 Good, so good afternoon everybody and From those who are listening in virtually from across Atlantic good morning or good late evening For those in Asia. My name is Klaus still man. I'm head of division in the directorate Supervisey strategy and risk of ECB banking supervision. Please. Let me make upfront a few housekeeping announcements We have in this session three papers and we have allocated its In the following way we have 20 minutes for the author of the paper to present then 10 minutes for the discussant and Then 10 minutes for discussion We take questions from those who are attending from this room and also from those who have joined online In order to help presenters and discussants to stick to the timeline Andreas will help us by waving warning signals when you're coming close to the end I may ask those who are here on the table that We need to press the button for the microphone because otherwise the people who have virtually joined will not hear us Keep this in mind now Let's move to the content of the first session. What is a common theme in the first session one way to look at it is to See that they contribute to assessing in different ways the effectiveness of the regulatory response to the global financial crisis The first paper looks at the forward-looking Provisioning approach the authors seen new provisioning roads in Colombia as an anticipation of the Expected credit loss framework as it was called as it was introduced by the IFRIS 9 This is work presented by Rosiluiz Pedro from Imperial College and Discussed by Skanda van der Heuvel from the Federal Reserve Board The second paper explores internal capital targets of banks and their interaction Both with the use of capital buffers and with lending supply also for different characteristics of banks Is presented by Cyril Qualier from the ECB and discussed by Florian Haider from Goethe University The third paper Inspects the impact of the ECB different recommendation on bank lending This represented by Ernest Dautovich from the ECB and discussed by Dejana Bonfin from Banco de Portugal and University Catolica Lisbon The second and third paper also have in common that they use the COVID-19 pandemic as a test case for the analysis So they also contribute to the corresponding strand of literature that is still increasing The key variables considered in all three papers First one credit provisions, second capital targets, third one bank payouts Also key variables for supervisors to carefully monitor and assess in ongoing supervision So the policy relevance is clear and needs I think no further Motivation so that in the interest of time, I'm happy to give the floor to Rosiluiz for the first paper Thanks a lot, thanks for the invitation So this paper is a quote from Bernardo Morais from the Federal Reserve, Gaifka from ESE Monica from Pennsylvania State and Miguel from the Banco de la República of Colombia So loan of provisions is the highest accrual for banks There has been until recently the incurred credit losses in which basically it resulted in two little provisions too late Luke has a very famous paper in 2004 if I remember correctly on these issues And basically the problem of this type of approach is that then in the middle of a downturn banks will have to increase very fast the specific provisions and this might create potentially a credit crunch in the economy and lower profitability and capital for banks So there has been introduced both in the United States and in Europe from 2018 to 2020 The expected credit losses and the big difference between the expected credit losses and incurred credit losses is that for the origination of loans which are still below 90 days of delinquencies if you want and they're performing and they're incurred credit losses, the loan allowances were zero and the expected credit losses they are positive This is the big change, so the big change is that when you originate a lot alone you are already thinking about future losses and you start the provisioning already there Now this is very important for accounting but it's also very important for policymakers and supervisors because in a sense this might create, this potentially might reduce the prosyclical quality in the banking sector and at the same time to better align risk taking incentives for banks As Andrea was saying before risk management has been one of the key issues in the bank failures in the United States Now there was the COVID-19 crisis, so one could have seen these issues in the COVID-19 crisis but they allow the supervisors, they allow banks to defer the introduction of the ECL and you know at the end it was a description of the banks to use it or not to use it And as I was saying before this is very important not just for policymakers for accounting So accountants think that this is the biggest reform even more than IFRS Now we check in Colombia, why Colombia? The first thing in Colombia is that this was introduced more than 15 years ago I will explain you exactly what it was introduced, it was more forward-looking provisions So in the sense that we, you know, loans at Origination or loans that are underperforming So below the 90 days or loans at Origination they have, you know, have to provision like in the expected credit laws And then just, this is an academic paper for identification, just for the issue about identification empirical identification, there is some discontinuity, a particular discontinuity which affirms, you know, which are very similar in all the things but they are very, just there is a random number of 50, sorry, of 2 million pesos Colombian pesos cop, around 2 billion, then you will have more or less provisions Just because smaller firms tend to have a higher probability of default than larger firms And there was just a particular rule and this rule is helpful for us just for the identification Basically, if you were a firm below these 2 