 Next presentation is going to be given by Jakom Brisco. It's going to tell us about the exporters, battery fund, and tax compliance. Jakom is assistant professor at the University of... Sorry, at the University of Tobingen in Germany. He has an extensive experience working with the public finance and taxation. He's one of the researchers who are giving a lot of information, research findings with acetide data. You are welcome. Presenting this project, so this is a project about VAT refunds. Title is a working title expected to be changing at some point. But this is a project with Marlise who's at UNU wider in South Africa and with Teju who is at UC Irvine. So, let's start by making sure we're all on the same page and we'll know what is a VAT refund. So, VAT refund arises whenever a firm has a lower liability on its output than it has input tax credits and thus is an owed net money from the government. There are a few ways that can happen, one of which is being VAT a destination based tax. Firms face zero percent tax rate on whatever they export but they're still in the VAT system meaning that they get to claim input tax credits and so they sort of naturally would have, especially if they export a lot, they would naturally have negative VAT liabilities. Same goes for a lot of countries have goods that are zero rated even when they're sold domestically in which case the story is very similar and finally you might have years or tax periods in which you have particularly high expenses say because of a big lump investment in which case that can also result in a situation where your input tax credits exceed the liability on your output. Sort of you can think of this as immediate expensive in a corporate income tax system. All of this to say you really really really cannot have a VAT system without having refunds in some way to administer refunds. So this becomes a very relevant aspect of VAT because you might think okay so we have refunds we issue refunds the end except if you just issue refunds that can result in some very massive evasion and here you know sort of the one way to think about this is that usually the amount you can evade on a tax is capped at the amount of tax you owe right whereas we're three funds you know the more input tax credit you claim fraudulent input tax credit you claim the more you can evade without any sort of natural cap. You know Madara Wasim has a paper showing this is happening in Pakistan. I was talking to Wainona who was helpfully providing some anecdotal evidence that it is happening in South Africa but it is happening in sort of richer countries as well. There have been cases of carousel fraud in the EU. You know if you want to spend a fun half an hour I suggest you Google Australia TikTok VAT refunds. I assure you that is going to be entertaining and yet TikTok is in the video platform where you know people upload dance videos or whatever. But so yeah VAT refund fraud is a thing so there are good reasons why you might want to pay special attention to firms that are claiming refunds on the other end you know if you never pay out these refunds right and you just hold on to them forever and ever and you just infinitely scrutinize these firms then you're really giving up a bunch of efficiency aspects of VAT that usually make it favorable to or at least in theory on the textbook make it favorable to other consumption tax systems like say sales taxes or turnover taxes. So that's sort of what we're and then South Africa you know is a particularly interesting institutional setting but let me not sort of linger on that because the point of this study is to sort of look at this end of the spectrum right and ask okay so clearly there are good reasons why revenue agencies are going to want to you know not go too fast with these refunds but what happens when firms experience delays to the behavior of these firms going forward and you know you might be thinking so okay there are clearly reasons why you might expect behavioral impacts of this sort of treatment of VAT refunds by revenue agencies and where are those concentrated and you know one thing we can show you is that the value of refunds is very much concentrated among zero rated sellers right those are the sellers that usually ask for the biggest refunds and indeed if we look at this graph about you know almost 90% of refund claims in value are claimed by sellers of zero rated goods either domestically or abroad but these refund claimers do not make up even half of all refund claimers which might you know might be telling in that you might actually think that the behavioral response is in fact not concentrated among among sellers of zero rated goods but rather other firms that are claiming refunds for other reasons like say lumpy investment and indeed we can see this we can confirm this when we're looking at sort of bunching graphs or just plotting histograms of VAT liability and here we're scaling by output but we can and you know scaling by output because exporter send to be a lot bigger than other firms in the economy but you know we can show you the same graphs and they're not going to make tell a very different story if we're not scaling them we can see that there's quite strong bunching at zero and this is sort of a fairly established fact in the VAT literature that firms would rather not claim a refund just to sort of forgo the headache of increased audit probability etc this does not show up when we're looking at firms that you would expect to be claiming a refund right so if we look at exporters