 Rather than answering my phone, I would rather give the floor now to Marcus Bain, one of our prolific researchers from the financial stability side of the ECB. When we now go back to the forward-looking provisioning that we heard about already in the first session, but this time more recent and made more European. Twenty minutes, Marcus. Yes, thanks a lot for having me here and thanks for the introduction. So I should say that this is joint work with Cyril who presented earlier today already on capital targets. And of course the usual disclaimer applies for us as well. So this is not necessarily the view of the ECB. Yeah, the topic of this talk is credit risk provisioning similar to the first paper that was presented today as already indicated by Philip. And the motivation for this is both general and also specific. And the general motivation is that of course adequate and timely provisioning is very important for banks. So it it covers for expected losses and make sure that they can absorb shocks and and it also creates transparency for both for investors and also for supervisors. And the specific motivation for looking at this now is that it's been a very hotly debated topic in recent years. So first during the pandemic where initially there have been quite some concerns on possible proselytical effects that could be induced by some accounting features and then over time the concerns shifted more towards the adequacy of credit credit risk provisioning. So in the in the latest stages of the pandemic, people got increasingly concerned where the provisions are adequate and there's also this discussion strengthened then after the outbreak of war in Ukraine. So that's one of the issues that we want to assess a bit in this paper, what has happened actually with provisions in recent years. And then the second specific motivation is that there is also a longer standing debate on accounting standards that dates back to the global financial crisis of the summer and eight. Where, in fact, the too little too late provisioning of the previous incurred loss accounting standard was argued to be one of the main drivers of proselyticality during that during that crisis. And in response to this, so-called expected credit loss standards have been implemented, including IFRS nine in Europe, and the idea of the standards was to tackle exactly this proselyticality issues. And that's the second thing that we want to investigate how this has played out in the recent period. So I think you're all more or less familiar with IFRS nine. So I don't want to go too much into detail here, but just to recap the core element, because this is really at the core of the paper. As I said, the main objective of this standard was to front load provisioning to earlier stages of the life of a loan. And this was precisely to avoid these large jumps in provisioning. At the time of default, that have been observed in previous crisis and that are illustrated by the reddish line here in this chart. So this shows the evolution of provisioning for a loan with deteriorating credit quality. And you can see that at the time of default, which is indicated by this red area at the end, these loans exhibit sizable jumps in provisioning. And I have as nine was supposed to address this by basing provisions on expected future credit losses. So to avoid these jumps at the time of default and then avoid the corresponding proselytical consequences, they might trigger. That's the theory of the standard. But then in practice, they have actually rather soon been discussions on whether certain features of this new approach might not themselves trigger proselytical effects as well. And this, I mean, the most prominent that many of you probably have heard is the so-called cliff effect that could occur if exposures are moved to stage two at an early stage of the crisis. So you can see this here on the chart in this blue line at the time when the loan transitions from stage one, so performing status to stage two, underperforming status. There's a switch in the provisioning horizon under IFRS nine. It switches to life cycle expected credit losses. And this triggers an increase in provisioning that would of course put pressure on bank capital ratios, possibly at the worst possible moment at the onset of a shock. Banks might react by adjusting lending and this would have proselytical implications. So to be clear, this is the same effect as under incurred loss accounting. What differs is the timing under incurred loss accounting. This happened rather late and potentially with a larger magnitude on IFRS nine. This could happen earlier and might also trigger proselytical effects. That's the first concern. And then the second relates to the usage of internal provisioning models. So under IFRS nine, banks are relying on their own internal models in order to estimate expected future credit losses. And there have been concerns early on that they might be tempted in order to exploit this discretion that this gives to them because this of course gives a lot of discretion in order to reduce provisioning needs and under estimate expected future losses in order to save on regulatory capital. So this is a bit similar to what has been observed also with internal ratings