 Okay, perfect. So the previous paper was about the climate cross-country climate policy coordination. And now the next paper is about supervisory cooperation, incomplete coverage in supervisory cooperation and cooperation externalities by Wolf Wagner from the Rotterdam School of Management. Wolf, please. Thank you. Thank you for staying for the last paper of today. So the paper is on incomplete supervisory cooperation. It's together with Torsten, who has conveniently disappeared at the moment, and Conselo Silva-Bostong from Universite Cartolica in Chile. So the paper, as Andrea said, actually is quite closely linked on the conceptual level to the one from Emanuela, so it's about cross-border regulatory arbitrage. But here the source is different. So we're looking at cooperation of banking supervisors, and this cooperation is effective in the sense of the countries involved in the cooperation, so their risks go down. But precisely because of this, actually there's pressure for the risk to go to unaffected or to third countries that are not in the cooperation agreement, and this has some kind of undesirable consequences for the global effectiveness of cooperation. So we think about the big banks in the world, right? So these are cross-border banks that operate among different countries, different jurisdictions, there are a lot of different regulators and policy makers involved, and this creates a lot of issues. So we know this. And to contain risks at these banks, domestic supervisors, national supervisors, they're frequently a corporate, okay? And this cooperation can take many different forms. So it can, for example, be some kind of joint supervision. So think about the banking union would be an extreme example where you replace national supervisors with one common supervisor, in this case for the large banks. But there can be also many other types of agreements, and actually in all data the most common is information agreements or information cooperation. So you're making an agreement to exchange information on your banks. So, for example, this is very prevalent in Latin America, so there are a lot of small agreements between different countries and they're agreeing on exchanging information. So in a sense that the banking union within our data set is not the most representative, okay? It's a very powerful cooperation, but in most cases, actually we're talking about these information exchanges across a smaller set of countries. So what's the potential issue with that? The issue is that these agreements are often made on the country level and so they don't necessarily cover the involved banks. So there's a mismatch between the cooperation area and the geographic footprint of the banks. So there's this incomplete coverage and this creates regulatory arbitrage, the potential for regulatory arbitrage. So cooperation between A and B presumably makes it more difficult for banks to take a risk in these countries, so they have an incentive to shift to C. And this may explain why cooperation is not effective for the very largest banks. So in our private paper in the JFQA, we find that actually if you look at the consolidated bank level, cooperation is effective, okay? But not if you look at the very, very large banks and these are the banks that also have subsidiaries in many countries so they can shift risks pretty easily. So the idea is that cooperation between A and B kind of makes it difficult for banks to take risk in this country and this creates this pressure to shift risk into third countries and there's quite some literature, mostly theoretical, unless people in the audience have contributed on this. So for example, there's this idea of negative externalities across countries. So if a bank from A is failing, it has negative spilloes on B. So once A and B cooperate, these negative externalities are internalized and this makes more stringent supervision optimal. There are also arguments based on heterogeneity, right? So if you have countries that differ initially in the stringency of regulation, so the theory suggests that there will be a tendency to only cooperate if basically you move to a stricter standard. But there's also a sense in which supervision is becoming more effective and it's simply because you have more information, right? So you can exchange information about cross-border movements and this makes it easier for you to take risks. So all of these channels basically suggest that it's going to be harder for banks to take risk. So basically in this paper we show two things. One is this incomplete coverage creates a third country risk shifting and second, this risk shifting in turn has kind of effects for the distribution of gains and costs of cooperation. So basically the costs are shifted kind of outside the cooperation area where the gains are mostly within the countries that are cooperating. So we have collected a large set of data, so hand-connected data. It's on formal cooperation agreements. So we have quite a large number of countries. So in terms of home countries we have about 100. In terms of subsidiary countries we have more. What's also interesting is that these agreements, they are bilateral or multilateral, so actually we have quite a big number of pairs or sets of countries or more than 10,000 potential country pairs that can cooperate. So we have now data since 1995. So for this second pair we updated the data to 2019. Now we look at simply here the existence of cooperation agreements. So we look at the data. So now in this map kind of dark is good. In the previous paper, dark was bad. So here dark is having higher cooperation intensities. So what you see is that there's a lot of variation across countries. So basically here we measure basically for each country with how many other countries in our data set you're cooperating. But you see it's in Europe it's relatively dark, reflecting for example the banking union. But you see that the cooperation agreements, many other places of the world, so there's a lot of cooperation in Africa and Latin America. But you can say that cooperation, I mean this should be conditioned on kind of having some banking links. So there's no sense of cooperating if you don't have any banks in common. So what we're doing is we're combining this with kind of a banking sample. So we have a large set of banks. So they have 600 subsidiaries. They spend 47 home countries and 100 host countries. And