billion, then, you know, even at the beginning with no arrears at all you would have to provision and you would have to provision more than firms which are a bit larger And this is just introduced, so before these, if you were not in arrears, you wouldn't have to provision at all So it was zero, okay, so it's the introduction of the provisioning for expected losses and a bit higher for smaller firms So what are the preview of the main results? The preview of, so I'm going to tell you two types of results, one more for policy makers and then the results of the paper So the preview of the results is, the first one is that these changes in provisioning implies a reduction in credit to firms These will have an associated reduction in real efforts for the firms And we are comparing firms which are extremely similar around this continuity, but they have more provision and less provisioning due to the rule The second thing which starts being very important for policy is that these effects are even stronger in bad times So the issue of prosyclicality, reduction in prosyclicality, we don't see it there The evidence of the mechanism is that this is truly through an increase in provisioning between the different type of firms and before and after And then there is going to be a complementarity with capital, these effects will be stronger for banks that have less capital Then there is regulatory arbitrage, you might have rules, and this is a question I was posing to Andrea So you might have rules, in this case regulations or accounting, but there might be an issue which is, you know, maybe banks react through unintended consequences And here the unintended consequences is that now the banks are making less profit from these loans So what we find is that they will do a search for yield within the high provisioning, within the loans to high provisioning firms They will try to, within these, they will try to go for a higher search for yield, especially banks with less capital And this will end up in more concentration and higher spot defaults through these unintended consequences So the key policy message of the paper is the first one is, I think it's important, there is no clear reduction in prosyclicality with forward looking provisions The effects are even stronger in bad times Second thing, there is a complementarity between provisioning and bank capital Of course, provisioning and capital, they are different in the sense that the effect profits potentially taxed differently One is more for expected losses versus unexpected losses And you can have part of the provisioning tier two and the other is tier one Nevertheless, both of them are coming from the shareholder funds So what we find is that the negative impact on credit on the higher provisioning is enhanced if the bank has less capital In fact, here in the string case, in this one here, there is this reduction in credit in bad times is zero With the new provisioning of forward looking if banks have enough capital, if they have high capital levels So in a sense, this is something that are not independent, they are complementary these policies The last point on bank supervision for this regulatory arbitrage is that there are unintended risk taking So the other point which is important, apart from the prosyclicality or the complementarity between bank capital and provisioning Is that you want them to introduce these reforms in order to have better risk taking And what we find is that not surprisingly banks adjust in the margins in which they can take more risk and they get higher yield Within the staff that they have to provision similarly, and this is what we find in the data This is an example of risk taking but exposed, it has more concentrated effects And you know, concentration is important, we saw it in the Silicon Valley bank just to put an example Or they have higher exposed loan difference And basically in these eight minutes that I spoke, I just gave you like the big overview of the paper Now the rest of the times are just details So the first thing is that we use credit register data much with firm and bank supervisory data The provisioning law is from May 2007 They affect firms in different ways and we are not going to exploit all these things But because in a sense our paper is academic and we want to exploit something which is a bit better on identification So we want to compare firms which are similar but they have these different provisioning And we will use this exogenous arbitrary threshold of two billion copes Also, we will have a long period of time, so we will have a crisis before and a crisis afterwards Like recession before, recession after And so we're going to see whether the implications of prosyclicality in recession times are different with the provisioning scheme of forward-looking provisioning Versus the all-in-care credit losses And the data goes from 2001 to 2016 So now you will see lots of stuff, but let me just go a bit So now you will see around the provisioning So these are firms which are more affected, which are the red ones, less firms which are less affected The provisioning of these firms just before the introduction were similar And after the introduction, these firms below the threshold, they have to provision more and indeed they provision more These are tables on the left-hand side is long volume And so we find that around the threshold, remember the threshold is two billion So once after the introduction of the provisions, firms that have to provision more, what they will see is a reduction in credit This is not a question of smaller versus a bit larger firms, they are very close to around the threshold and we can be as close as possible to the threshold But if you see thresholds which are lower, so this is from 1.5 to 2.5, this is from 0.5 to 1.5, you don't see anything And these are smaller firms, or if you see thresholds above, you don't see anything So