S&P Global Inc. (SPGI) Earnings Call Transcript & Summary

Hello, and welcome to the second webinar in the Banking Essentials series brought to you by S&P Global Market Intelligence and the European Banking Federation. Today's discussion will be about managing credit risk during the COVID-19 crisis and the recovery phase in Europe. We're seeing some easing of lockdown restrictions in some countries. But the pandemic is still very much in our lives and affecting businesses across the continent. So credit risk is clearly a big issue that banks must contend with. My name is Ben Meggeson. I'm an editor in the European Financial News team at Market Intelligence. And I will be moderating today's session. I'm joined by Gonzalo Gasós, Senior Director of Prudential Policy and Supervision at the EBF; and Arsene Lui, Senior Quantitative Analyst at S&P Global Market Intelligence. Gonzalo and Arsene will first talk us through some key themes that they're seeing in the management of credit risk. After that, I will be putting to them some questions from you, the audience. But before we start, let's take a look at the welcome poll that we ran before the opening. And the question that we asked was how big will the total pile of European COVID-related nonperforming loans be. And it looks like 35% of you think that the total will be less than EUR 1 trillion; 32% between EUR 1 trillion and EUR 1.5 trillion; 19%, EUR 1.5 trillion to EUR 2 trillion; and then 14%, more than EUR 2 trillion. So clearly, people erring towards the bottom end of that range. But whatever the eventual size of this pile, it's going to be a big pile of bad loans. And so we're going to be discussing how banks and the credit management industry can best manage that. Before our speakers start their presentations. However, I'd like to just say a couple of housekeeping items. Firstly, this session is being recorded, and a playback will be made available on the websites of both the EBF and Market Intelligence afterwards. Also, speaker slides will be made available after the event for your reference. I would also like to invite you to submit your questions for our presenters at any point during the webinar, which you can do using the questions box in the webinar tool. So go ahead, click on questions and tell us what you want to ask are experts. So talking of our experts, our first panelist is Gonzalo. He has been Director of Prudential Policy and Supervision at the EBF since 2015. He's a member of the Advisory Board of the European Banking Institute. From 2010 to 2015, he was an adviser on banking regulation at the EBF. From 1995 to 2010, he developed various positions at Santander Group in risk management and the implementation of Basel II standards, and he began his professional career in 1990 at Accenture. So Gonzalo, over to you.