now we can compare, we can calculate a different metric. So here, so darker is again higher coverage in this case. So basically what we're doing is we are calculating metrics at the country level. So we're looking for country at all the banks incorporated like headquartered in that country. And then we look to what extent the parent subsidiary, the home host relationships are covered by cooperation agreements. And you see that we get a bit different pictures. For example Europe kind of gets a bit lighter and the reason is that yes we have the banking union but we also have kind of relationships, European banks in Latin America and other places in the world. So this kind of reduces the coverage. Another interesting example is Latin America. So actually they don't cooperate much but they cooperate exactly where, so they have a few banks that are really locally overlapping and they're cooperating exactly in these areas. That's why they're very dark. They have low intensities but they have high coverage. Okay, so this is from our kind of first paper. So basically it's kind of the direct effect. So we're looking here at lending in a subsidiary. So let's say we think about lending as a measure of risk taking. And here basically the home and host countries start cooperating and we see that lending slows. So that's kind of motivating this idea of making it more difficult to take risk in A and B. Now here we are looking at the third country effect, right? So let's look at Royal Bank of Scotland in 2008 and 2009 and let's look at risk shifting into the Argentinian subsidiary. So arguably in 2008 there was not much risk shifting pressure because there were eight other subsidiaries and only one, so dark is indicating a cooperation agreement with the UK. So only one is covered by a cooperation agreement. Now in 2009, so basically UK signs cooperation agreements with many other countries. So basically then how many subsidiaries are covered by cooperation agreements. So you can argue that in 2009 the pressure to shift risk into Argentine has increased. So that's basically what we're going to study. Regression. So the setup is, so basically we're interested in how a risk allocation into a specific subsidiary, how this depends on cooperation coverage in the rest of the group. So we are regressing lending as a measure of risk taking on a group cooperation. What is group cooperation? It's the share of subsidiary countries of the banking group for cooperation agreement with the home country excluding the subsidiary itself. It's like the cooperation coverage in the rest of the... And for identification what is interesting is here that we are basically relying on third country. So when third countries sign cooperation agreements with the UK we look at how this is influencing Argentina. But think about what we can do is we can, in our setting really control for parent subsidiary country time-fixed effects. So we can have a lot of time-fixed effects. So basically we are comparing two subsidiaries in the same country so they have the parent bank is also in identical countries. But these banking groups they have different geographic food points. And our third countries are signing agreements with the UK that this creates differential risk shifting. Okay so we find that lending indeed shifts out. So it shifts into the subsidiary if group cooperation increases. Now this seems to be also indicating a general change in risk so it's also that it's not just substitution away from other activities. It's really an increase in the size of the subsidiary. Leverage increases so the liability side becomes more... risky. And if you look at the Z score as an overall measure of default risk it also indicates higher riskiness. So we do an alternative exercise at the loan level. So the first exercise is really looking at the overall subsidiary. And now we look at individual loans. And basically what we do is we look for a given loan. So we take a loan as given through which subsidiary a banking group is originating the loan. Okay. And we look at how this is relating to residual group cooperation of this subsidiary. And basically we find that the same kind of type of result. So basically there's this risk shifting pressure. And here the loan level what we can also look at we can look at explicit measures of lending risk or loan risk. We find that if the loan is ex under risk here you know this risk shifting pressure is more pronounced. So then we look at several kind of determinants of this risk shifting. We look at what kind of country characteristics from the subsidiary perspective affect the extent of risk shifting. And one thing we look at is supervisory stringency. So the idea is if it's motivated by risk shifting then if supervisory stringency in a subsidiary country is high then this should mitigate the risk shifting. And that's also what we find. So we have different measures of relative stringency. So subsidiary country versus subsidiaries in the rest of the banking group. Okay. So we construct relative stringency measures. And so we have different proxies for those and in each case they indicate that you know if the subsidiary country you know basically they can protect themselves against the inflows by having you know a stringent supervisory framework. So in the last part we are looking at basically the distribution of benefits and costs from cooperation. And basically if you think about this through basically there's an interesting implication here. So if you look at the effectiveness of cooperation purely from the perspective of country A and B actually it's increasing if banking groups can shift risk into C. So we also have a model underpinning this but basically the idea is that yeah so if basically there's no risk shifting opportunity into country C for banking groups so if they face more stringent supervision in A and B they will reduce risk a bit but you know they want to still keep up their risk so they're doing some risk shifting within the countries. But if there's opportunity to shift into C they will shift risk at least partially into C and this lowers risk. So if we just look naively at A and B it looks like supervision has become more effective in terms of reducing risk in these countries. Of course the risk on the world level is just shifted around. So the prediction based on our model is that if A and B do not internalize this effect so then they don't