FSB virtual workshop on understanding and addressing systemic risks in NBFI - day 2

Key Takeaways

here at this uh third session um of this very important conference uh yesterday and today on understanding and addressing systemic risk in non-bank financial intermediation it's a great pleasure to see so many of you as as guests and participants in this discussion and it is my pleasure to share this third session today following the very important topical sessions that we had yesterday on the liquidity imbalances in bond markets and application for systemic risks and then secondly the interconnectedness in onbank financial intermediation today in this third session we have an opportunity to discuss data data and tools to enhance risk assessment and monitoring of the nbfi sector and of course no need to emphasize that this is extremely important for the financial stability board and the whole community of central banks and regulators supervisors because we are we've been over the past years so extremely engaged in actually monitoring what happens in the non-bank financial universe risk monitoring and analysis on non-bank activities is actually an explicit mandate of the fsb after the global financial crisis and on top of that for the last years we even had uh and have established the annual global monitoring report on non-bank financial activities as a dedicated channel for assessing and reporting non-bank risks and of course there is a constant need for us to work and enhance and refine the monitoring in in in this context and this is why it is so great that we can look at the latest thinking around assessing these risks and monitoring these risks um so we have important topics on the agenda for today and very distinguished distinguished experts we'll have three presentations to look at um first of all a presentation on systemic risk pro cyclicality in the european financial system uh uh which will be presented by peter cincinnelli and i'll say a word on the presenters in a moment and will be discussed by when kian huang from the bank for international settlements uh secondly we'll have the contagion from market price impact a price at risk perspective and third we will have the increasing corporate bond liquidity premium and post-crisis regulations so three very important topics that can help us cast light on the latest thinking uh around risk assessment and monitoring the first uh presentation like i said will be uh held by peter cincinnelli he's an assistant professor in finance at the department of management at the university of bergamo and in italy and an associate research fellow at the center for econometrics analysis based business school in london like i said the discussant will be when chiang huang she is an economist at the monetary and economic department and financial systems and regulation at the bank for international settlement settlements the uh second uh presentation will be held by matthias sudo he's the team leader for financial stability at the directorate for general micro potential policy and financial stability stress testing and modelling division at the european central bank his commentator in the session will be sylvia piccini is a senior manager on financial stability surveillance at the hong kong monetary authority and the third topic will be brought to us then subsequently um by um he's a phd student a graduate from new york stern at the in new york at the university and he's joining chk as an assist professor in finance the discussant will be yoon pan chair um professor and professor of finance at shanghai advanced institute of finance uh in china it's great to have all of you around the table uh and we're very much looking forward uh for to your insights that you are going to share with us and also to the discussions we'll have subsequently the way we will proceed in the usual format is that we'll hear uh each presentation followed by uh by discovered by the discussant and his or her remarks and then after the three presentations we will have an opportunity to open up the panel and discuss the topics in depth and also take questions from the audience so please in order to post questions that you might have in the context of this panel do not hesitate to post them and if i understand correctly mateo everybody is invited to post questions that they may have through the chat function uh in this call with that it's my great pleasure to hand it over to uh what we've what we're all looking forward to namely the presentations and the first presenter is peter plieter please get started thank you very much for sharing this with us thank you stefan you can hear me yes all fine and i share the presentation and i hope you can see the slides yes we can see the slides if you can turn it to full screen that would be useful thank you better fantastic okay thank you thank you let me thank the organizer for including our paper in this very interesting conference this is a joint paper with elizabeth elizabeth bellini and juvenile from bayes business school since the global financial crisis the financial system has undergone deep and remarkable changes on the one hand in the run-up of the crisis credit and asset crisis increased and deleted from that fundamental trend during such period of exuberance financial intermediary lending activity and their stock of debt are high due to an expansion in the aggregated demand on the other hand instead asset prices decrease the value of collateral diminishes and the borrower's profitability deteriorates as a consequence the level of the credit supply in the economy is reduced that is financial system is procedural um several authors investigated the consequences of leverage prosecutors in the financial system also there is a number of contributions arguing that higher financial leverage especially short-term leverage induces bank to engage in liquidity and risky landing as well as securities activities that result in the widespread failures that said in our research we test the following hypothesis first of all we test if the leverage of european financial institution is procyclical second one we test if the leverage of european non-bank financial institution is per cyclical and last but not the least we test if non-bank financial institutions are sustained risk for cyclical in addition to traditional banks we also evaluate the prosecutor considering the following superior pre-crisis subprime global financial and sovereign debt crisis and the post-crisis period let me start by presenting some issue that we will address throughout the presentation in particular so we need to identify first of all non-bank financial institutions or new bank financial intermediaries second how to measure leverage