Thank you, Ben, and good afternoon, everyone, if you are in Europe; or good evening, good morning, wherever you are in the world. And also thanks to the audience for being slightly more optimistic than what we know about the upcoming wave of NPLs that we are talking about today. In my presentation, I will refer to 4 topics: the credit quality trend, the banks' constraints to manage their deteriorated assets, the options to deal with them and what can we expect from the NPL market. First, let's talk about the credit quality trend. Many of you will have watched that film about the tsunami that hit the coast of Thailand, starring Naomi Watts and Ewan McGregor. You will remember the first scene was an unusual moment of calmness followed by a strange wind and rumbling noise preceding the blow of the tsunami. On screen, you will recognize a similar pattern in the credit market today. In the first chart, you see that we are living a moment of calmness in the NPL ratio during the last decade. Even if there has been the outbreak of COVID, the NPL ratio continues dropping until the latest data we have. However, there are hidden signs of asset deterioration. If we look at Stage 2 assets in the last months, an additional tranche of 3% of the credit portfolio in Europe is believed to have experienced a significant increase in credit risk. This can bring us to about 10% of deteriorated assets in the overall portfolio. And latest data published by the ECB last week indicates further growth up to 13%. Whether this is the preamble of an NPL tsunami is a topical question these days. But unlike realized tsunamis, which caught everyone unprepared, in our case, the banking system at large has taken preventive measures to absorb the shock, in the first place, a comfortable level of capital built up during years and, more recently, a substantial increase in the level of provisions. In the next slide, we see that -- nonetheless, what is clear is that the level of NPLs will rise sharply. The legacy NPL in Europe amounts to less than 3% today. If only half of today's Stage 2 loans become NPL in 2022, the average NPL ratio in Europe could jump to nearly 9%, which is just the peak level of the past crisis, however, in a shorter period of accumulation, which will put under stress the capacity of banks to deal with them. There are several factors constraining that capacity. First, human resources to analyze the recovery capacity of a massive load of unlikely to pay loans and later to engage and execute on recovery proceedings. The second constraint is the size of many portfolios affected by the COVID crisis, which will not pay off the cost of due diligence, analysis and recovery actions. Number three, the identification of viable businesses will require a lot of soft information, which is not readily available. And also regulatory capital will set the limit to the amount of loans that can be worked out internally by the banks. With the backstop provisioning calendars, which were frozen during the moratoria, we'll again consume capital via minimum provisioning. So let's have a look at the equation of what banks can do with the NPLs starting from this year. Due to the sharp increase and the vast constraints, a large volume of new NPLs will need to be transferred out. Banks will have to select which loans are worth staying on the balance sheet, employing limited capital and human resources to work them out. With this option, the bank can obtain the real economic value of every loan, assuming a partial loss that is represented by the yellow tranche that you see in the graph. NPL securitization could be a widespread option as well this time around, following the successful experiences in Italy and Greece, including with the GAC system. And asset management companies, let's not forget, were the most important vehicle in the past NPL wave. The question here is where to fix the transfer price, therefore, the upfront loss for the bank and whether the AMC will be able to recover the economic value of those loans without incurring in state aid. Finally, direct sales to the market have been widely used so far. However, this time, prices could be depressed if a huge supply of NPLs in a short time doesn't meet sufficient market demand. In those conditions, the estimated market value, that is represented there in the graph, could sit far away from the net book value of those loans today with a sizable gap, represented there in green color, from its future economic value if they stayed on the balance sheet of banks. Then interestingly, we can see the emergence of electronic trading platforms, which could be an outlet for multiple portfolios, especially of smaller loans. Regulators are designing standard templates that can be helpful if defined in a pragmatic manner. Now let's look at the market. We have seen a very dynamic and liquid NPL market developing in the last years. However, most traded assets were legacy NPLs, which valuation was determined just by the value of the collateral. The amount of UTP loans placed in the market was just a fraction. You can see that in the light blue bar, the light blue tranche, shown in Italy at the top bar. That is the only fraction of UTPs that were traded in the last 6 years. So the fact is that we don't have sufficient experience in UTP trading. And the next NPL wave will be full of UTP exposures, the majority of which might be unsecured. This fact puts into question whether we can use conventional measures or a different approach is needed. It is uncertain how much market appetite will be there for European UTPs, and it will determine the extent to wait AMCs will be necessary. In the meantime, the only fact we know is that the NPL market has slowed down in 2020, as you can see in the bottom graph, which is a bit in line with that moment so well represented by Naomi Watts and Ewan McGregor, the strange calmness before the tsunami. I will be happy to take questions and to learn from the many risk managers in the audience about how to mitigate the effects of the next NPL wave during the Q&A session. Now over to you, Ben.

Thanks, Gonzalo and really, really interesting to get that overview of the preparedness for this oncoming tsunami and some of the challenges that await banks as well. We certainly will get to questions but not before I turn the floor over to our next speaker, Arsene Lui. Arsene is a Senior Quantitative Analyst at Market Intelligence based in London. He's responsible for the analytical development, maintenance and ongoing validation of credit risk models and products across Market Intelligence's credit analytics tool set, which are used by financial institutions to measure and manage credit risk. So over to you, Arsene.