kind of consider also this negative effect on country C then actually their gains on cooperation they're higher if there's risk shifting opportunities for their banks. So the prediction is basically if risk shifting opportunities for banks exist they can shift risk into third countries so the gains from cooperation they're higher for A and B. And basically what we do is so we look at existing kind of frameworks existing models of the benefits and costs to cooperation and we kind of amend them by including this new risk shifting term. So specifically what we're doing we're now explaining cooperation so we are regressing the propensity of two countries to cooperate on kind of controls general controls for cooperation costs and benefit and then this new risk shifting term and how do we measure this? We have kind of several ways to do this in the bearer but the most intuitive one is so you look at kind of the number of third countries in which banking groups that are active and A and B operate. So this kind of measures the opportunity in terms of the country set the exploit for shifting risk. And what we find is that so this risk shifting measure this risk shifting opportunity measure is positively related to cooperation across countries and so basically what does it mean? So when a situation appears to be more effective from the perspective of country A and B because a risk can be shifted into C A and B are also more likely to cooperate. And this indicates like a negative kind of a cost so this is a potential source of inefficiency because these are kind of gains accruing to A and B but actually come at a cost of C. So here we have a graph, a map where basically we look at this risk shifting potential so this is now from the subsidiary country the perspective of a subsidiary country and basically we look at for country we take all the subsidiaries from global banking groups and then we look at how these groups are cooperating again in their residual subsidiary structure. So this measures basically the potential of the risk shifting pressure you get from outside the country. And on the map again it changes so the darker is indicating higher risk shifting exposure and now it looks like a bit like kind of developed developing country divide. So like let's say Europe and North America they do quite well also because they're mostly they have the parent group in their countries but then for example South America they host a lot of subsidiaries of for example European banking groups so they're exposed to this risk shifting pressure. Okay so let me wrap up and come to conclude so basically the main thing we want to point out here is that there's this mismatch of cooperation area and footprint, geographic footprint of banks so this is what we call incomplete coverage and this creates risk shifting into third countries and this undermines on one hand it undermines the global effectiveness of cooperation agreements but it also potentially creates this incentive program so if the set of countries that are thinking about cooperation differs from the affected countries then basically you know the incentives might be distorted because the cost of cooperation they're partly shifted to the country's upside. Now the policy I'm mentioning is very clear so it's really about this mismatch so what it means is that when you think about cooperating I mean you should think kind of big in terms of the countries so you should involve all countries as much as possible the countries where the banks have a meaningful footprint and what we also argue that this also speaks in favor of multilateral agreements where basically on one shot you know you're trying to set up a big agreement on many countries at the same time so this kind of reduces the negative potential side effects of cooperation. Thank you. Thank you Wolf. Perfect timing. I'll discuss in this David Marcus from ECB's research department please. In a nutshell what this paper is doing is to show that an increasing cooperation and cooperation is a proxy for supervisory scrutiny. Banks expand and where would a bank expand so they would go to greener posture so what we call greener posture is places in which there would be less scrutiny because scrutiny is expensive for banks and the mechanism as Wolf explained better than me is that countries do not internalize effects in third country so that's in turn this will lead to an increase in the propensity to coordinate, right? So very graphically we've got a cat which is the supervisor chasing the bank because he doesn't want the bank to take excessive risks on other people's money and what the mouse does is to hide and obviously the cat will find another cat to cooperate with and reduce the cost for the supervisor so this is in a nutshell what this paper is doing. Now my first job at the ECB around 20 years ago when I was in policy is to attend the banking supervisory committee in which all the EU countries coordinate and have informal coordination agreements so I'm very sympathetic to this type of work and also very skeptical to some extent on how this could be nailed down with the current setting so is the question large enough? I think so, I think the power of supervision has been hardly studied in the literature I think it's super important can they answer it? I think the authors do a great job I think it's a very important paper but to my mind you are trying to answer too many questions and again I love this paper because you are actually looking at a major question and normally as you know there tends to be like a trade-off I will talk about it but I'm very sympathetic about this type of work and I think it's very important Now, why am I convinced about the power of supervision? So let me give you an example very close to home as you know in 2012 the SSM was actually unexpectedly announced and there were talks for the need of a banking union in Europe and a size threshold for banks to be inside the SSM was announced it was 30 billion, 30 billion, sorry so banks have a whole year until the ECB assumed the supervision of the largest banks so I think this is going to be helpful to bring the point of the authors home in the following sense if you look at the distribution of banks' assets in the year prior to the announcement which is this is basically the distribution of banks according to size and as you can see in 2002 it's a kernel distribution so you see how the