Unlock eSignature Legitimacy for Supervision in European Union with airSlate SignNow

How to eSign a document: eSignature legitimacy for Supervision in European Union

welcome back let's uh now go through the presentation of the third paper in this session this is titled is titled restoring confidence system system systemically systemically important Banks SSM effects on Bank performance presenter is Bard Ron paper jointly produced with Michael uh Sigman from the National Bank of Austria thank you for inviting me as it was said this is join work with M Sigmund and I have to mention that he did all the hard work he did the hard work with the data and did the estimation so I had to privilege to do much of the conceptual work and writing this thing up so obviously he's the older one of the younger one of us uh so uh the views are our own views and and now I think we can can jump into the presentation so the this structure of the presentation is I give an introduction this will be very short because uh when we present this paper at academic conferences you will be surprised people don't know most of nothing about the SSM so you have to talk a while and a while about the SSM they don't know much about it I can skip this basically I talk only a little bit about the data uh we'll spend much time on our empirical strategy because uh this is talk is a little bit different than usual because we use we use uh methods conceptual methods which uh were originally developed in computer science and and are slowly moving into economics but um expect are not very well known and so I take the chance to introduce them in every earlier version of the paper we refrained from from putting it in and talking about but we in the end we decided we have to make it clear how we came to our identification how what is behind what we are doing so to be completely transparent we've we've now introduced uh took much more time to introduce the identification of the effects but talk about the results which are very in favor SSN people will really like it so so uh so I think and the results of uh which I was very very uh very glad to see that the results are perfectly consistent with with the results of the two earlier papers so that's also kind of nice to see that they used very different methods different data sets but in the end the results are are consistent and then uh I also want to take some time to talk about to explain a little bit how we how we try to explore the sources behind SSM effect what nature are these effects and then a conclude conion will be very short so as I said the introduction uh can basically uh be skipped you know all this uh so uh I leave switch second Slide the uh the thing that I want to mention here is all the which is on the slide is well known to all of you but uh to our to our uh wording we say to the significant banks these are banks above 30 million billion euros which or which are economically important or or uh or have significant crossb activities we we simply call them SSM bank which is obviously technical not correct all banks are the SSM Banks they're significant less significant but for most readers it's easier if you say SSM banks are banks supervised by the ECB directly that's how we how we use it it's it's it's a short it's a shortcut you know uh so we are well aware that this is not technical not correct but it makes the presentation easier um we also know many large Banks were heavily affected by the by the crisis the two crisis so we asked did the SSM which was a reaction to the crisis help to improve performance in soundness of the SSM Banks and what are the main sources of this these effects we take we look at three uh outcome variables with simple return on asset on risk weate can average risk rate and the return of risk rated assets which in some sense is a combined measure of performance which which uh combines income with risk taking to some extent some sense and as I said we we want to we want to uh explore the resources of SSM FS or likely likely in sources uh what we are we doing so we estimate SSM effects what what uh is different to other papers we estimate direct indirect and total SSM effects I want to explain this to you in a minute and to do a bunch of robustness checks to see if our results survive these tests and explore the sources in a in a separate section so what we find in a nutshell is SSM has positive effects on return on assets of SSM Banks it is negative or no impact no impact negative to no impact on risk rats depends a little bit on the model we using it has a positive effect on return on risk rated asset may come from more income less risk taking or both and SSM has mainly Direct effects so this means uh translated more confidence better risk management things like that not so much doesn't result so much only from adjustment much much comes from confidence and these more uh let's say uh qualitative sources of or reasons of SSM effect uh so in the end we come up with finding that uh the components of the returnal risk rated assets when we uh when we look at them they clearly suggest uh more confidence and better risk management in this is M so these are the results in a nutshell the data uh come from the official SNL Financial database we have uh unbalanced panel data we have about 2,600 Banks about 116 as SSM Banks we do some data cleaning uh with other data resources Bloomberg eurosat ECB I'll just show you now a picture which is now in the appendix of the paper but it's simply to show you uh how how the uh the three outcome variables look uh develop over time for SSM nonm these are size adjusted triggers so the effects of size has been have been already removed from this St data uh this is for people who do different and diff they look they look at these pictures uh very very heavily but here you just see to get an impression how the data look for SSM and non SSM Banks I will come to all this uh in a minute I skip the summary statistics you see Bank specific variables we're using uh for certain in certain models we also have Micro variables which are not here in in this Statistics but I will explain why we use and not use these variables in what we are doing in a minute so now I come to the Perle strategy which I will