Thank you, Ben. Good afternoon, good morning or good evening, everyone. So I will talk about some results on the empirical analysis of credit risk assessment. So first of all, to understand how the credit risks have evolved during the COVID crisis. The first thing we need to know is how to quantify the risk. For the company having their bonds trading in the market, it is well straightforward. So we can just look at the market information, the bond price or the CDS spread. And besides the market data, so the ratings produced by the credit rating agencies, such as S&P Global's rating, Moody's or Fitch, can also give us a quite direct and holistic assessment on the credit risk of the company. But sadly speaking, the number of rating companies are so tiny when compared to the millions of operating companies. So banks or lenders have to build their own version of models using whatever data available to them or sourcing the data or models from data vendors. In S&P Global Market Intelligence, we have leveraged on the financial data and the market data we collect and to develop a number of quantitative models that span the full spectrum of credit risk assessment from a full cycle to a point-in-time model. Each model have a different methodology, use different input factors and what's trained on a different target level. So the usage of each model depends on the availability of the input factors and also the training level being used. So we also combine the output from all these models to create a model called RiskGauge model, which gives a user a more comprehensive view on the credit risk of the company. Next slide, please. So first of all, let's have a look on the change in credit risk during the COVID from the viewpoint of the company's operations, that is their financials item. And this slide is showing the output from our PD Model Fundamentals, which is a statistical model that's trained on historical events and the financial ratios of the company, such as the profit margin, return on capital or maybe, let's say, EBITDA-to-debt ratio. And the primary output of the model is a forward-looking 1-year probability of default. And then we map it into a credit score according to the PD value. On the left-hand side on this slide is showing the distribution of credit score of 3 major European countries, the U.K., Germany and France. Before the COVID crisis, we can see the peak of the distribution is around BB level for U.K. and Germany and a slightly lower B+ level for France. And for the fiscal year 2020, so we see a notable shift in the distribution to the right, with the peak of the distribution around 1 notch lower than the pre-COVID level. The chart on the right-hand side is showing the comparison of the credit score between 2019 and 2020 of all the companies in our database. Around 30% to 40% of the companies got worse credit score during the COVID crisis, which is slightly higher than the historical level and only 20% to 30% of the companies has a better score. So we can also see from the figures, France, Italy and Spain has a more severe deterioration in the deferred risk with the percentage of company downgrade excess 40%. Next slide, please. The PD model fundamentals provide a medium- to long-term view on the deferred risk assessment. However, during the COVID crisis, the economy and also the business environment were evolving so fast. So it would be more useful to have a more point-in-time model, which can give us some light before the release of the next financial report or financial information. And the PD model market signals used the equity market information and the merchant model approach to estimate the market value of a company and also the distance to default. And the distance to default is then mapped into a probability of default value. And this model is able to provide a daily update on the PD estimate. So before the COVID crisis, the median PD of these 7 European countries were all below 1%, and the median PD start to increase at the beginning of March last year and were at a high level until it starts to drop again in June. We can see the second wave of the credit risk deterioration in around October and November. But for this time, it seems like Italy, France and Spain, again, were more severely impacted than the rest of the countries. And we can also see that the most recent PD has already come back to the pre-COVID and which indicates a quite good recovery or a sign of recovery. Next slide, please. Okay, in this slide, I want to show the result of the analysis on an alternative data set, which is the trade credit data set. So the trade credit data represent the short-term liquidity need of the company. And it also tells us how well the company repaid their trade credit obligation. In the figures, we summarized the monthly trade credit data using the day beyond terms measure, so which is just a dollar rate average number of days beyond the contractual days of the invoice that the company owe to their creditors. So based on the trade credit data, we also developed a quantitative model called PaySense, which forecast the potential payment delays using the historical trade credit data together with other information such as the company-specific information or some macroeconomic factors. And the PaySense score ranges from 1 to 100, with a score of 100 as the best. In the figures, we can see the significant increase in the day beyond terms during the COVID crisis. And majority of the company were paying their suppliers on average 1 to 2 months' late. And the trend of this trade credit dataset is different from the peer analysis on the market signal PD. The day beyond terms and also the PaySense score is not going back to the COVID level even if we look at the most recent data. Probably, this suggests that companies are still in short of liquidity and they need to stretch their trade credit to pay for other, more urgent expense. Okay. All right. I think that's the end of my part and now passing it back to Ben.

Thanks so much, Arsene. Really, really interesting to get some of those more granular details of credit risk analysis and see how the trade credit data can really add some color to that analysis. So we will now move on to the Q&A section of the webinar. And first of all, we will be putting up the second polling question on the screen for our audience. So if everybody could have a look at this and answer. Which will be the most common way for banks to sell the NPLs they cannot deal with internally: securitizations, direct sales, asset management companies or electronic trading platforms? Which one of those methods will be most commonly used to work out bad loans that can't be resolved in-house in banks? I'm going to put that question to our panelists as well while we wait for the results of this question to come in. So Gonzalo and Arsene, maybe in that order, what's your take? Which do you think will be the most common?