tails are going down at the end this is purely by construction but effectively in 2012 the red line banks were very smoothly distributed around the threshold of 30 billion in the year after you see that many banks shifted because of anticipation of stronger supervisors so I think Amber is sympathetic to this in Europe at least to the issue of the supervision and even on expectations of supervision this anticipation effect I think is also interesting so on the data in most papers there is a big trade-off between a big question in terms of scope and the identification of a very little component which is very well specified this paper tries to do both it focuses on the identification but to my mind still very little on the data for instance before 2014 the data has been hand collected from annual reports, regulatory websites and newspaper articles this is the core of the paper I would like to know more about this maybe it's my background but I wanted to understand how they coordinated more with with the countries around that I would like to know more about that now I fully understand that in a central bank we have the luxury of data but I have a very big quirk with the use of syndicated loan data because in the end this is just one way of expanding direct lending they also do cross-section a cross-selling of other products and they occurred very very rarely in theory to nail supply effects which is what we are talking about you will need to have two different syndicates before and after and in Europe this happens very very rarely it's around two firms a year that happens syndicated loans happening from two different syndicates before and after my last point is about cooperation the authors pulled together all types of corporations from NUMOU which for the people in the jargon is memorandum of understanding to an actual supranational supervisor so you are bringing together legally legally binding agreements with an informal paper which has been signed at one point in time now people in the business who have been doing policy for a long time they know that cooperation is very often toxic for the lack of action and you have been living in the Netherlands we know what happened with NB and AMRO I can tell you there were a lot of cooperation agreements down there there are also implementation problems time consistency it's very easy to sign a piece of paper with another supervisor now in a different country it's a very different thing to put fiscal money when the financial authorities need to intervene so how about also considering for failing banks which cooperation agreement had prior to failing I'm convinced that the majority of banks who failed had a cooperation agreement in place econometrically you could use local projection act adding different types of cooperation agreements as a predictor of failing bank and then of course the other issue is how about the incentives to cooperate not to implement the stricter supervision I remember the light touch regulation from Gordon Brown in the UK before the crisis often coordination is an excuse not to increase the supervision of banks so a set cooperation sometimes in this country this is the blue light you are familiar with it's something we are worried about and I would like to see more about that so to summarize and to let you go all of you after thanking Andreas again for a great supervision a great conference you see this is a Freudian slip for a great supervision we used to work together I would like to know more about the data I would like to know more about the type of cooperation this is a relevant question very well executed by excellent co-authors and I think it's an important paper with great potential that I would love to see very well published because I think it's a very important contribution thank you so very much and I don't want you to go now because we still have a speech so thanks David thanks David for your kind words and thanks for your discussion so then we go to the last question and answer around this afternoon and the demand is high Harry, Giovanni so I wonder whether to what extent these cooperation agreements that is whether a country enters into agreements with other countries whether that could reflect say local supervisory employment because in the end you need people to carry out and do this cooperation and then local employment of supervisors could also affect bank risk-taking and therefore the quality of supervision so I wonder whether there's an employment issue here Giovanni Bassani here from the SSN it's quite interesting I was probably the question actually should be on your previous paper because I think you mentioned that you found out that even if there is this consolidated supervision at the end of the day that doesn't catch actually the risk in those countries and for somebody doing supervision this is something a bit surprising in a sense because we are we are we are supposed to think that through consolidating supervision you should be actually able to supervise risks in all the perimeter of consolidation if that doesn't happen because you don't have enough information coming from a country you should actually ask the bank to consolidate the subsidiary which then the potential treatment is going to be much more so it seems to me that maybe it's true I mean I'm not saying that it's not true but that the the findings and again maybe more from your previous paper are against the way consolidated supervision works in a sense so this supposed to work which will be something significantly depressing in a sense and also I wasn't sure how you actually defined this what countries are because I think you have a very extensive definition of what the country are I guess all Eastern Europe from from the from the charts you show will become a country as opposed to the SSM and also all South America is basically a certain country so we basically supervise banks on a consolidated basis if there are countries where we can actually get into we should actually take them out for what consolidated supervision is so we have a country right now you can imagine which one which presents a lot of problem but this is sort of an exception you seem to actually generalize this aspect very much I mean it's amazing that it work and you mentioned that I mean there's a lot of variation on cooperation intensity and I think David also appealed to this I would be very interested to learn if you