spend most of my time because I have the chance to to show you something many of you might not have seen that often before so what I put here on this slide is a is a csal graph so this Cal graph is a stylized graph which so shows the basic identification problem so the point here is you have the SSM effect you have the SSM this is the treatment and you want to you want to measure the effects on the outcome variable which is the Y variable and there are Direct effects this is the red arrow and there are also indirect effects they go VI variables M mediating variables variables uh that react where where the bank for example in our case these are these are variables where the bank can react and the effect of TSM is transmitted via VI this reactions to the outcome variable we have uh criteria for becoming an SSM Bank this determine these are determine why you are an SSM bank or not why you are significant or not and we have a bunch of we have economic environmental variables this could be micro variables structural variables everything and in this interplay the goal is to identify direct and indirect SSM effects the indirect SSM effects I start with the indirect one this is the effect as I already explained which are running we are the mediating variables we think of these variables as variables under control important variables under control of the bank so have an example here SSM Regulators Force Bank to hold more Capital which may affect the outcome variable easy example we have a direct effect this is what I we call like confidence effect for example Market think or customer think the SSM banks are better are better supervised than on SSM Banks and this Mak deposits and enable cheaper funding by the way when we think about SSM we think about that the ECB is the tougher and has more resources and it's probably the toughest supervisor than National authorities so this is the this is the let's say kind of background how we think about SSM because SSM as was mentioned before is can can mean many things but this is how we how we basically see it and then we have a total effect this is the direct effect plus the indirect effects so this is what we want to identify so now to do it you have have to introduce some things how you can how you can how you can uh uh come up with uh with an identification strategy before I do it I have to shortly introduce uh three causal patterns which which you should know so behind all that there's a deep Theory coming coming from computer science and a beautiful Theory which is uh beyond all this which behind all this but which I cannot cannot explain here time is great books about it and it's really really interesting so but we only need a few elements of all of this we need to know causal PS and they're talking about three of them there is the fork there three variables X Y and Z and in the first case the fork as the name says the variable set causes X and Y so and this leads to spous correlation so the variable y set leads to spous correlation between X and Y in the chain X uh causes Z and this causes y this is the like in indirect effect this is the chain and so the variable in the middle mediates the effect from X to Y the collider is a completely different animal the collider is a constellation where X and Y are both causes of Z and Z is so Z is a joint outcome of X and Y why I'm telling you this because the effect of conditioning in a model is is different is different depending on on a constellation you're having if you have a fork if you condition on set if you hold set constant that's what you do in a usually want to do in a regression you block the the Cal path between uh X and Y and you remove this bous correlation that's what you want to get an unbiased estimate of the effect in the chain if you condition on Set uh you block the path between uh X and Y this s to kill you're killing the effect you're blocking the effect you're killing it if you're doing this in a collider is completely different if you condition a collider on a joint outcome you're creating correlation so this is because you uh you introduce selection Biers this is important uh if you don't if you don't uh condition the collider things will be as they are if you condition it things will become dependent so and you will introduce buyers so two more things that I'm basically done with this uh from this what I told you before you can derive blocking rules you can you can think about how you can block Cal causal Pathways between variables if you as I've already explained if you have if you have a if the if you have a chain for example or a fork you can break you can block the pass by by conditioning on this variable uh W which is which is either in between on the path if there is a collider you don't condition on it so you must not condition on it otherwise you open up the path so this is important this is basically everything you need to know about this and the backr backlog Criterion simply I don't want to go into this in detail simply says that uh that you should remove any spous correlation and you should leave all the paths between the uh treatment and the outcome open this is what and if you do this you get you get the identified effect this is what it in a nutshell what this says so this is enough now now so one more slide so also don't go into the detail but I just tell you if you want to uh identify the total effect uh interestingly this is the easiest thing to do the total effect indirect plus indirect effects uh requires only blocking for the selection criteria nothing more this is size of the bank and economic importance so this economics importance we capture because it's unobserved with a fixed effect type thing this is all you need you don't need any other control variables to do that if you want to estimate the direct effect it's becoming much more challenging because you have to conditioning on the selection criteria for becoming an significant Bank you must condition on the mediating variables but in the if you do it if you do it in that minute uh you open up another path because this is a collider so you also have to control you find something to shut down the path and so in this case you have to also to control or micro and environmental