Well, Ben, I think that a mix of different options to transfer out of the balance sheet of banks the assets, the deteriorated assets, that banks cannot deal with it in a short period of time is the key question. However, I think what we need to consider is that for the client and for the bank, the best option for viable clients is to stay on the balance sheet of the bank because after all, those clients that have been hit by an external disaster like COVID, many of them will still be viable when the economy recovers there. And therefore, they represent also the longer-term client relationship for the banks. So I think at this moment, banks are very much selecting which loans deserve to stay on the balance sheet of the bank. And for the rest, I think a very interesting option that we have learned in the last 2 years, especially from Italy and Greece, is securitization. Regrettably, securitization was stigmatized in the past financial crisis. But I think we are now acknowledging the benefits of securitization because not only it permits to transfer those assets out of the balance sheet of banks and be placed in the market but also they can make room for new fresh lending on the balance sheet of banks. One of the problems we face now in the banking sector is that the accumulation of NPLs doesn't leave sufficient space to grant new lending. So securitization, in my opinion, is the most beneficial option. It has grown a lot. And the GAC system that has been developed in Italy is a best practice to look at. Then obviously, the AMCs, but I think that should be the last resort. And I think there is also an interesting potential in electronic trading platforms for smaller portfolios. Those are working very well, and it could be an outlet in this crisis very much. Let's see what the audience thinks about it.

Well, thank you, Gonzalo. Well, it would be quite interesting to see the result, okay, because I'm more eager to see the audience to choose the electronic trading platform because you mentioned the benefit of the trading platform, then more transparent of the data set, and also it allows the bank to reach out to a broader set of investors. So as a model developer as well, I think the benefit of the electronic trading platform will become more and more important, especially when we need data for modeling and we need data for testing and we need data for validation.

Yes. Thanks, Arsene and Gonzalo. And yes, like you say, Arsene, interesting to see our audience thinks that AMCs or banks will be the most commonly used method to resolve NPLs outside banks entirely working them out. Perhaps there's some recency bias there given how popular those organizations were after the global financial crisis. And I note that Gonzalo, in your presentation, you pointed out the EUR 400 billion worth of bad debt was absorbed by all those European AMCs. We have a question -- I'm going to move to questions from the audience. I do have 1 or 2 of my own that, hopefully, I'll get a chance to work in. But there's one that segues neatly from the issue of securitization from one of our listeners. When we're talking about the mitigation of the tsunami, Gonzalo, that you mentioned, should banks proactively securitize Stage 2 loans before they hit Stage 3? Perhaps, Gonzalo, if you could tackle that question first and then Arsene?

That is a very interesting question because it is a new paradigm in the COVID-19 aftermath. So far, I think we have a lot of experience with NPLs that have been legacy NPLs for years, but now unlikely to pay loans is a different animal and we really need to think about it. And banks are preparing their strategies. However, it becomes a challenge in the first place in terms of valuation, when you securitize a portfolio of, let's say, mortgages. You have some real estate indicators, housing indexes and the value of properties and so on. But when you securitize unlikely to pay loans, which many of them are unsecured with no collateral, retail unsecured, that's really a challenge in terms of valuation, but it also makes the risk function more interesting. There are companies developing sophisticated credit analytics to help banks and also investors assess the value of those portfolios. But I think my conclusion is that there will be securitization of UTP, of course, but that will require some developments in terms of pricing.

Yes. I agree with this. And the challenging part is to getting the data to accurately, let's say, value the collateral and the likelihood of not paying. And then especially for discount of exposure, the traditional models are usually not that applicable. I'm not saying it is not useful, we can still use it as a reference, but not applicable to dose exposure, and we need to think about maybe some other alternate data set in order to give us a more accurate -- or some insights on how to evaluate those exposure.

Could I ask a follow-up to that to you, Arsene? Given that we've had such a sort of once in a lifetime experience with this coronavirus crisis, when we think about statistical models that were trained before COVID, do we still expect them to work in the current context? Or do we need to retrain those models? And what needs to change going forward?