based capital regulation where also there has been a lot of discussion and and also evidence that banks have been underestimating possible losses with this approach. So these are things that we want to investigate in this paper. And in order to do this, we focus on the period since 2018. So the period that was characterized by both the pandemic and the outbreak of war in Ukraine. And then we use a granular data set from the European credit register. So on a credit to see what happened with provisioning dynamics during this period. And specifically, we compare provisioning dynamics for loans that are using IFRS nine to those for loans that are using national accounting principles. So some banks in the year are still used national accounting standards in order to report the provisions. These are mostly of an incurred loss nature. They also contain some forward looking elements. But what's important for us is that they have been there for a long time. So they they have been there well before the global financial crisis. So they offer a good group with whom we can control dynamics under the new standard. That's one of the cuts that we're looking at. And the second one is bank capitalization. So we also want to see whether dynamics for better and less capitalized banks are different. And the reason for that or looking at that variable is that less capitalized banks might have stronger incentives to be more lenient on provisioning. Because, of course, increasing provisions puts additional pressure on the capital ratios in order to avoid that they might be tempted to delay or avoid provisioning. So, yeah, I mean, that's that's what we're looking at. And we use, as I said, this very granular data set that that allows for sophisticated econometric techniques, including a very granular set of fixed effects that ensures that we can look at provisions by different banks to the for loans to the same firm. So can really systematically control for a borough risk, which of course greatly hates terms of identification. And before going into the details, let me briefly summarize the main findings. So what we see is that some features of IFRS nine seem to be working as intended. We find higher exxon to provisioning as intended by the reform. So there is front loading of provisioning. We also find a more risk sensitive reaction in response to shocks, which is also an objective of the reform. But in terms of dynamics, we actually do not see so much differences between IFRS nine and an end gap lens. So in particular, there's a sizable jump in provisioning at the time of the fold under both standards. And we interpret this as suggesting that in terms of prospectability, the implications of IFRS nine might actually be not so much different from from that of previously existing end gap standards, because as I show you, the dynamics are very similar. And on bank capital, we also find that that bank capital has a strong impact on provisioning practices. So we find that better capitalized banks generally provision more than less capitalized banks, which which is what you would expect if capital management motors have a have a strong role. And we also find that this tends to be stronger under IFRS nine. So and this suggests, as I hinted before, that there might be more discretion under the new standard for banks to discretionary adjust provisions. Okay, that's it in a nutshell. Let me briefly talk about the results in a bit more detail. So I'll show you three sets of regressions here. First determines of provisioning in the whole sample of loans. And then I'll focus on the dynamics as I said that this will be particularly relevant when it comes to deriving conclusions about prospectability. And here we'll look both at EU syncretic credit risk events, so default events and also a correlated credit events or macroeconomic shocks because these are, of course, particularly relevant when it comes to deriving macro conclusions. Well, let me start with a with a whole sample of loans. So this is using the entire sample of 60 million quarterly loan observations that we have in the data set. And our baseline regression simply regresses the provisioning ratio at the bank firm quarter level on the set of fixed effects and the number of bank and loan level variables of interest. So I'm showing only the ones that we are most interested in here on the slide. A dummy variable that indicates whether the loan is NGAP or IFRS accounted and CAPHAT, which is the capital headroom of the bank. So the capital space that is available on top of regulatory requirements. And what you can see here in the first two columns is that, as I said before, NGAP loans tend to have lower provisioning ratios for similar exposures to the same firm. So in other words, IFRS 9 has higher provisioning and this is what was intended by the reform. And the second key finding that is shown in this table is that capital headroom also has a significant impact on provisioning in the sense that better capitalized banks tend to provision more. And again, this is for the same firm in the same period. And we are also controlling for a large number of loan level variables like maturity, protection, all things that are included in unaccredited in order to ensure that