it's just around the threshold in which some firms they have to provision more than others Is then bands will react by cutting credit to those firms This will have negative real effects, it's not only a reduction in credit But those firms around the threshold that they have to provision more, they will have not only less credit volume, lower liabilities, lower sales, lower assets, and lower investment So there will be some negative real effects at the front level As you can see here, I cannot close because I need the micro, but as you can see here Everything comes just after the introduction of the regulation, both the provisioning, both the reduction in lending And you know, it's not only a reduction in lending, these firms are less, you know, you have to provision more So not only you give them less lending, but you increase the interest rates and you reduce the maturity Reduction in maturity is another way of tightening lending standards And so on, this is immediately So this is the first part of the paper, short windows, but we want to analyze crisis times, so economic cycles And we have, you know, like recessions before in the sample in the global financial crisis and even later Even before in the credit sample, the sovereign CDS of Colombia was relatively high and unemployment was also high So in a sense we have, you know, bad moments before the introduction of the policy, the policy was introduced in May 2007 Then we will have short unemployment in the global financial crisis And then, so once we analyze the cycle, now as you can see here, so this is what I showed you before After the introduction of the policy, firms that are more treated, they have to provision more There will be a reduction in lending for those firms, and this effect will be stronger when there is a weak GDP Or when there is a recession And again, all this only happens around this threshold of firms, which are very similar, but one have to be provisioning more than the other But it doesn't happen to other type of thresholds, arbitrary thresholds, in which all the firms have similar provisions There will be also associated real effects, I don't need to go to all of these things, but what you will see right now on credit It will also happen for real effects in the sense of liability, sales, assets, investment and survival of the firm This is, everything is related to provisions, everything is related to provisions, meaning like you have to provision more You know, for these firms that have to provision more, indeed they provision more, and especially in bad times So this is the problem of the prosyclicality, so this is better credit loss for banks to provision more in crisis times And so the prosyclicality that you know was at the idea of reducing prosyclicality, and we saw this in the COVID-19 20 That's why I know you guys here stopped, I mean, you kind of watered down this type of provisioning, both in US and in Europe Because you were afraid, you know, you were afraid, and also there was a substitution with capital, no? Capital buffers went down So there was a watered down of this, and it's precisely because of this, because they do provision more in crisis times When there is a weak GDP Then we have a result on bank capital and provisioning, which is also the second type of result, which is important So as you can see what I was telling you before, that they have to provision more, the firms that have more treatment There will be a reduction in credit because of the more provisioning This is somewhat reduced, but not eliminated, but somewhat reduced if the banks have more capital Again, only this happens around the threshold of being more or less provisioning more or less, but it doesn't happen in other firms Now what happens in the crisis times? In the crisis times, it's the same thing And now there are many tables and many rows, and I'm very used to read these things, and it's very difficult for you I show you many tables, it's very difficult, but just trust me For instance, in this one there is a reduction in credit in crisis times for the firms that have to provision more after the introduction of the provision But this coefficient is completely killed if banks have more capital So basically what this is saying is that these negative effects of provisioning on the reduction in credit in crisis times could be completely mitigated if banks have enough capital Because remember, there are many differences between provisioning and capital, and I explain this to my students But at the end of the day, both come from shareholder funds, so in a sense if you have a lot of capital, then you can, you know, the increase in provisioning Even in recession times will not imply a reduction in credit if banks are highly capitalized, this is what we find In the final thing, I have four minutes, so regulatory arbitration Now, these firms are making less profits, provisions imply a reduction in profits automatically, everything else comes Now, the time when you lend to firms that require more provisioning, more because of the forward-looking provisions, the expected credit losses Then you are making even less profits in those firms, so how do banks are going to react sometimes? They are going to react by going to risker borrowers, not because of the sake of taking more risk, but because they can search for a higher yield And this is what we find here, so, you know, again as before, if you, these firms that you have to provision more, you will cut lending to those firms But you see, in these firms, you will completely upset that by taking to the highest yield firms with the higher provisioning With the treated firms that you have to provision more, on the high yield firms, exposed, you will start lending more to those firms And you completely kill the effect So in a sense, you have to provision, you will cut lending, but within the high provisioning firms, you still want to