variables so estimating a direct effect is much harder than estimating a total effect which is much easier indirect effect is difference between total and direct effect in our case because it's linear our model are linear in general this will not be true in our case we make things simp very things very simple this will be the case so now I'm done with this and so that is the logic behind behind the identification we are doing so to recap if you want to have a direct SSM effect it's efficient to control for size and unobserved effects if you want to have Direct effects you have to control for a bunch of other things different things so now as was uh mentioned before and discussed SSM we measured SSM by a simple D this is obviously can subsume anything and this obviously a proxy if you would have a better variable better proxy it would be nice if you do it but in the end it will always end up with some kind of proxy if you measure measure the SSM or aspects of the SSM and this was also discussed that many other things went on during this so if the SSM would be in isolation and we could completely uh then we would have unification would completely be done but here we have other things going on so we have to think about it a little bit this has to do because we use a dami not identification strategy per se so single resolution mechanism this applies to all area Banks became effective in January 2016 so we find SSM effects before that since the SSM became effective so this is a sign that this is not a problem low interest rate environment was also before the SSM it affected all banks to some extent and uh the SSM effects stay there when we control for Direct interest low interest rate environment parcel 3 was already mentioned has been gradually introduced there were o buffers things like that some uh puzzle packages change risk rates uh but they effect survive when we control for for in some another way for these things uh so uh I want to also make a prbo test uh with a treatment period 2010 2012 uh 11 with data until 2011 we find no effect we also look at strategic self selections some have argued that some have tried to reduce their assets uh to not become a significant Banks this does also doesn't affect it so we have three types of models uh we have a fixed effects model we have uh another model called theice model which is uh the first two models these two models use within variation only it's important because everybody's talking what's the control group in the first two models the same units are compared before and after treatment the other the other uh Banks contribute to estimating the uh beta parameters on the X varibles but do not directly contribute to the estimation directly contribute to the estim estimation of the SSM effect so there is no separate control group in the first in the first two models the the units before after treatment so they are their own controls in the first two approaches the third one is a standard diff and diff model that what what most people do here is different here use uh here use you you compare two different groups so you compare compare within variation of treated and untreated for treated and untreated group so this is conceptually different in the first two cases you could run the regressions without even without any untreated Banks it would work it works you can do it in the third case you can't do it because you do it if if you need a control group so to clarify this the results in a nutshell are we have we just to time wearing effects and as I already uh mentioned uh the effects are positive for uh for uh return on on assets we just show effects for uh fixed effects in the fice model we don't use a diff and diff model to compute Direct effects because logically if you if you introduce more control variables you make the control group uh more and more let's say you you Bel for more and more dissimilarities is the more where to put in for more dissimilarity you allow in your comparison so we don't want to do that so we used fixed effects in the F model the F model has the has the other advantage that you don't need a some such things as par CH assumption and so on so this is you don't need to assume it in it's a more flexible model has his own uh things so for the uh risk weats we find they go down in in the in the uh fix effects model which by the way the results are very similar to the diff and diff so if you look at the fixed effects diff and diff and fixed effects give similar results P model are a bit different here in the other cases the results go completely in the same directions uh the return of risk weed assets is also qualitatively similar whether you use a fixed effects or a f model so positive effects on on on uh onal risk assets so as I told you we had made did a number of ress checks looked only at large Banks core periperal countries there are some difference differences but the effects don't disappear self- selection as I told you uh we also made a placeo test and we also made resampling uh similar in a way like like cross validation did a huge resampling exercise to see whether the results are robust against outliers things like that so the results survive all this all these tests so in the skip this one is these are just uh in the end we also wanted to look at sources of of this SSM effects we already saw the indirect effects are very small most of this is is direct effect so there's a huge chance that this will be confidence and better risk management qualitative things but we also broke up the return of risk rated assets into components and did run the F model for each component the F model because we don't need a parallel tent assumption so this is less restrictive in that respect that's why we used it in this exercise and as you see here the results in the nuto uh deposit uh Lan growth was positive which was found in the other paper uh also loan loss reserves went down this was also found in the other paper operating expenses went up a little bit so tougher supervision is more costly so this went up so the effect on the net interest margin was was basically uh close to zero so this were the effect all this all this uh risk rates are yeah equal or a little bit lower so so we think this uh shows the