these are really similar types of exposures. And you can also see that this effect is prevalent under both IFRS 9 and NGAP, so there is some discretion under both accounting approaches. Okay, let me move to the dynamics. This is the main specification that we are estimating there. So here we focus on the sample of firms that defaulted during our sample period. And then we plot the evolution of provisioning ratios around this default event. And we do this in a regression framework again, but we have the provisioning ratio on the left-hand side, fixed effects on the right-hand side. And then interaction terms between dummy variables that indicate whether the loan is IFRS or NGAP and the time to default. So default is time zero here in this chart. And what you can see, I think, are three main messages. First, the message that IFRS 9 has higher provisioning pre-default is confirmed here. So the red line ahead of the default is always clearly above the blue line. The second message is that there is no real gradual increase in provisioning for IFRS 9 loans. So if you remember the conceptual chart, we should have seen something like this. So a gradual increase in provisioning towards the time when the loan defaults. This is not happening. I mean, the line is rather flat and it's evolving in parallel to the line for NGAP loans. So the bulk of provisioning for both types of approaches really continues to occur at the time of default of the loan. And the last finding is that there's a reversion in provisioning after default. So then here NGAP is above IFRS 9 and this is driven by less capitalized IFRS 9 banks in particular, as I'll show you in a minute. So yeah, I mean, overall the patterns are quite similar between the accounting approaches and this made us wonder what explains this. So why is there no bigger difference between the two approaches? And the reason for this is two-fold. The first is that the timing of moving loans to stage two varies across exporters and tends to occur rather late before the default event. So you can see this illustrated here in the chart on the right, which shows the share of loans in different stages ahead of a default event. And what you can see is that, or actually what you can't see because there's my picture in front, but two quarters ahead of the default, which is the second-last column, there is still 50% of the loans in stage one. So these are all loans that default then and one quarter ahead of default, 35% of the loans are still in stage one. So this illustrates that it seems to be rather difficult for banks, maybe not surprisingly, to clearly identify those loans that will default. And this then explains why a lot of the provisioning still occurs at the time of default. That's one reason and the second reason is that also the loans that are already in stage two exhibit some increase in provisioning before default, but still also for those loans, the bulk of adjustment continues to occur at the time of default. So overall, the change in provisioning patterns were quite similar between the two approaches. IFS-9 did not fundamentally change the patterns and what we think this shows is that it's really the incentives for banks that drive these patterns and not so much what exactly the accounting standard says. So also a model that moves to an expected credit loss approach does not necessarily induce more timely provisioning if the underlying incentives for banks to delay loss recognition remain the same, which is likely the case. And actually the built-in discretion under IFS-9 may even facilitate this type of delay in loss recognition. And this is what we look at in the second part of the paper when we look at capital headroom. So here you see a similar regression as before where this time instead of differentiating between IFS-9 and NGAP, we differentiate between well and less capitalized banks. And the red line is for well-capitalized banks, the blue line for less capitalized banks, and you can see that provisioning for well-capitalized banks around a default event is always above that of provisioning for less capitalized banks. In line with capital management motors and provisioning as much as you can afford strategy as we call it. And this effect is particularly pronounced under IFS-9. Actually for NGAP it's mostly statistically insignificant. We have some specification where this becomes a bit stronger also with NGAP. So there is some discretion also under NGAP, but the effect is stronger for IFS-9. So this approach seems to have enhanced discretion that is available to banks. Okay, I mean in the interest of time I can be very brief on this just to say that of course banks have two levers in order to adjust provisioning needs under IFS-9. One is to adjust the ratio itself for a given exposure in a specific credit risk stage. And then of course they also have flexibility in sorting exposures into stages because there are no fixed rules for that. So we also tested whether the likelihood of moving alone to stage two depends on the level of capitalization and this table shows you that it does. So we find that less capitalized banks are less likely to move exposures to stage two. So they are using both of the levers that are available to them in order to