make the old profits So you will go for the risker ones because they give you an exante yield And all these things are driven by lower capitalized banks, in higher capitalized banks, you don't have these things Again, another complementarity between bank capital and provisioning And this only happens around the firms in the reform, sorry, around the firms related to the reform, but not in the other type of firms In which basically these ones are placebo tests, in which all the firms are similarly, so the level of provision to these firms is similarly changed And the final table, there are already many tables, let me just give you the headline result on the top This exante is taking, and these unintended consequences imply that these banks that they do these, they end up with higher concentration of loans And with more response at default, so you do care as a supervisor of these things So let me, in the last two slides, just go back to the results And let me just pay more attention to this one, to the main policy messages So the first thing, and this is in a sense what we kind of saw in the COVID But in the COVID there was endogeneity of banks taking this provisioning, and then there were the regulators softening this provisioning So these forward looking provisions, they're going to increase the provisions in recessions, because there is much more risk at that time And this will lead to a reduction in credit, so the prosyclicality will be even worse Now, this is what we find Second thing, there is this complementarity between bank capital and provisioning Meaning that despite of many differences across the two Remember profits, taxes, expected losses versus unexpected losses, TR1 and potentially TR2 Both come from surholder funds And this implies that the reduction in credit due to higher provisioning will be reduced if you have more capital Because you have more surholder funds In the string case, which is very powerful, in the string case, this reduction in prosyclicality, this increase in prosyclicality This reduction, stronger reduction due to the provisioning in bad times will be completely mute if banks have high levels of capital That will be, you know, like, you cannot treat these reforms independently They might be complementary, and our evidence, both in good times and in bad times, is that they are complementary And the final point is, supervision is important You might have rules, you might have rules, but you need the supervision over and about the rules Why? Because banks, they will adjust on the margins, they will adjust to the old times I always put this slide on my students, on this book by Tomaso del Ampedusa, Ilgato Pardo They will always try to find ways to have the same privileges as before In that case, in the book, it's about aristocracy in Italy, in this case, it's about getting profits And this might not imply anything, this higher standard is taking, but exposed, they end up with more risk, both in the concentration And in the exposed loan defaults And the final slide Thanks, thanks a lot And the discussant of the paper is Skanda von den Hövel from the Federal Reserve Board Thank you, Luke It's a pleasure to be here, thanks for inviting me to discuss this paper And it's so nice to see so many familiar faces Oh, here we go, I have to just say that the standard disclaimer applies, these are just my own views since I'm at the Federal Reserve Board So the main message of this paper, the way I took it at least, was that it really calls into question the desirability of the expected credit loss provisioning And including the notion that ECL, as it's known, can reduce prosyclicality So it really raises some questions about that More specifically, it finds that the introduction of forward-looking provisioning in Colombia in 2007 had the following effects It led to higher provisioning and higher loan interest rates It led to declines in credit and economic activity at more affected firms, as Jose Luis explained And moreover, these effects were more pronounced in times of economic stress And for banks with less capital And there's even some evidence that these banks searched for yield within groups of firms lending more to riskier firms, higher yields So, you know, if incurred credit loss provisioning is too little too late, as some have claimed It begs the question whether ECL is then too much and too prosyclical Or still too prosyclical So that's sort of what I thought of as the main takeaway from this paper Just a little bit more detail and then I'll give some further comments So the setting, as Jose Luis explained, is the 2007 Colombian reform that replaced ICL and incurred credit loss provisioning With forward-looking provisioning, which is, you know, if you look at the details, it's similar to an ECL approach Expected credit loss, that seems to have been also the intent of the reform Under the new model, you have these prescribed probabilities of defaults, PDs, and loss-given defaults, or LGDs And they depend on some variables that you'd expect, the number of days past due, the collateral But also borrower size, and, you know, that will come back in a moment Overall, the reform raised required provisions, in addition to making them more forward-looking Most of the causal identification is based on a somewhat arbitrary size threshold And so the paper in a lot of the exercises compares firms just below the threshold and just above the threshold Right? And arguably these firms shouldn't be that dissimilar because they're just below and just above And then you can identify the causal effects, even if there's other things going on at the same time in 2007 Which can always be the case And so the key thing there is that post-reform, there were higher provisions for loans, two firms with less than 2 billion in assets So overall common, so, you know, the paper has a compelling identification of the effects of the reforms