 Thanks for having me here. As Maria said, I will present this preliminary result about the value-added tax gap in Tanzania. We are a research team with Amina from UNWider, Ezequiel and Oswald from the Tanzania Revenue Authority, Vincent from the University of Tarasalam, and me, as Maria already said. So a brief summary of our result up to now. We gather administrative tax data with out-in data for a specific tax region to estimate the bad gap between the 2014 and the 2019. Up to now, we are doing these in three steps. First, we use the data, okay, to provide preliminary evidence, evidence about the strategic behavior of firms. So basic alert firms look like they declare taxes to be in zones where are less auditing process. And the second step, we follow a bottom up approach to estimate the bad gap. In our preferred estimation, we estimate the gap of 48 percent for the audited tax region, and we compute 53 percent for the whole country. We also see that this estimation is decreasing across time, being more or less stable between 2016 and 2019. And also, we show that this is a lower bone estimation, so if we can have better data about large firms, we expect that those number will rise. And finally, using our prediction, we can see how is the behavior of firms or the evasion across the bad distribution and the sales distribution, showing that firms at the beginning of the bad distribution evade more. For the sales distribution, this happened for small and large-sized firms. And also, we document that firms try to declare more bad purchases in order to pay less taxes, and these produce that they are mimicking smaller firms, and this always could distort either the bad distribution or the sales distribution that we see. A bit of institutional background, the VIT is collected by the DRA, the Tanzania Revenue Authority. Tanzania is divided between tax region and some geographical region, so depending on the geographical region, we can see that more than geographical region for one tax region or vice versa. Regarding taxes, we have zero-rated product, mostly export, exempted product, and the VIT in the mainland Tanzania is 18%. Also, regarding the VIT, if firms have an honorable gross sales larger than 40 millions of Tanzanian chilling, they should be registered as VIT agents. In the case that they have turned over larger than 100 million of Tanzanian chilling, they should have a VIT ID. And in this case, as in many other countries, the VIT is basically sales less purchases taxes, and also firms can ask for coming forward some credit. Regarding audits, this is annual plan by each tax region, and what does this mean is basically could be some more general strategy, but we can interpret it that each region can decide specifically what will they do. This is based on the auditing process in the taxpayer turnover trends and payment, so this is something like a risk assessment estimation. Regarding our data, we have audit data at a firm level between 2018 and 2022. This is where the audited were conducted. And the period that they audited goes between 2013 and 2021. Basically, we have, as I told you, where the audited start and when it's end and the period covered, we have the different type of auditing and the amount recovered per type of taxes. This is important because we can see if some firms have zero VIT but have positive evasion in other taxes, and we can denominate these firms as bad compliers. Regarding the VIT declaration, we have administrative data, so basically we have all the bad forms that firms monthly feel to the TRA. This is between the 2011 and the 2021. And we have the taxes and the untaxed declaration, so we can know, for example, how many of zero-rated sales they are declaring. And also we have gross and some tax paid amount. So to be clear, we have how much of taxes product they sell and how much they pay for taxes for these products. Also, we have information from firms, basically the ID. For those who accomplish the VIT requirement, we also have the VIT ID. We see the tax region, the postal city, business activity, and the industry. So to show you a little bit more of the data, you can see that the red bar of the bar are the audited tax region, the number of firms, so mostly the majority of firms come from the audited tax region. The same happened with the total output and the total input. So basically the audited tax region or the firms there represent a large part of the VIT, the output, et cetera. The rate of audited firms in the audited tax region is around 50%. We also document that audited and non-audited firms bunch around zero bad declaration and the VIT recovered is significant if we consider that the average percentage of firms audited is around the 50%. This is our first preliminary evidence. This is the VIT evasion. You can see in this part in your right across the bad group and in your left across the sales group. What we do with the bad group is basically divide the sample between negative, bunching at zero, and positive, one month before the auditing process. So to not have any bias. You can see that this is the mean of VIT recovered from the auditing process with the confidence interval. Our main conclusion is those firms who are in the bunching at zero group evade more, but the dispersion