on the left or zero rated domestic sellers on the right we don't see nearly as much bunching as we see in the graph for the total population which is by the way even stronger if we look at only non-zero rated sellers why is there bunching? well like we said you know there are good reasons to expect that there is bunching and here we can show so we have this sort of we actually just today got access to the audit data so you know expect more on this but in the VAT data as we have it right now we have this assessment variable which is sort of a very broad term as we understand goes from SARS followed up with the firm again for because there was a typo in the tax return to there was a very intensive audit but what we can show is that there is some probability of assessment depending on what bin of VAT liability you fall in and that there is strong discontinuities at zero I'm going to come back to this post pre-October 2018 thing at the end here but here I just want to show you there's clear big jumps in the probability that you are assessed if you claim a refund then if you don't of course there's selection here going on it's not random who falls on which side of zero but given the bunching graphs you might all the more expect that the same firm would expect higher audit probability if it crosses the zero threshold so we're currently working on a cleaner causal indication but what we have today to show you is sort of mostly motivational evidence mostly correlational evidence so what we do here is we simply say okay let's say let's create a dummy variable zero before you experience the delay and it's one thereafter so it becomes one if you are ever in the top quartile of waiting times in any given year and then look at what happens to some outcomes so what we can show and then we can control for year fixed effects firm fixed effects we can control for other sort of year variant controls but what we can show is basically after you waited a long time for a refund you are less likely to claim refunds in the future you tend to invest less relative to your total output and you tend to have pay more VAT or at least claim less refunds so we find this very interesting motivational evidence for this one thing you might say is well gee but of course it is endogenous who it is that experiences refunds and there might be selection in this sample what we can show you to reassure you of that is we can play the same game conceptually it's like a parallel trends check we can play this game in reverse we can say okay well if it's just selection what if instead of looking backwards meaning whether you ever experience a refund I construct a variable saying whether you will ever experience a refund in the future and what we can show is all the signs flip relative to the preview stable why well because the mechanical effect here goes in the exact opposite direction if you claim more refunds it's more likely that you're eventually going to experience a delay if you claim more refunds it's more likely you're investing a whole lot etc so now let's get to sort of what we're thinking of using for causal implications so we found that in October 2018 there was a leadership change in South Africa in SARS that led to significant reductions in waiting times for refunds in particular for waiting times in general you can see but for refunds in particular and we are thinking of a nice identification strategy that can exploit that variation as exogenous to firm decisions and I already showed you the probability of assessment here the discontinuity probability of assessment dramatically changed between pre and post October 2018 I can show you here the waiting times and this is in days by the way on the vertical axis dramatically dropped post October 2018 what we don't see is big behavioral responses in terms of claiming refunds by firms that are claiming refunds for the first time so what we're arguing is there's not huge selection who is claiming a refund for the first time firms are just particularly inexperienced refund claimers are sort of going into it thinking well it turns out I have a refund let me ask for a refund but the waiting times dramatically drop almost by two thirds here and still some back and forth with SARS that is going on about exactly what we're measuring here in terms of waiting times but there's clearly something going on right before October 2018 we see average waiting times exceeding a month and after we see them at least halving relative to that standard so this is what we're planning to do we're sort of thinking of doing diff and diff pre post first refund request and we're treated as first refund request happening after October 2018 and control group is first refund request happening before 2018 which is going to allow us to check for parallel trends of course that all still needs to happen but we're very much looking forward to that and to any of your comments so what we saw here is that the value of refunds is dominated by zero rated sellers but the behavioral effects of refunds might be concentrated among in whole other parts of the economy that we see strong bunching among non zero rated sellers but not among zero rated sellers which again confirms that hypothesis and we can show some early evidence suggesting that waiting a long time for a refund decreases investments and decreases the likelihood of claiming a refund and increases your net liability going forward and like I said future work is going to focus on a more causally well causally identified strategy so thank you very much