reduce provisioning needs. Okay, and then the last finding that I quickly wanted to talk about is the change in provisioning around the energy price shock that occurred in 2022. So here we look at the change between the second and the first quarter of 2022. And we regressed changes in provisioning on the same variables as before. And we also include an interaction term between our variables of interest and an energy exposure measure at the firm level. So this essentially calculates the degree to which a firm is dependent on energy price related inputs. So the idea is that if you are more dependent on such inputs then of course you are more affected by the energy price shock. And what you can see here in the table is that as the shock occurs, loans that are using the IFRS 9 approach for those loans provisions are adjusted in a more risk sensitive manner. So banks increase provisions more for loans to firms that are more affected by the shock. And this is this risk sensitive reaction to shocks that was also one of the objectives of IFRS 9. And when we look at bank capital we find that again more capitalized banks increase provisioning across the board. So you can see this by looking at this at this coefficient for the for the standalone variable. So if you have if you can afford it you afford it you increase provisions across the board. If you cannot if you have less capital headroom you adjust in more in a more targeted manner and focus in particular on those exposures that are more affected by the by the energy price shock. So the exposure to more energy intense firm and this is this coefficient here at the lower right of the table. Okay, then let me conclude since time is up. So what what what we find in this paper is that IFRS 9 has delivered partly on its objective of fostering more timely and prometer provisioning. So we indeed observe higher exam to precautionary provisioning as intended by the by the reform. But in terms of dynamics the bulk of provisioning the bulk of the adjustment continues to occur at the time of default. Which as I said suggests that in terms of prosecutability maybe the implications are not that different for IFRS 9 and and then get. We also find that there's evidence for for capital management motors playing a strong role in provisioning. And this is particularly strong under IFRS 9 and we are not saying whether this is I mean we're not saying whether this is good or bad because there can be good implications of this and there can be bad implications. If you think of the start of the pandemic we actually told banks or supervisors told banks to use the discretion that is embedded in the IFRS 9 approach to avoid the significant increases in provisioning. Precisely to avoid these possible effects. But over time then of course there might be concerns about possible under provisioning and of course also if this happens too much then it counters the transparency objective of IFRS 9. And lastly on the on the on the on the overall adequacy of current provisions so of course it's difficult to say whether they are adequate or not. We don't want to take a stance on that but what we can say and what our findings suggest is that less capitalized banks might be at greater risk currently of being under provisioned. Also because of built in discretion that is offered by the IFRS 9 approach. And that's it so thanks a lot for the attention and looking forward to the discussion. Thank you. Dustin Kassent is Harry Heisinger from the Tilburg University. In order for the online participants to get prepared I was reminded that you are warmly invited to also ask questions. So I'm willing to give the first question to somebody online if you if you raise your hand or what is the approach to that. And or chat. So thanks to the organizers for inviting me. It's a pleasure to discuss this paper today. So this paper is about provisioning by banks in the Eurozone after the introduction of IFRS 9 in 2018 that introduced a forward looking provisioning in Europe. Now some of the main findings that also I will discuss are as follows. So you always find higher provisioning for banks that apply IFRS rather than NGAP and they find higher provisioning for well capitalized banks but lower provisioning for guaranteed loans. Now overall I think the paper is very interesting and very rich but of course I have to have a few comments. So here are some comments. So one thing I'd like to discuss is the choice of banks whether or not to adopt IFRS 9. The second issue is whether the relationship between provisioning and bank capital whether it just reflects what the bank is doing but or alternatively it also reflects what the firm is doing which would complicate it. Then there's some policy issues that enter into the relationship between capitalization and provisioning some of which I mentioned some of them are not in the paper. So loan guarantees the capital relief under IFRS for COVID period. Then there's supervisory expectations for provisioning for NPLs and there's a supervision by the ECB. It all enters I think the analysis and then there's the issue of the timing of when a loan becomes non-performing which should enter into the discussion of the overall timeliness of provisioning. So in terms of the