banks take take interpreted Banks take less risk so uh okay so uh skip the results as well because these are simply the regression results so uh already uh told you basically the the picture showed you what the effects are uh again this these are just the regressions uh but strong positive effects on on on on Lan growth but this is a really large but this has also to do which which with the way how how device model computes the counterfactual don't want to go into into detail but it has to do with with the conture looks like in this model uh so the negative s SSM effects on the loan loss reserves with the positive effects on Bank Landing suggests as we also saw in the other paper that Banks can uh land more without increasing exposure to risk your borers this was also Central result of the other paper so this leads us to assume that this are improvements of risk management so I now come to the conclus conclusions and I don't want to repeat much so in the end that SSM had positive effects these seem to be quite quite robust at least in our in our setting and uh we uh our exploration of the sources of the SSM effects lead us to conclude this is mainly uh more confidence in SSM bank so uh from from Market participants customers and better risk management which which might might explain the results which are probably consistent with the results so in the end uh uh we we conclude at least the SSM improved the performance and the soundness of the SSM bank so it's a I think up to now it's a big success story and so yeah this this is how we how we we how we what the data tell us about about there a m effects in our setting so thank you thank you thank you very much Bart uh discussion is Alex for the AC all right uh good afternoon everyone any thanks to the organizers for the invitation um uh pleasure to read this paper uh important paper as you saw already uh with very pleasing resultss for SSM supervisors um these views are mine only that's always an important uh disclaimer in this time in particular they are not the views of the ECP most certainly all right so the motivation of this paper is that uh European Banks were dangerously weak in the early 201s uh largely due to uh the twin financial and sovereign debt crisis um contributed to the Sovereign Deb crisis and that created the danger that we would enter some sort of a permanent Doom Loop between sovereigns and banks at the time caricatures like this one were uh omnipresent American tourists looking at the pza tower must be a bank so we needed to do something as policy makers uh because Bank weaknesses have multiple consequences one is individual and fin systemic Financial instability the other is reduced ability of banks to support the real economy which we think of as a primary function of banks so the banking Union was concocted and it has three legs the SSM the single resolution mechanism and the European Deposit Insurance all of these aimed at stabilizing Banks uh it's not yet complete as we know and the evidence is not fully conclusive on how it has affected the banking sector and the real economy so this is where the paper steps in uh analyzes the effect of the introduction of the on Bank risk and finds that it has a positive effect on Key Bank parameters such as return on assets income lending risk-taking uh result is mostly direct they have this very interesting scheme of uh separating direct and indirect results so their claim is that it's mostly via higher profitability and lower risk rather than bya portfolio reallocation uh it's stronger in peripheral than in core countries something that Bart didn't have time to mention uh does not predate the SSM and the conclusion of the paper is that the SSM has contributed to the recovery and stability of SSM banks by enhancing confidence in their soundness so something that we sort of hoped for when the SSM took over very pleasing uh my assessment of the paper is that it's very interesting paper on a very important question we need more work like this the analysis as far as I can say is very rather solid the results are believable of course need to discuss it so I have some quibbles and uh now every discussion ideally should be uh useful to the authors and uh interesting to the audience uh I've noticed that it's very difficult to achieve both things simultaneously so I'll try to achieve them sequentially so I'll have first part of my uh discussion where I'll be I'll play the stern referee uh trying to give you some advice as I see it on how to make the paper more the analysis more robust so everyone over the age of 45 should recognize Jack Nicholson in The Shining that's how I imagine my referees look like when they're rejecting my paper uh and then in the second part of my discussion I will play the absentminded philosopher everyone under the age of 45 should recognize laavo xic one of the stars of modern philosophy so here I'll give you a bit of a sort of Bird's eyee View and um big picture um ideas about maybe your next paper Okay so uh first Theory uh your paper lacks motivation uh what what do we expect to find uh and and as we know the theory is ambiguous um on the one hand supervision by local entities may be more rigorous uh this argument has been made by the literature local supervisors can monitor better the banks or they may extract Superior information that relates to Miguel's discussion this morning uh on the other hand of course centralized supervision may be more effective because it reduces the risk of banks engaging in some sort of crossborder Arbitrage uh higher supervisory Independence breaking the Doom Loop and so on and so forth so theory is ambiguous and I would have liked to see some sort of a framing of where your results fall Within These competing theories um sort of lead me from the beginning what am I to expect and why why should I expect it uh second uh broad comment is on your sample um so you have these exercises where you don't have a control group and I I have to say I really don't understand this well because we always need a control group the idea is that the control group is the the counterfactual right how the treatment group would have behaved in the absence of the SSM so in the exercises where you do have a control group um