can somehow explore this further for instance I mean I wonder if there's agreements that are focused on specific banking groups like Santander compared to very general high-level agreements also how does this link to the the actual cross-border banking structure I mean for instance if these are agreements between countries that are mostly hosts of foreign banks I don't know I think there's tons of interesting questions to be asked there I appreciate you're trying to finish up so I'll be brief Mary let's bet the magmon I'm from the Central Bank of Ireland thank you very much for the paper I couldn't I had to stop myself going to the question of effectiveness and I don't know whether through your research you kind of tested for the cooperation effectiveness in crisis times and you know whether the existence ultimately of the cooperation agreement I think you touched on this David really did it show anything and then secondly for me given that you had you said you had I think 93 countries you've over 10,000 of the kind of relationships whether we know that these group structures change and more you know continuously through your data gathering do the cooperation agreements change or do they remain static just a few comments it's really a very interesting paper covering understudied topic it's also brave to present this paper to a room full of supervisors let me tell you there would be no change in legal form of cooperation between SSM supervisors between 2012 to 2014 and now but the scope effectiveness the standardization is two different words so also sometimes we've been even a single group of supervisor following a single legal framework 10 years would make a huge difference if it's a project like creation of SSM and this is just one of the aspects where we would need to look into much more detail in terms of what cooperation really means to have a meaningful benchmark in any case what I would also join Giovanni here look into and perhaps you could comment this how those countries that are those third countries treated a bit as let's say lower standards jurisdictions how they react to this status and whether they were being aware of this status whether they don't increase actually the benchmark I think we would have real life examples of such countries maybe Eastern Europe is one of those where some jurisdictions are perfectly aware that the subsidiaries located there might be treated in the way you describe and then ring fencing is a natural reaction so maybe a comment on this would be welcome many thanks okay please a lot of questions thanks a lot maybe working in reverse order so I would be interested in your real life example so we talked to Latin American supervisors and they are aware of each other so they know something happens in this country it's shifted into my country I react so there's this notion of interdependence it's clearly there and I think there were two questions on I mean related on kind of overlapping cooperation so basically our data set is dynamic we are updating over here so it's a better but basically the banking union would be a one because if you just look at the existence so basically you want same you jump from 0 to 1 so if cooperation agreements evolve and come better, so you wouldn't Yes, there was some comments on the type of agreements. So we have some limited information. I think we can do more work on this. So the previous paper, we split it up. And we found that each individually significant, but we couldn't find significant differences in the effectiveness. But yeah, so maybe we should move on. So basically, Giovanni on the consolidated solution, I think it's probably just a language problem. So I didn't talk about consolidated solution. I meant the consolidated banks in terms of accounting. So I'm not looking at any of the subsidiaries. So I'm just looking at the total group, the risk of the group. And I consolidated parent and subsidiaries. This was the statement. And there we found it is effective. So it does work. Except so the overall sample, except when we split up and we look at the real big cross-border banks. So Harry had a question on kind of the indigeneity of cooperation, right? So kind of employment. So it needs manpower to kind of implement these corporations. So I agree. So one of the concerns is that cooperation is endogenous. That's why we focus on this paper on third country effects, right? So it's driven by third countries. So it's not driven by Argentinians, but by third country signing agreements with the UK. So basically then these endogeneity concerns, I think they're less, you can still find some concerns, but I think they're less first order. Yeah, David, thanks so much for the comments. I think it was a very nice set of comments that we can take on board. So data collection, because the second paper we were kind of a bit brief. And in this paper, I think it's good to know that people want to see more, because it was like a multi-year effort to collect this data. But basically, so we have like a sequence of sources. And we start from the relevant authorities, the national authorities. And we look for information there. And it's all imperfect, but what helps us is that we have several countries, right? So if it's at least about other agreements of one country as kind of patchy information, we can still hope to get it from the other country. So in many cases, this helps us that we actually can get these agreements from different sides. Yes, I agree. And you mentioned a very important point on cooperation and bank failures, right? So that's something we also thought about. So it would be really nice. So we look kind of at ex-ante risk, right? So it's like the Z-square of banks. But it would be really nice also to show that this kind of also leads to less bank failures. But then we thought, yeah, we don't have too many kind of cases where we observe failures. So we thought about, in the end, we're not too confident to run regressions on this. So it was really on our mind. But the only thing we did is that, basically within the context, again, in the first paper, so we said, OK, so this risk effect on the Z-square, does it go down in a crisis? Or is the effect we estimate in normal times in terms of risk reduction, is it still present in a crisis? And the answer is yes. But it's not about failure, it's just about balance sheet. Thanks so much for your comments and to everybody else.