choice so banks are principal they can manipulate the choice that is whether to be IFRS bank or not taking it to count what will do to their provisioning. Now as the paper explains some banks and those tend to be the smaller banks they can choose either to be IFRS bank or a GAAP bank but that immediately raises a puzzle because it seems to be that if you choose IFRS you only have disadvantages. Of course you have to pay or you have to have higher provisioning that's what they find under IFRS and also if you apply IFRS it's more complicated and therefore more costly. So why would you ever choose IFRS? So this choice I think is unlike the analogous choice that we've seen in different context where banks have to choose between advanced approaches versus standardized approaches in risk rate calculation. But then there is a real trade-off because the advanced approaches are more complicated and costly but they do yield lower risk rates so the same trade-off does not seem to appear here. So that does raise the question whether if you only look at the sample of banks that actually make this choice is it still true that banks that choose IFRS have higher provisioning? It may not be true for this sub-sample. Also I think given the analogy of the two choice sets here maybe the share of the advanced approach that you see in risk rate calculation could be used as an instrument to predict whether a bank or not will use or apply IFRS and could be the alternative to the propensity matching score matching approach which is currently in the paper. Then I also wondered whether you can actually look at this issue of what IFRS 9 does to provisioning while sidestepping the issue of this choice of the accounting standard. So one thing you can do is look at the introduction of IFRS 9 in 2018 and you can look at the same bank before and after and see how the change in the regime affects provisioning. Another avenue might be to realize that for IFRS banks after the introduction they have this transition period during which there is some capital relief that is they can add back to capital some of the increase in provisioning that comes from IFRS 9 relative to the previous alternative. Now this suggests to me that banks during this transition period they have to calculate provisioning under both IFRS 9 and the predecessor. So this would be supervisory information so I don't know whether researchers have access to this. I don't know what the level of aggregation would be. That's just a thought. So then there's the question of what's the firm doing in all this. So essentially the paper looks at the relationship between bank capital and provisioning but the firm may be active here as well. So there's a paper by Chantarelli and what they show is that in the case of Italian firms firms are more eager to service their debt which is debt provided by well-capitalized banks. And the rationale is that if you keep paying your money to well-capitalized banks that's a high chance the bank will still be around so in fact you invest in the bank-firm relationship. Now some of this should be going on here as well, right, that firms will repay more eagerly to well-capitalized banks. So that's true then also just having these firm time fixed effects as in the paper following Kwaji and Mian is not sufficient to control fully for everything that concerns the firm, particular here firm behavior. So the question is whether the data that the authors have and I think the answer is yes, is sufficient to actually try to sort out these two channels. Okay, so now the policy issues. The first is loan guarantees during the COVID period and the authors they estimate a negative relationship between there being a loan guarantee and provisioning. And it's a bit surprising because there's other research for instance by Nicola that shows that banks they tend to provide guaranteed loans to riskier firms. So that suggests that there should be positive relationship between whether you there's a guarantee on the loan and the subsequent provisioning while the authors find here a negative relationship. So why do we find a negative relationship? So one reason if you look at defaulted loans what could be driving it is the supervisory expectations regarding provisioning. It says that for guaranteed for secured loans, which include guaranteed loans. There you can actually take seven years to do full provisioning of unsecured loans. You have to do it in two years. So that suggests less provisioning for defaulted guaranteed loans. Now this does not apply to the non defaulted loans. So then we still have a puzzle and I wonder what's going on here is that the banks they give these guarantees for loans to riskier firms and then they start evergreening these loans. Okay, so you provide guaranteed loans to secure the previous loans and keep doing this because you know in the end you're going to get your money back from the government anyway. So I wonder whether this is an explanation or could be looked at. Then there's the issue of capital relief, which IFRS banks got during the COVID period. So connecting the first paper and the current paper in the sessions. We know that the IFRS here can be very pro cyclical. So if a new crisis arrives, there's a lot of bad news. So