looks like that we can find similar amount of evasion in all the other groups. So it's not so clear that those firms are who evade more. Regarding the sales group, we divide the sample by the total sales per year. So we have small size, the first 33%, small, the following 33, and large firms the last 33. In this case, it is more obvious that small and large firms evade more than the middle size group. This could be also because some firms are evading certain amount of taxes that they move to the small group. So what we see could be is that some evading firms increase this number. Also, we document that firms with positive and negative VIT declaration are large-sized firms show more auditing process. So if we interpret this as a probability, those sounds are more prone to receive an auditing process. And if we normalize the evasion by sales, we identify that firms that bench at zero VIT declaration on small-sized firms evade more. And for this reason, we said that looks like firms declare in a way strategically to avoid sounds who show more auditing process. Now, our empirical strategy, basically, we estimate this equation for the audited firms. Since we have monthly data, we can impute the average monthly evasion that is detected by auditing. So we have the monthly evasion and we control for all the sales items in the bad forms, all the purchases input in the bad form, the net profits, which is basically total output, less total input. And we control for the bad distribution, sales distribution, date, tax-region, activity, city, on industry. Again, this is only for firms that we have in our data that face an auditing. Later, we use this coefficient to input evasion for those from who didn't receive an auditing process and for firms who receive an auditing process in period that they are not auditing. Using that, we can estimate the VIT gap, which is basically the total amount of evasion over the total amount of taxes that they should pay, which is evasion plus the taxes that they already paid. We did this per year. And to estimate the country VIT gap, we assume that the VIT gap rate between audited tax-region and the country is the same as the tax paid. And to avoid some problems, because of the sign of some audit declaration, particularly when they input credit, we use the absolute values of the tax. So this is our first result. This is our preferred estimation, where we can see that we can estimate the VIT gap between 2014 and 2019. The average VIT gap is 48.5%. And between 2016 and 2019, this become more stable and fall a little bit to 44%. If we don't consider the large taxpayer department, which are the larger firms, because we don't have so many auditing process there, this estimation increase about 20% touch point. And because of that, we said that this is a lower bone estimation, and we are trying to get more data about auditing in this group to improve our results. This is regarding the country. We see that same patterns as before still hold, but now the average VIT gap increase to 53%. And finally, using this evasion prediction, we can see what happened with evasion across, again, the bad distribution and the sale distribution. To normalize this and give a better intuition about that, we normalize bad sales. And we can see that regarding the VIT distribution, evasion is monodinodecrescent, with firms at the beginning of the VIT distribution evading more. So we can suppose that firms that bunch at zero and who have negative VIT declaration are those who are evading more. But more interesting in the sales case, we see a U-shaped form, which means that small sizes firms evade more, but also large. And this is interesting because something is happening in the large firms that they are evading more. So our preliminary conclusion in our preferred estimation, we document that the VIT gap is around 84% for the audit data region and 53% for the whole country. And this is a lower bone estimation. We also document that firms declare more purchases about the VIT to increase evasion. And as I told you before, this produced smaller VIT declaration and firms are mimicking small sizes. What are the policy implications of our results up to now? The first one is that to estimate the VIT gap, we need to consider not only the heterogeneity across the VIT declaration, also we need to see what happened across sales. Something happened in large sizes firms that needs attention and it's not so clear, not only the incentive, also the amount. We can expect that with better data, perhaps this part of evasion will rise a bit more. So this could be something similar that we're talking about, what is documented nowadays with personal income data with leaks. And finally, we also document that firms are a strategic firm. So what is relevant is that they see auditing as random as possible to avoid that they declare strategically and be in some zones that face less auditing process. And finally, to share all of you, what are our next steps? First, we improve the estimation of evasion, basically matching learning and without that econometric method. Later, we want to study the determinant of evasion using some difference in different strategy. And finally, we want to go further in the revenue consequence of this. So thank you very much.