax policy proposals, any changes in this is bringing you in. And the rest that you're actually collected will then be your tax administration. So how do you measure the efficiency and that's coming back to the lady that has asked the question. Now if you say well this is how much I've collected, how do I improve on that collection. So if you look at your tax administration efficiency you've actually got four pillars to look at. All your registrations done. Everybody filed that should have filed. If you had your payments, if you haven't then there will be debt. And we've seen especially in low and middle income countries that the debt is actually becoming a growing problem for a taxpayer debt. So that impacts on that tax collection that is in all this models on a cash basis. So how's your debt doing and how are you administering your debt. We have seen that if we put more effort into that we can actually increase our tax effort quite considerably. And then lastly will be then your accuracy of your declarations. So that will tell you what is the attitude of your taxpayers declaring accurate declarations and your audit data will give it to them. So from a revenue administration perspective the best you can do is to see what you're doing with your revenue administration. And that will then in the end determine if you take what the economy is giving you for what is happening in the economy, what policy proposals are giving you. Then that is actually been telling you well this is your tax effort and where you can actually increase that tax effort from an administration perspective. Okay. We'll give the presenters one minute each just to wrap it up. And if anybody wants to engage them you can meet them during lunch I guess. Okay thanks once again. Firstly to Peter. So essentially for the paper we did we estimated tax effort over time for I think it's 160 countries. And I didn't talk about it today but we do observe some change over time. And as you sort of meant a lot less change over time than we might have seen from some other approaches. And we've put those scores for the whole time series on the wider website so you can find them under the sort of revenue portal there. And I think those as the distribution is slightly less skewed I think also those scores do change less over time compared to some of the other methods. So sorry I didn't go into it today but like those should be there. So let me pass on. So thanks on the questions for the new approach. I encourage and go to talk more on the corporate tax you know in principle is a hard question to answer. If you assume that firms are owned by people that's an easy one but if you assume that reality is different is. So I guess I will not have more to say and we need to develop it further on the two poor to tax another very hard question. But you know I think then the way how we try to approach this is to look at the very broad definition of income. And then you can maybe detect some of those that don't have income but have assets etc. But another conceptually maybe easy one but actually difficult to measure. So encourage work of us to develop more that concept. I pass it on. Same for me I was also going to donate my one minute to lunchtime. Okay thank you all for coming. We thank the presenters that discuss and thank you.