It's really, in a sense, the reform is looking at the joint effects of higher provisions as well as forward-looking provisions Now you might say ECL is always going to entail higher provisions I don't know if the Colombian reform was particularly harsh on that, particularly stringent, I should say The paper is very well executed, placebo tests, the right fixed effects, right standard errors, so at least I thought it was very well done It also has the benefit it can show the persistence of the effects that it finds and that's because the reform was already a while ago Whereas IFRS 9 and CISEL were much more recent, so papers looking at that can't go that far out The impacts are economically large as well as statistically significant And you know, maybe one question one has about papers that look at a particular historical episode We should always ask, you know, what's the external validity? I'd like to maybe know a little bit more about what's special about the Colombian banking system Or maybe the reform compared to CISEL and IRFS 9, how much can we extrapolate? So in the remainder of my comments I'm going to talk a little bit more about the last two points, they shouldn't detract that I really like the paper But I'm actually going to take the results really seriously and see, you know, try to put them together by looking at the economic magnitudes Okay, so let me try to do this in five minutes So the reform raised the prescribed PDs, probability of defaults, at origination from zero to 1.1% for larger firms or 1.6% for smaller firms Okay, so the difference is 50 basis points, right? That's where the paper is going to get most of its identification You'd expect the differential effect on provisioning to be less than 50 basis points, right? Because LGD is going to be less than 100%, LGD depends on collateral, other things, but sort of in general not more than 50 basis points The difference in provisions observed between the firm groups is 1 percentage point, excuse me Or 1.4 percentage point depending on whether you're just looking at straight up difference or whether you're controlling for other loan terms So it's the same ballpark, but it is substantially more and so it made me wonder where that's coming from It could be that there's a number of outstanding loans because the paper, the effects in the paper are not just on loans at origination but the whole portfolio of each bank And so maybe there's some outstanding loans that are between 90 and 150 days past due where the difference between the pre and post reform and the two groups is larger So is that the reason or is it something else the paper could check I think and it would be nice to see Going further along and this just shows the effect of provisions for the two groups which Jose Luis also showed this picture Going further in the economic mechanism, once you know provisioning the next link in the causal change is to look at the lending rate So doing a really crude back of the envelope calculation, 1.4 percent higher provisioning rate if we just take that as given Should raise the lending rate by 0.01 four times the required excess return on equity right how much more expensive is equity relative to other bank funding costs So if that's 10 percent which I think is a high number then it should raise the loan rates by 14 basis points If it's 20 percent let's say it's high in Columbia then it would be double that What do they find they find that the lending rate goes up by 0.6 to 0.8 percentage points So maybe against same roughly ballpark but a substantial amplification that we see and so I was wondering you know where does that come from is that evidence for financial frictions Maybe it's really hard for Colombian banks to raise equity that could well be although the effects appear to be persistent So you just kind of expect that to over time they'd be able to retain earnings and solve that problem And then going one step further again even if we take this differential effects on interest rates as given so 0.8 percentage points on an average loan interest rate of 18 percent We see a really big impact on lending volumes that impact is about 24 to 31 percent depending on the specification So normal elasticity's where CNI loans in the US are two and so you'd expect something much smaller you know in the order of 2 percent What's going on there I'd really like to know more what's going on there it's not substitution because if you aggregate at the firm level you still get this result And it has real effect these at these firms total assets go down sales go down by big numbers like 20 percent It made me wonder if there's perhaps something going on like a switch from lending to subsidiaries to parents could that be what's going on I don't know enough about the Colombian economy to know whether that's a factor or not Maybe it's just credit rationing that banks just cut off credit in addition to raising loan interest rates to the more affected firms So I have some other comments but in the interest of time I'll stop here and and do the remaining ones bilaterally Yeah, I think so it's kind of I would like to collect first questions and then ask the author to address them all in one go Three quick remarks I would like to do we collect questions first from those in the room and then from those who have joined on joint online I will take only three questions but I see only three hands and the rest we need to do afterwards I guess Please speak your first first your name and affiliation when asking a question and for those in the room please wait until the microphone has arrived so that other people online can can hear you So we start from the left to the right the left hand side Paul if you could come Hi, thank you. It's very interesting presentation. It's an atolli Segura from the Bank of Italy a question related to the standard validity When it comes to to the discussion of potential