and third how to measure systemic risk for cyclicality and regarding the latter we will use three prominent systemic risk measures in the literature namely delta quarter margin expect shortfall and as risk regarding the identification of non-bank financial intermediaries we follow the guidelines provided by the financial stability board and in particular we start by definition of non-bank financial intermediation as a wide measure of all non-bank financial intermediaries such as all financial institutions that are not central banks banks or public financial institutions non-bank financial intermediaries play a clue a crucial role in the financial system and they are involved in credit intermediation process and as a financial stability board stated in december 2021 they grew considerably to almost half of global financial assets compared to 42 percent in 2008 and they become more diverse over this period moreover no non-bank financial intermediaries provide an alternative to bank financing and helps to support real economic activity however their growth implies that risks are increasingly being intermediated and held outside the banking sector with implications for global financial system resilience they can become a source of systemic risk regarding our data sample it's a panel of 590 european listed financial institutions between 2005 last quarter and 2019 last quarter regarding the types of financial institutions we consider both traditional banks and those financial institutions fully or partially outside the regular banking system and we consider 287 finance services 181 real estate finance services in addition to 129 traditional banks let's focus on the specification regarding the leverage process quality equation one shows the relationship between the change in leverage and the change in total asset where delta leverage as our dependent variable is the quasi-market leverage ratio growth defined as the market value of asset over market capitalization alternately we use we consider accounting leverage ratio growth as the ratio between total asset and total equity without considering um total asset valued at fair value um that aside instead is the size group as the natural logarithm of total asset and the associated beta2 coefficient that is positive means that an increase in asset valued at fair value leads to an increase in leverage leverage at time t minus 1 captures financial institutions reaction to the leverage level uh in the previous quarter while financial institution is a set of dummies capturing fixed effect for each type of financial institution while time is a set of dummies capturing fixed effect for each quarter we also specialize our analysis by testing how differences in the financial institutions business model could affect leverage management of a national institution in particular we have this dummy variable and bfi which takes value one for final services and real estate finance services together and zero for traditional banks in addition to evaluate the effect on leverage for cyclicality for finance services and real estate finance services separately we had two dummy variables and the first one is fs is a dummy variable taking value one for finance services only and zero the wise it means zero for traditional banks and real estate finance services and the rest is a dummy variable taking value one only for real estate finance services and zero otherwise let me uh go to the to the result and in particular let's focus on the full period we find that we find that leverage is procedural in the sample of the european financial institution as a whole and that good market and accounting leverage turns out hardly statistical significance however when we specialize the regression to consider the impact of the non-bank financial entities well how the results emerge and they are very interesting because the active pro cyclical management of leverage concerns not only traditional banks but it also extended to finance services and real estate finance services and in particular duress during stress periods let me now focus on different sub period and start with the global financial crisis during the global financial crisis if we focus on the quasi-market level leverage it is worth emphasizing that when the market leverage is considered neither traditional banks nor non-bank financial intermediaries are procyclical and this result shows that while financial institutions may actually manage accounting leverage market leverage is mainly um determined by market forces moreover we find that during the sovereign debt crisis only traditional banks should leverage prosecution and one possible explanation is that what the persistence of sovereign debt crisis and the wheat outlook for earnings made difficult for traditional banks to reduce their leverage ratio significantly further as for a systemic risk for cyclicality is concerned equation two shows um to what extent the change in the fair value of assets may translate in the risk capital financial institution management and in this equation as dependent variable we use the growth of systemic systemic risk measures namely delta var margin expect shortfall and as risk the rest of the equation is the same as the first equation and let me go to the um results and i want to focus on the pre-crisis period both finance services and real estate finance services are pro-cyclical during the pre-crisis period and the result confirms that the prosecution of systemic risk may be amplified in models with the value at risk constraints when volatility is low such as the pre-crisis period value at risk constraints enables higher leverage and size regardless of prime crisis instead we noticed that the prosecutionality of non-banking financial institutions vanishes and signaling that mainly traditional banks expanded their balance sheets becoming precyclical uh i'm let me focus also during um during the global financial crisis in particular the finance services during the global financial crisis procyclical in terms of systemic risk in addition to traditional bank and one possible explanation is that they become more and more related to the financial markets dynamic and that they manage balance sheets aggressively and utterly the same happened during the sovereign debt crisis instead during the post-crisis period we noticed that real estate finance services is much more procedural in terms of systemic risk in addition to traditional bank and to finance services we also conducted some robustness checks and and in particular we implement the transcend mother to study potential asymmetric effects in the level of leverage we also replicated our analysis after excluded financial institutions belonging to um big