Well, that's a very tough question to a model developer and to every model developer. Well, I think there's no doubt the performance of a trained statistical model before the COVID crisis are not performing as good as before during the COVID crisis period. But I don't think it means that the model are completely useless. So one major factor affecting the model performance is the accuracy of the model input. So during the crisis and many shops and companies were closed due to the lockdown or they can only operate at reduced capacity. So the significant drop in the revenues or cash flow will be reflected on the financial statement. And if we use a model that's trained on the financial items, it will trigger an increase in the PD assessment or PD estimate. But if we look closer into the actual situation, we will see that some companies are actually getting maybe debt moratorium or some other form of support from the government. So a lower maybe income-to-debt ratio may not be sustainable before the COVID-19 will be a norm or a new normal during this unprecedented time. So if we make a proper adjustment and maybe a little bit forecast on the company's financials by considering these factors, then the statistical model can still fit the purpose to some extent. And also, besides the issue of the model input and other big concern on statistical model is always where the calibration of the model's output is still correct or still appropriate during the COVID period. So the PD level generated by the [ statistical ] model, whether it matched the actual default frequency if, let's say, we observe right now for 2020. So we know our model, statistical model, may not be performing well on this, but there are other type of models such as the model -- using the merchant model methodology, such as our PD models, market signals, which use the market data information. So -- and the output of their models -- of this model is more point in time and that we can just use this model. Maybe the output are not available for every company because it required, let's say, equity market information and only applicable to the public and listed company, but we can still use it, let's say, to get a flavor on the general level of the change in the [ ODR ] or a change in the PD. And then use this as a proxy to adjust or even recalibrated output of other, let's say, more traditional statistical model. So if we are aware of the limitation and also the characteristic of each model, either a statistical model, using financials or more point-in-time model using market information, then we carefully adjust the input and also the output, and I think we can still get some useful insight from the statistical model trend before the COVID.

Good to know. Thanks, Arsene. So just to zoom out a bit, and Gonzalo, this is a question that I'd like to put to you. The amount of NPLs last time -- when I say last time in the past-Euro area crisis peaked at about EUR 1 trillion. And according to the ECB, this time, COVID-related NPLs are estimated to total about EUR 1.4 trillion. And we've seen from our earlier poll, our listeners today think it might be less than that, but it is going to be probably more than the previous crisis. What is similar between current crisis and the previous crisis?

Well, that's the key question about the total figure of NPLs, which comes to the front page of papers, of course. But I think even if it is quite similar to the past crisis, the components of that NPL pile is completely different because, in the past, we had legacy loans, legacy loans that had been sticking to the balance sheet of banks from the times of the real estate bubbles, the archetypical loan in the past Euro crisis was coming from the real estate bubble. It could be a resort, a building or an airport built in the middle of nowhere where the only strategy was the liquidation and collateral realization. Of course, there was more to that and there were mortgages and so on, but it was completely different in the type of NPLs we have to deal with this time around. These are, as we've mentioned, mostly unlikely to pay loans, many of which might have a clean track record of payments in the long past which makes them completely different. They will all come in a single vintage. In the past crisis, we had different vintages and the vintage of the loan would explain the probability of default and, most importantly, the loss given default. But this time around, we have a single vintage, the COVID vintage, that will emerge massively in 2022. And a significant part of them might be retail unsecured, which makes valuation more difficult. Concentration, instead of real estate, it will be more on the side of corporate and SMEs. Therefore, the strategies to deal with these loans have to be necessarily different. Also, we also need to take into account the social angle of these NPLs because, at the end, this comes out of an external disaster. For that reason, obviously, policymakers last year took measures to freeze the forbearance rules, for instance, during the moratorium. So there is also a social aspect involved here and there is willingness to help viable borrowers to reach the economic recovery and come back to normal, which is another characteristic we shouldn't forget about. And all this makes analysis more interesting because as Arsene has mentioned before, probably when we analyze these portfolios, we should put more weight on the point-in-time analysis. Of course, the financial ratios are there, and the statistics are always right because they are factual. However, when it comes to forecast, you need probably some judgment and a lot of point-in-time analysis as the models that Arsene is building.