if provisioning forward looking, you'll get a huge increase in provisioning and therefore a big reduction in capital. Now during the COVID crisis, what we saw is that are regulated to provide a temporary capital relief only to IFRS banks and not to the banks. So the question is then is whether this type of capital relief, whether it's actually inherent in the selection of IFRS. So do we expect if there's a big crisis, we get this type of capital relief, which means it would be an integral part of all the relationships that are looked at here. So IFRS adoption, provisioning, and capitalization. So that's the question that I throw out here. So then the authors, they look at another shock. This is the second shock, which is the shock of the energy crisis. So the authors like this shock very much because it's a relatively small shock. So this shock was so small that it did not trigger your preferential capital relief for IFRS banks. So it makes it interesting for the researcher, but of course less irrelevant because we only, you know, we more care about the bigger shocks, right? So it leaves me with one question in the room, perhaps, also for the authors. So how big does a shock have to be for the regulator to provide preferential capital relief to IFRS banks? I don't know whether the regulator knows this, whether they have a cutoff, but at least the COVID shock was big enough and the energy shock was not big enough, apparently. Okay, then that's the issue of supervisory expectations regarding provisioning, which I think does enter the relationship here that's being estimated. And as already mentioned, we have these supervisory expectations. They came from 2018. So then what they say is that for unsecured loans, then the bank, they have to fully provision for a loan after default within two years. Well, that's extremely that's what this policy was enacted after the previous NPL crisis. Of course, this is a very stringent requirement to fully provision within two years. So the question is whether it is binding and I suspect it is binding. Whether it would be differentially binding for IFRS versus CARP banks. But if it is binding, then yeah, you don't get the same relationship between capital say and provisioning that's being estimated in the paper. In fact, you know, ecometallically, the relationship, you know, the coefficient capital would be zero because everything is driven by the regular requirement. And that's done, you know, being estimated in the end. So the question is whether this supervisory expectations, whether they leave a trace in the data. I think that that can be looked at whether that's the case. Further issue regarding supervision is supervision by the ECB. So we know that the IFRS banks, they tend to be larger. So they're more likely to be directly supervised by the ECB. Now, then the issue is whether the ECB is actually stricter and more diligent in, you know, triggering provisioning by banks. I don't know what the answer is, but if that's true, it could be that if you don't control for ECB supervision, you know, you get a positive effect of the IFRS regime, which actually should be attributed to ECB supervision. So final substantive issue for me to cover is the timing of the moment at which a loan becomes non-performing. Because the whole issue here is the timeliness of provisioning. So what the authors do, they look at the dynamics of provisioning around the moment of default, which I think is a part of the story. So another important part of the story then is when does the default occur? And I think the reason to expect that the timing of the default will be different under IFRS 9 versus GAAP. We think that IFRS 9 is forward-looking. So you would take into account forward-looking information in addition to backward-looking information, which suggests that you would expect a quicker status of non-performance for loans under IFRS than under GAAP. So I think this would be an easy extension for the authors to look at to also look when they look at time on this as an overall issue, to also look at the timing of when the loan becomes non-performing. So then just to conclude, I think this is a nice interesting paper for which we learn a lot about provisioning in the eurozone, but of course it raises new issues. So Andreas, do we have a question from the online audience or do they let us down? Unfortunately not. There are many people still out there. I promised that they would give the first one, but then all the others are. Normally we never pre-commit, but... So I see we have many questions, so please ask your questions shortly, briefly, concisely. We have a question from Kostas. A point similar to that made by the discussant about bank incentives. Do IFRS provisions at loan inception or early in their lifetime for loans that end up in stage three higher than those loans that remained performing? If the answer is negative, I guess that there are limited gains for the bank in IFRS. Okay, we collect. So maybe I make a tour. So we start with Vito Constancio and then Jean-Édouard. Philip, at the beginning of the discussion, you stimulated us to compare this paper with the paper by José Luis Pedro, who is here and he will speak perhaps better than myself about it. But nevertheless, I