what I worry about is that the SSM banks are many fewer and they're much large so what we need to do is we cannot compare all banks to all banks we need to make the sample a little bit more symmetric look at the smaller maybe SSM Banks and the larger uh less significant institutions so closer to the threshold uh you have something like that in Table Six column one but I think it should be your Baseline sample um comment second comment is that SSM and non SSM as you call them so siis and lsis uh differ across multiple Dimensions not just size uh and I would like to see that so I think in your summary statistics you should compare them uh uh give me show me averages of of the various variables Bank specific variables that you reported and how these differ across the two distributions and I think you should also again in the test where you do have a control group uh do some some little bit of matching uh so that you choose pairwise similar Banks based on their observed characteristics again the whole idea is to make the control sample and the treatment sample more similar with each other because we know that the the two types of banks are not um my third comment is about the empirical model so there is a Time varying variation that is to all kinds of banks within a country and I think you should include Country Time fixed effects in all of your regressions because then we run the risk of somebody mentioned it this morning comparing large French banks to small Slovenian Banks so what you really want to do is compare within the same country the siis and the lsis again looking at those that are closer to the threshold because otherwise the effect becomes contaminated by all kinds of stuff which I guess in your computer science uh inspired approach you try to control but I'm more old school I would think that you know without doing this demand growth opportunities all kinds of omitted variable bias comes into play so it needs to be taken care of um some effects are observed already in 2013 as you showing your charts so that's obviously the effect of the comprehensive assessment SL asset quality review in 13 and 14 so I think you should maybe discuss that and make it an explicit part of the analysis and maybe look at a short sample rather than starting in 2005 when the world was a very different place um finally the effect might be driven by a global Trend whereby larger Banks became increasingly sound once the sovereign debt crisis were resolved was resolved for reasons unrelated to the SSN globally speaking um stock markets were booming you know whatever the reason so I would suggest a placebo test where you identifying sort of banks that would have been supervised by the SSM and banks that would not have been supervised in the SSM in countries that are not part of the SSM and run the exact same test for those uh and if you don't find anything then it's really an SSM effect if you again find the difference between larger and smaller Banks then it's some sort of a um global global effect that you are attributing to the SSM but it's actually has nothing to do with okay now my uh big picture comment um it's about growth and stability and uh the the the big philosophical question about growth and stability is are they complement or substitutes and the economic literature has looked at this for a long time and hasn't reached the unanimous conclusion so there is one strand of literature which believes that stability generates growth so in that sense growth and stability are compliments uh there is a famous AR Ry and Ry paper from 95 where they show that countries with lower volatility grow on average faster and they explain it with this idea that stability breeds growth whereby um by sort of making the return on investment more predictable investors invest at a higher level and that generates growth so and stability is a great thing for growth uh there is the opposing however argument uh ransier and cther in the QJ argue that maintaining high growth basically necessitates the occasional crisis because both both crisis and growth are driven by the same forces which are let's call them risk-taking uh the the the willingness of people to take on risk to fail you know some projects survive others go into the the drain and uh and once in a while you have the occasional crisis because too many people are trying to do maybe crazy things but uh uh this attempt to to lead to the to the long to long-term higher higher growth so in that work world if you clamp on the forces that are responsible for the occasional crisis you also reduce long-term growth so in that world there are substitutes so I've always wondered if we can use the SSM implementation to learn something about whether the growth and stability are complement or substitutes and I'm using the same ccure as um the previous speaker yes this morning which means that either great minds think alike or that fools rarely differ uh but I was more interested in the thank you it was a caricature from the financial times where Martin wolf I think uh the title was are European Banks too feeble to Spur growth and he was thinking exactly about this do we need to bank to make Banks safe and that would lead to more growth um but we don't know the answer to that and and U the truth of the matter is that the last decade uh has has seen the most c growth in the history of Europe since the second world war both in terms of gpn in terms of um productivity and I'm not saying it's the ssm's fault it kind of coincides with the introduction of DM it's not the ssm's fault but maybe we have moved too much in the direction of putting stability on a pedestal and we have you know uh built this Cathedral of stability on the back of the European economy and maybe it's dragging it down uh so what I would like to know you show that that the SSM has made Banks safer but I would like to know how this this has been translated into the the real economy have we you know have the safe have Banks become so what has this safety come how has it come to be uh is it because we have eliminated uh dangerous management practices at Banks or is it because we forc Banks to to always land safe and that would be inconsistent with with the idea of long-term productivity