countries we validate the relevance of explanatory variables using bonferroni adjusted p-value and we also test for causality relationship between as a growth average and systemic risk by an an extensive grandeur causality intelligence panel analysis and all the results are confirmed let me conclude we analyze the relationship between bank financial intermediaries leveraging systemic risk we find that leverage is procyclical for traditional banks while it only becomes procyclical for non-banks in periods of stress and that in general non-banks contribute significantly to systemic risk as policy implications uh we argue that there is a need to carefully monitor in addition to bank law activities also non-bank financial entities which are involved in credit intermediation process uh which could lead to potential risks within the financial system as a further step we are also investigating whether the intermediary capital ratio defined as the value of market equity divided by the market value of equity plus book debt could be considered as a counter-cyclical driver or of systemic risk and in particular we want to address we are addressing this question do non-bank financial institution intermediary net capital ratio differ from traditional banks and let me show you quickly some interesting stylized facts and in particular let's remember that the intermediary equity capital ratio is the net worth of intermediary sector and it is considered as a key determinant of its marginal value of wealth and here we can see that an increase in the net worth leads to a decrease in systemic risk while a decrease which it means that a financial intermediaries is required higher compensation to take more risk okay we notice a decrease an increase in systemic risk we also notice that a different trend in relation to different financial institutions and in particular in order to conclude this presentation we notice that the intermediary net capital ratio is greater than in this case we only consider the delta covalent system in residential during tranquil market periods such as pre-crisis and post-crisis period and it is lower than systemic risk during financial market tournaments such as the prime crisis global financial crisis and solving that crisis so we are uh analyzing these uh aspects and that's it thank you for your for your attention and i'm available for a q a thank you very much peter this is hugely interesting uh and and very sophisticated um and before we move we move on to the second item of course we've got the opportunity to hear directly um a first view from a discussant in this case it's wang kyan guang from the bank for international settlements uh please please forgive me if i've mispronounced your name um but uh in any case we're very much looking forward for your comments when we went crying please thanks a lot stephan uh thank you peter uh i really enjoy reading your paper so this has been chiang paong from the bis before getting into the substance let me say the usual disclaimer applies the views expressed yeah online and not necessarily of the bis so i think peter has done a great job in presenting the paper i would just briefly provide a recap so this is a very comprehensive study of leverage and systemic risk in the european financial system i think the authors they have a really extensive data set on banks and mbfis so this includes some banks some finance service firms and also real estate financial finance service firms and authors also uh embark on the study on systemic risk measures following the literature and their key findings is that first leverage is procedural for banks and it is procyclical for mbfis especially after the financial crisis and then mbfis in particular the real estate finance service firms increase systemic risk so i think overall the paper is the is the is very rigorous analysis on the european financial system and i recommend everybody to read it my comments would mainly focus on the three things so the first is the practicality of leverage well so uh yeah the literature has shown that leverage of dealer banks uh is uh for cyclical so here on the left-hand side i took the uh one of the graphs in adrian and shin 2010 where they showed me the average growth is positively correlated with the total asset growth and i think that is the what has been captured uh by by peter and co-authors uh and on the on the on the right hand side show the uh another graph from adrian and sheen 2014 where it shows that you know like yeah of course if you have the you have the leverage as the asset over equity and asset is the equity plus debt if equity doesn't really move when you expand assets only that increases or shrinks together with assets then leverage growth would co-move with asset growth and the main reason behind that analysis shows that so their analysis shows that the main reason behind is the market risk constraints such as the value at risk constraint so i think all i think uh uh what the peter have shown is very consistent with the results uh traditional banks they are they they show really consistent positive uh correlation uh between the asset growth and the leverage growth i think although the regression results are very significant but when it gets to the graphical analysis it seems like the results will be driven by some small subsamples so i would suggest the authors to uh follow you know agreement shin 2010 and 2014 and maybe split the sample and focus on the banks and mbfi's that are more subject to market risk constraints such as risk contract for instance maybe the dealer banks uh maybe the the the you know the trading firms in the financial finance services i'm not sure to what extent you capture those entities but as long as it's possible i think it would be interesting to split the sample and focus more on those ones that subject to market risk constraint so my second comment is about practicality of systemic risk i think it makes sense to measure to measure prostheticality as the positive correlation between leverage and asset growth when it gets to the leverage discussion but when it gets to a systemic risk the definition of physicality is a bit tricky because here if we only look at the correlation between firm's asset growth and its contribution to systemic risk then i'm not sure what kind of what kind of physicality it is measured right because it's about systemic risk and um and and it's different from a leverage which has a more clear micro foundation and it's to me i find that physicality can be better defined when we discuss systemic risk so here i think a classic uh paper from the literature is boreal for fine and low 20 20 2001 there they said that a financial indicator is specifical if it tends