prosyclicality of IFRS 9 that we had here in the Euro area and have been also some academic contributions The issue was regarding the reclassification from stage one to stage two at the beginning of a recession that was forcing the banks to to estimate the expected losses from one year to the entire lifetime of the of the loans So my question is is that feature present in the in the example you are you are considering and if not I guess the the the implication would be is that if anything IFRS 9 would be more Procyclical than what you find in the Colombian case. Thank you. Thank you. Next is Peter in the last row, please. Peter Raupach, Deutsche Bundesbank just for a better understanding having seen these huge loan rates. Are we here in an environment of extremely high default risk or is it rather just nominal rates and We have a high inflation. So if we are in the in the first variant, then I wonder whether this is let's call it a healthy cleansing effect or is it just other figures than the ones we are used to. Thank you. So in the right hand side, Jacob. Thank you. Jacob from the B.A. Hi, so Luis. So I was going to say a nice paper right I guess I have a slightly different take, which is that if if what you're really looking at is banks seeing that what they see as a request for over capitalizing because the machine changes that make that it makes a lot of sense to what you're seeing is that the banks fight back and try to avoid the over capitalization by taking on more risk. And I particularly like your the last slide which is actually they undo each other perfectly right which to me is like, let's say a standard Lucas critique finding right that actually you change the rules you should expect the system to respond. Aside from all the other stuff I think that will be my interpretation but I'd like to hear you read if you think I'm way off or where I am. Thank you. Thank you. Any questions from our online community. If not then we can also go. No, nothing. I have one question. Costas. He was wondering if the paper distinguishes between the persistent prosyclicality of addressing forward looking losses in forward looking LLPs as opposed to the prosyclicality of the impact of truly through the cycle forward looking LLPs on landing. Thank you. I think that's all then we have space for one last question please. I think that this reform gave firms a strong incentive to grow or at least a great gave a loan officers strong incentives to tell firms to mark to market some assets that had grown in value and so on. And these could explain why Scander is skeptical about the high effects because of the companies that remain below the threshold are weaker along an unobserved dimension. So it will be and I would think that this effect becomes stronger and stronger as you go in time and firms have stronger and more time to adjust. Then we close the question over here and I give back to Jose Luis to respond to the discussant and also the questions please. So I will go into this order but first let me just say thanks Scander for the excellent suggestions. So on my Asunta. Yeah, so in some regressions we don't do regression discontinuity just because precisely because what you just said. And so and then in some regressions is this paper in revision so it's the lots of robustness. So we take we move away from the threshold to avoid firms which are say the very good firms which will be close to the first threshold that we want to switch to the other one. One thing that we should check is the economic effects whether they are reduced going to your suggestion. So cost as yet there are persistent effects and persistent effects especially in crisis times. So the crisis times is the strongest. Jacob I completely agree on that. I mean it's just a point that I wanted even to highlight more here on the supervision. And that was my question to Andrea that you know like at the end nothing is about is not only regulation is just supervision as well over and above the rules because the banks will try to adjust. Yeah Peter long rates high sometimes it's like, you know like in this long sometimes the number we should check very well these things because sometimes the numbers are high because you have the fees, the rates and sometimes if the, you know the drawdown is very small so the the loans are very small. So the interest rates could be high so we should check on the interest rate whether it's driven by these smaller numbers. And not only, you know states one is also important because there have was a change in the state one and two. You know we used to only have state three and the introduction is the state one and two and both are important. And on a standard we will check the numbers. We are checking the numbers on basically the big, big number in the long rates is what I answered to Peter. And there is the issue and Maria Sunta as well and then there is the issue of the external validity. And here external validity. These are forward looking provisions. You know, what one problem that I might say more philosophical is that in medicine they do lots of you know test in order to be confident in a particular drug or something like that. So here in order to understand better a particular law we have to do we have to have many lots of evidences. And this is one. The external validity is that it's forward looking. We have the different recessions in good times and bad times both with the policy and without the policy. There is a state one and two versus purely state three and the effects are, you know, in a sense they're related if you think about the COVID-19 crisis in Europe and the United States, but if they follow the expected credit loss provisions will go up very high and there will be a drop in credit. That is my interpretation why the regulators supervision supervisors, you know, water down these chains in provisioning at the time of the COVID. So thanks a lot.