growth so it's a big picture common it's not for your paper uh but but I would like to think of stability not just and safety not in an isol ated sort of way but as a part of a bigger bigger picture which includes long-term growth uh okay so in conclusion important timely paper I liked it I learned a lot uh SSM banks are safer and more profitable good to know we need more work like this but again put in a larger context I would think good luck publishing your paper and getting your message through thank you thank you very much Alice I can tell you that has been both very constructive and very entertaining for the organ to achieve your goal okay uh ra over there first I in thinking about uh the impact on profitability I can I mean you have taken the approach of looking at accounting measures return on assets but of course there are Market measures and it is well known that that European banks are not performing very well in terms of the book to market value and and so I think they would be interesting not perhaps for this paper but to look at the SSM effects from the perspective of Market measures of profitability let me just follow following question to this one because another alternative would have been actually to look at return on Equity to the extent that uh actually Capital has gone up uh but they they have gone up in different ways for different typ types of banks that should have a sort of opposite effects on both return asset return on Equity I know whe you have tried actually to look at that where the results change just hold on okay you will repair later over there please thank you um so I have two questions one is um related to the effect of rised assets larger Banks tend to use internal ratings more intensively so you should maybe try and disentangle where there is a the effect is driven by the intensity of usage of Vis v stand that approach as smaller Banks than to use the the the latter more than the former and this is first question the second question is as far as I remember uh you put on the slide as for the the the say the the source of revenues that affected the the returns by by more of the this was were fees so probably if you because you introduce a matching procedure at some point controlling more directly for business model will help understand whether again is a large is a size effect or a business model effect I think think this is important to take into account thank you more questions can I a question as well on the on the indirect effects it's very interesting you say that Direct effects are very important Direct effects you define d effects as sort of a an impact of confidence which does not translate into standard radios with sheet variables uh so youbute to to those are effects an important component of your results but then at same time you explain your result by saying that an important driver of a higher profitability for same Banks is actually that they they actually lend more how do you link confidence with more lending this is not totally obious to me uh okay on other questions afterward have still a few minutes do you okay so uh from just one one word to the control group issue this is the classical issue because this is uh to you because I don't can go into the details but if you in a different if setting you have a control group is separate in in in a panel model you your own unit a control group if you look at it uh you will you see it so and by the way we we got the same results with and without with different approaches so the control group the different seems to be okay still okay it can be made better better But the results in the paper they don't differ really much so so in that sense we are we are confident that this is this should work but what well taken points well taken you can always come up with better control groups uh those is certainly another a paper so that's way way too broad for us to to look at in this paper book to book to market value uh we didn't we didn't uh look at look at that we we uh we looked at we have another paper which is uh cost of equity which a very new one now where we look at cost of equity of SSM Banks and also find there that the cost of equity went down was good for good for cost of equity but it's very new work it's it's it's in we have a first version of this no no cost of cost of equity we didn't look at return on Equity cost of equity so this is uh this is this uh risk red assets uh we controlled for that what you said in the regressions I didn't show it here but in the regressions when we compute the Direct effects we we exactly control for different approaches that were taken to to compute risk rat so this is in the paper I didn't tell it here for for some reasons this is done but business model sure I mean if you look at if you look at uh this indirect effects and also qu qualitatively if this this is a point but as as as was as I tried to explain when you want to compute the total effect you don't need to explicitly control for that if you do it a direct effect you need to do it exactly so this is depend what type of effect you want to compute with your approach so it depends what you what you are after and the last was uh I think I haven't got the confidence the confidence and the link with credit the confidence and the link with credit so yeah that's that's a good question at least uh we uh we uh we found that that uh that uh the SSM banks have a have more deposits I think it's good good for for the for deposits on the one hand and on the other hand they they lend to to less risky that sense to less risky uh people so that's that's basically the same result from the other paper this is this is how we interpret it so but I have to think again I have to think more about about this point make it more clear sure any thanks more questions okay uh it's 1229 so we just some time for break for lunch let me just say that um has been a very interesting morning you have heard presentations that provide interesting empirical evidence that the SSM has not not only made Banks less risky better provisioned more able to give credit to a real economy but even more profitable but with this overwhelmingly positive note I think we can break for lunch so thank you thank you very much me yes