 So this is a paper entitled Taxpayer Responses to Greater Progressivity Evidence from Personal Income Tax Reform in Uganda. This is joint work with two of my wider colleagues, Pierre Atenhooper and Maria Jouste, and then two colleagues from Uganda Revenue Authority, Tina Kaidu and Joseph Okel. So this is a little bit different paper than the three previous one, but it does have a link also to inequality, at least towards the very end. So let me start off with some motivation. So actually when you look at the personal income tax rates in sub-Saharan Africa, they are low. So here's information from a recent study by McNabb and Granger. What it shows is the average top, so the highest marginal tax rate in sub-Saharan African countries versus then the same thing in the OECD countries, either without social security contributions. And without the social security contributions, the average top personal income tax rate in Africa is below 35% or so, which is way lower than people in developed countries face. Whereas of course we know already from these papers as income inequality is at a very high level, and there's very little redistribution taking place by the state. So I don't know, I mean these numbers are probably not visible to you, and Uganda unfortunately is not included, but it's very similar. So this shows from a study by Oliver Parkeen and co-authors how much the state or the government redistributes income via the whole tax and transfer system. And the numbers are very minimal for countries like Zambia, Ghana, Ethiopia and Tanzania, so that the difference between the disposable income and the market income is minimal. So if you take these two things together, the question is would it make sense to raise these low personal income tax rates in sub-Saharan African countries. In order to understand if this is a good idea, we can go back to all the way to the optimal tax analysis initiated by Jim Murlees, and what it says is really that the socially desirable tax rate, other things equal, is relatively high if inequality is the higher level, and society's inequality aversion is strong. And on the other hand the top tax rate should be relatively low if the taxation reduces the tax base to a large extent. This is measured by then the response of the tax base to changes in marginal tax rates. So the key is measured this responsiveness, and that's what we are now aiming to do in this paper for Uganda. And notice that there's very little evidence on how income tax changes influence the tax base in developing economies, especially at the lower end of the distribution. So what do we do in this paper? So we examine the elasticity of taxable earned income on the basis of a policy reform that took place during the fiscal year 2013 in Uganda using the same data that Markus is using. And the reform changed many things, but we are focusing on the top. And at the top, which pertains to the top 1% of income earners, the highest marginal tax rate increased from 30% to 40%. So that's a relatively sizable increase. So we look at the impacts of the reform on the reported revenues and then also on inequality. Taking into account of course in the inequality analysis the behavioural reactions that stem from this reform. So what the reform did was that the tax bracket had been kept the same fixed for many years, so bracket creep inflation being higher than changes which were minimal to the bracket results levels was a phenomenon. And the government wanted to alleviate the tax burden for lower middle income people and then recoup part of the revenue by increasing the tax rate at the top. And the new top tax rate was introduced on persons earning more than 10 million Uganda ceiling a month. And this is very close to the administrative data to the top 1%. So for the whole population it's even a smaller group. So here's what the reform did. So what we have here is that the marginal tax rate pre-reform is the yellow one and then the marginal tax rate post reform is this light gray one and it basically pushed the brackets to the right. So we have done the analysis also for the other groups but really in this presentation we are focusing on the top. So we are comparing those who were affected by the higher marginal tax rate at the top to a control group which is just below, which corresponds to the next 9% in the income distribution in the admin data. Who didn't have any change in the marginal income tax rate, they have a slight reduction in the average tax rates because the marginal tax rates at the lower end were adjusted but it was very small when you move towards the bracket threshold level. All right, so the data set is the same. So it's the universe of the admin tax data from the URA monthly pay return. We also look at the same for the business holders who don't have corporations who pay the personal income tax rate but they were so few affected by the reform that it didn't make any difference to the analysis. So this covers now wage earners. We focus on three years after the reform and that's when our data set starts so we have only two years before the reform. And as already Markus said, although employers have a tax identification number, the individuals necessarily don't have so this is in a sense a cross-sectional analysis which, by the way, isn't necessarily a bad thing according to Emmanuel Saier's survey on elasticity of taxable income research. He actually says that the benefits of the panel data are not clear when it comes to taxable income elasticity analysis because of things related to mean reversion, etc. So he in fact favors using cross-sectional analysis so we are following that approach. We do some sample adjustments to sort of rule out persons who were not affected by this reform. So basically as you can guess, this is a difference in difference approach following the elasticity of taxable income or the ETI literature for short. So we have these pre-reform years and it's quite customary to look at three years after the reform in the literature. Here, in bold, the treatment group we have in this presentation is at the top 1% taxpayers and the control group are those who were right below and roughly speaking they are then the rest in the top 10% admin data. So when the outcome variable is low reported taxable income then the elasticity of taxable income the elasticity can be derived by just dividing the difference in difference estimate by the relative change in the one minus the marginal tax rate so in a sense to the keeping, keep rate of the, which was affected now by this reform. So here's normalized developments of the log mean incomes for these two groups. The treated one is the dashed line with the confidence bands and then the control group is the line here. What you can see here is that after the reform there seems to be some action so that the incomes of those who were now subject to this higher tax rate clearly declined after the reform. When you do the same in a regression framework this is the different estimate and here we report four different versions so we first have a simple different in the first column and the second one is weighted by incomes because that's what people typically do in the taxable income elasticity if there's any heterogeneity in the estimate this should be then the correct one. But then we are worried that they happen to be certain very high incomes before the reform so what we are doing, we are leaving out 1% of the top 1% of incomes when we are saying that these would be then the outliers and if we drop those then the elasticity of the point estimate and the corresponding elasticity which we also report here drop quite significantly. So in a sense what we have as our preferred elasticity estimate is here which is at the ballpark of 0.5 and for reported labour earnings that's way higher than people for developed countries typically estimate so they hover around 0.1, 0.2, 0.3 or so. So the summary of the results so that was what I showed you was for the balance sample of employers the elasticity is much for cut to a... So I'm benefiting you hopefully you are not suffering too much. Elasticity is increase if all taxpayers are used so elasticity is higher among the upper half of the treated group the basic salary less response less than the other income units and there's also suggestive evidence of income shifting taking place because the firms for whom the taxable income for top earners reduced the most also report a greater increase in dividend income after the reform so it could have been that part of the drop in the labour earnings was then compensated by higher dividend income. So what are the revenue implications of this? So obviously there's the mechanical increase on the revenues from the top group because the tax rate went up but we are worried that the behavioral response eats away something about the response but according to our calculations even if you take into account the behavioral response then this increase the mechanical increase dominates so it increased revenue so we can also do we can calculate the revenue maximizing top tax rate and there's the well-known science formula and according to it with the Yukandan parameter values the revenue maximizing tax rate would be around 55 percent and the current one now including consumption tax is around 50 so that meant that the reform didn't increase the tax rate about the revenue maximizing one which is not necessarily optimal so we also care about this top earners welfare but if it doesn't then the revenue maximizing would be the highest that you would get and finally what are the implications for inequality so here we have numbers for the inequality in the whole data set before the reform and after the reform assuming now that there hadn't been any behavioral changes then the actual after the reform and of course now when the top incomes decline then the behavioral response further reinforces the decline in income inequality following the reform so here's the summary we studied the personal income tax reform in Yukandan using the different approach our preferred elasticity is around 0.5 it's certainly greater than for developed economies and we simulate the revenue and inequality impacts and the behavioral response is either way some of the revenue impacts but they reinforce the inequality reduction and that's it, thank you