think that the conclusion of this paper is that there is no much difference in the prosicality. But the prosicality, as it is seen in this paper, is assessed practically only by looking to the moment when the provisions jump very much. And as indeed that moment is not so much different or the amounts are not so much different at the time of default, then the conclusion is that, you know, they are very similar, so no more prosical. The analysis of the José Luis paper was totally different. It was about the impact on credit behavior and its effects on the real economy. So it's a different approach to the concept of prosicality. And this is not done here. So we should not indeed conclude that the outcome or the takeaway of the two papers is different in what regards prosicality because they are using two different concepts of prosicality. And the question is, add you the opportunity of also testing the effects on credit and what, in that case, were the results. Okay, let's continue collecting. I'm afraid, Andreas, there are so many questions we would be slightly over time, but not too much because I have to go all the way around. Okay, I like the paper a lot. I like the empirical results. It's just a comment about how to interpret this because I will give you my interpretation. You can just tell me whether you agree or not. So I think the point of IFRS in general has nothing to do with regulation is to give you a better view to the creditors and the shareholders. And so it's logical that as we improve these accounting standards, they are going to reflect better and better the position of banks and they are going to be more and more proscical. And so I think there is a temptation in regulation to try to go against this movement and I think we should not. But if we think that they become too proscical, then we should have more counter-scyclical buffers and these kind of tools instead of tweaking accounting standards. And one big tweak we have is in the way we treat whole-to-maturity securities and there is a big bank stock sell-off in the U.S. right as we speak. And that's pretty much due to that. So I think it's a very, very interesting question in general, this interaction between accounting standards and regulation. So that's just a comment from me. Thank you. I wasn't sure who was next. Hi, Joseph Meichnitz from the Austrian Central Bank. A quick one, you mentioned that banks still have a lot of discretion at their hands when it comes to stage two and free migration. And I was wondering whether the same is true for countries. Did you have a look at the country effect? Because I remember when the pandemic first hit, we saw stage two migration just in the past. A few countries and I was wondering whether there is a strategic element to that and whether that was just a one-off thing or whether you've seen that in the data as a structural effect. Thanks. I saw Ernest and then, oh, we are a bit running out of time, unfortunately. You showed a chart of 40% of loans are actually recognized at default only at the end. So at the end of provision before. So I was wondering if that chart is before the COVID period or not, because at COVID, or if this is a standard average number that comes up with the loans. Because at COVID, you might have a lot of firms defaulting all of a sudden, sudden death. That's why banks, some firms, I don't know, because Anacredit is also small firms, a lot of small firms in Anacredit, micro firms. So if that chart was before COVID, after COVID, how these numbers change? It was 40% of recognized defaults at the end. Sorry. I'll be super quick. Okay. The marginal cost will be low. More to benefit the huge. Are there tax consequences of adopting IRFS 9? And if so, could you use that as an instrument for the choice whether to adopt it or not? Sorry. So pair. Last one. So I was going to be very quick, just on the point about lowering the risk weights for energy firms, right? That wasn't really the debate, right? The debate at the time, which is not, I think nobody in this room, just to be clear, is the commission wanted to possibly make Malinian rules for collateral posting to CCPs. So they wanted to make it easier for banks to give loans for posting to CCPs only. There was never any debate to lower the risk weights for loans or other things as such to energy firms, right? So that wasn't really, and I can say second comment, I think at the supervisory table, there was very little appetite, which was also reflected in a number of statements made to the commission that, you know, please don't make this a banking problem when you've got bad energy firms, right? So I think just to make that, make that point. Yeah. Okay. No, thanks a lot. First of all, for the very good discussion and for all the interest and the questions in the paper. So I think we are very glad that there's so much interest. Let me pick up a couple of points. So maybe on the differences between IFRS 9 and ANGAP because some people ask questions about that. So indeed, I mean, it's not completely free choice whether or not a bank adopts IFRS 9. In fact, if you're a publicly listed company, then for the highest level and reporting at the highest level of consolidation, then IFRS 9 is mandatory. And only