Indeed. And talking of probability of default, we've seen those reduce significantly since the height of the COVID crisis in all geographies. And as Arsene pointed to in your presentation, are there -- and this question I'd like to put to both of you. Maybe Arsene could take it first and then Gonzalo. Are there specific countries or geographies or regions within Europe that are more vulnerable to a spike in probability of default rates than others?

Well, actually, the analysis I did and the results I showed on the presentation slide about the European countries, but I also extract the data for the other countries that our model covered. And of course, the impact of the COVID on this country depends on the industry and then the industry and their importance to the country. But in general, we see the country in East Asia and also in the emerging markets may not be doing that well in the recovery, especially if we look at, let's say, the median or the mean market signal PD on the country level. And I think it is also consistent with the existing situation of the corporate right now and the successful way of the vaccination in those countries compared to the Eastern European countries and even to the North American countries.

And Gonzalo, could I put the same question to you?

Yes, I think Arsene has given a more global perspective. If we look to Europe, I think more than country wise, the impact will be -- we need to look at the sectoral analysis. At the beginning, after the outbreak of COVID, there was a general contagion of the unlikeliness to pay. But now the focus has been narrowed down into the certain sectors, which obviously will take longer to recover. And therefore, some countries are more concentrated or more dependent on those sectors, and that is the reason why they might be hit harder. But in principle, it's -- unlike the Euro area crisis, which obviously was correlated with the sovereign spread, obviously, in this case, the nature of the crisis is different, and it might hit companies in different sectors, irrespective of the country they are headquartered.

Interesting, yes. So people be keen to find out which sectors might be more at risk. Okay. Well, we, I think, have time for one more question. We're going to be wrapping it up in 3 minutes or so, but there's time for one more. And this one, I'll put to you, Gonzalo first, and Arsene. But if I could ask you to give brief answers, that would be appreciated. So do we think the banks are adequately provisioned for the upcoming wave of defaults? And do we have expectations for default rates by rating category?

I take the first one about bank's provisions. I think you're never safe in this world, obviously. But I think at the level of capital and provisions that banks have built in the last years is huge. So that makes you feel comfortable about that. It depends on the bank, obviously, because there are varying degrees of safety. However, if you compare the starting point of this crisis with the previous one, it's completely different. We are very much on control of what is coming also in terms of regulation, in terms of supervision. We are on a stronger footing, obviously, but we need to remain attentive and take decision as the crisis unfolds.

Well, regarding to the default rate of each rating categories, so it is a tough question, but maybe I can answer this question with the data from the S&P Global Rating side. And S&P Global Rating, they published the default and rating transitions starting every year, and the latest report for 2020 was just published last month. So in their research -- and this is summarized default they observed on all the rated companies last year -- on an overall level, for the investment grade company, the observed default rate increased from around 1.3% to 2.7%. And for speculative grade, the increase, I think, is maybe from with 2.5% to 5.5%. The exact number can be found in the report. And also the more breakdown of the default rate into different rating category industry and country are also available. I think that if we want to know what happens in the past and then use it, let's say, predict the default rate in the future, that report is a must-read.

Great. Thank you, both Arsene and Gonzalo, for the succinct answer. I realized that was quite a tough question to seed in the last couple of minutes. Well, thank you, everyone. We've arrived at the end of Banking Essentials webinar. I just want to remind you that today's session has been recorded, and the playback will be available on the MI and the EBF websites. I'd also just point out that we will -- if there are outstanding questions that we haven't had a chance to get to, we will have a go answering those over the next couple of weeks. So look out for those as well. And yes, we look forward to welcoming you back to the next In the Banking Essentials series, which is on the 14th of September, where we'll be looking at ESG in the private sector. And if you stay online just for 2 seconds now, you will also have a chance to sign up for further information if you should so desire. So to wrap up, thank you so much to both our panelists, Gonzalo and Arsene. And a big thank you to all of you for attending today, and see you next time. Thank you.

Thank you, Ben. Bye.

Bye.