those operating that are not listed or reporting at sub-consolidated level can also choose to report under ANGAP. So there is a choice here. There is some selection definitely going on. And we addressed this in several ways first by conducting a propensity score matching approach that ensures that entities are still somewhat comparable. And then more importantly, in Anacredit, there's also reporting by the same bank under different types of approaches. So if a certain sub-consolidated entity is part of a larger group that has adopted IFRS 9, there may still be occasions where that sub or solo entity actually reports under ANGAP in Anacredit. So that offers us a nice chance to focus on those banks that actually report under both approaches in the dataset. And then see whether our effects occur also for that subset of firms in order to address this issue of choice. Because then we're looking at very similar banks and we find that also in that subset actually there is still the same pattern or the same findings emerge essentially. On the IFRS 9 introduction, unfortunately, we cannot look at that because we do not have data before 2018. So Anacredit starts only in 2018. But I agree that on the transition, I mean you're right that essentially the banks need to calculate the provisions under both approaches. We don't have this data. If it's available somewhere here on the supervisory side because we are on the financial stability side, there's always a bit of a data sharing problem. So if it's here in this building somewhere and you are willing to share that with us, then we would gladly look at that as well. On the loan guarantees and the negative effect there. So maybe one of the reasons that could explain why we find something a bit different there is that we are also looking at the same firm in the same period. So we include firm quarter fixed effects. So in that sense, I mean you're right that banks might use guarantees more for loans to risky affirms but then controlling for that and looking at the same firm. I think it's plausible that you would then expect that those that use a guarantee might actually have lower provisioning attached to it. Yeah, on the support measures so that's also key aspect of course. So during the pandemic there were actually two types of support measures. Banks were allowed to add back certain parts of the provisions to their capital ratios. And there was supervisory guidance to be not too pro cyclical in the assumptions used for the models. And we thought a bit about this and what this would do to the results and how it would affect the results on the adbex. It's actually uproarly not clear how that would affect our results because it could also induce a bias in the other direction because if the banks know that they can add back the provisions to their regulatory capital, they might have additional incentives to provision more under IFRS 9 and this is not what we observe. For the guidance this is not really I mean this didn't really constitute a challenge for us because essentially what supervisors have told banks was applied a standard. So they just have told them look there is a lot of discretion in IFRS 9. You don't have to increase it very strongly after the shock and make use of that in order to avoid very positive increases. And so this is essentially the test that we are the thing that we are investigating how this has played out. Yeah, maybe just some questions on the on some points on the questions from the audience. So on the comparison with with paper by Jose Luis, I fully agree with you that it's a different thing. So the approach is different. The country is different. Also, this paper is not focused on IFRS 9 specifically. Therefore, and I also think that many of the findings are consistent. So both papers find a strong impact of bank capital, for example. So therefore I am not concerned at all. Looking at credit is indeed an interesting suggestion. So I think we will definitely do this. I mean, we're anyway playing to do this for the period since the pandemic. What we cannot do is do analysis like Jose Luis where we test the effects before and after because we don't have the before reform data. Okay, then there were a lot of specific questions on country effects, whether we also looked at country effects. So we haven't done so for now, but it's also a good suggestion on the on the share of exposure. So that was indeed for the whole sample of of loans. So from the period for the period from 2018 and until 2023. But we don't find that this differs a lot. And on the tax consequences. I don't I don't know I have to admit, but it's something that we can that we can also look into. And finally on it on the energy firms. So here the idea was not that we that we test for adjustments in risk rates, but more that the economic argument was that we would expect a stronger impact. On expected future losses for firms that rely more on energy related inputs. So firms that purchase more from electricity or other energy related sectors. So we would expect a stronger increase in provisions for these type of firms. And this is what what indeed we observe or for IFRS my loans. So I would stop here. Thanks a lot for all the comments and happy to continue afterwards.