to amplify business cycle fluctuations right so it's really important to define the business cycle or in the context of financial stability the financial cycle i think there's a growing literature on the financial cycle measures so for instance on the global financial cycle there's uh a price based measures which essentially would be the global factor based on the asset prices uh by miranda agrippino and co-authors and there's also quantity-based measures uh which could be the first principal component of the ratio of gross capital inflows to gdp for instance by uh other sorrow and co-authors and yeah additional measure could also be uh the bis global liquidity uh indicator and this would this would give give the the authors and also the readers a better sense about what kind of physicality we are talking about it's it is for security with respect to the global financial cycle or in some cases maybe we also interested uh in the in the physicality with respect to country specific financial cycle so in elder zorro and co-authors they also uh provide this kind of country specific measures so i would suggest the authors to request these systemic risk measures on these financial cycle proxies so that we can have a more meaningful discussion of practicality for systemic risk and then the last comment i have is about the impact of leverage on classicality of systemic risk so in the paper i think the author took a took empirical analysis the empirical specification in paper is that you split the sample of to the high leverage low leverage subsamples and then see whether whether the correlation between system risk and asset growth is more positive or less positive i would suggest to do a more comprehensive study here and use the leverage of banks or mdfis if the interest is on mbfi's leverage and use this the leverage as a state variable and see how physicality of systemic risk varies across states i think there's a well-developed methodology in smooth transition regression for instance the l-star model that has been used in the financial literature so given time strength i would not go into the details of the l-star but i think that methodology really fits to this kind of analysis that authors would like to pursue i'll stop here thank you very much wonderful thank you so much wendy and for these initial uh thoughts on the paper three very important comments um and uh peter i think we i would like before moving on to the next presentation uh we can uh look at peter whether he has any initial reactions to your comments that you've made peter well thank you thank you stephan and in particular thank you so much thank you for your time and for your for what you did and for your precious suggestions uh really it's they are very appreciated um in particular i had the opportunity to uh reflect some uh to think about the first the first point and thank you and it is an information that we have and that we can view in the course about the evaluatory constraint in order to select to identify some financial entities and thank you so much uh about the second point um i yeah we decided to uh to consider delta size as a main explanatory variable uh by looking at the uh main literal academic literature as an example adrenaline bruno mayer lopez and espinoza and so so for regarding the main drivers uh the main financial accounting which could uh have a sensitive relationship with the systemic risk and so we decided to use this kind of proxy and we thank you and we thank you also for this suggestion and in particular thank you for the third suggestion because i think and i totally agree with you that this kind of empirical analysis as you suggested will improve a lot in the paper and thank you so much thank you so much really thank you peter thanks for that first reaction um i see in the chat box there is a technical follow-up which might be worthwhile looking at um right away before we move on to the next presentation art partners asks please clarify how you work determine the causal relationship between leverage and asset size peter yeah thank you for for the comment uh yeah we start uh first of all uh following the adrian and sheen and jerome financial intermediation paper first of all and so we simply apply the that model and in order uh to test the causality relationship between delta leverage and delta asset side we also applied but i i did not have the time uh but uh there is a huge sensitivity analysis in our paper where we demonstrate that there is a closer relationship between total asset particular asset and the um delta leverage and so we we we conducted a huge sensitivity analysis and particular robustness check about that okay wonderful peter for that clarification thanks to you peter for the moment uh uh for the uh for the presentation uh went young for the first thoughts and comments on this one and now we can move on to the second uh paper we have the pleasure of seeing today it's on condition contagion from market price impacts a price at risk perspective it will be presented by matthias sudo from the european central bank thanks matthias for joining us and subsequently i'll call on silvia pizzini from the hk mate to discuss and comment but first of all over to you matthias thank you stefan many thanks for having invited me to to present our paper here just give me one second i'm trying to share now the presentation which hasn't worked before that's why i was silent we can hear you very well so that's uh that's already a good sign fantastic so let me see i'll do it from basil matthias that's fine yeah but i think i found it now let me just try because i have one disclaimer that i changed here i think it looks like my player is moving first so you are moving first okay all right okay pete uh matthias over to you thank you thanks a lot so i cannot see the presentation yet mateo can you share in it okay very good so the paper that i will show you today is work done at the european european central bank in the context of monitoring risks that are also coming from the markets this is joint work with gabor mikiel kaiser and lucamino narelli and all the views that are expressed here are the views of myself and should not be considered as views of the european central bank or the earth system just as a usual disclaimer now why do we care um we are looking at quantization from the perspective of the markets this paper and what is important in this context is we know that agents overlapping portfolios can provide a channel of contagion there risk stemming from this channel cannot be taken into account by any counterparty in the system though because they have limited information the regulator however can capture the full picture if data and tools are accessible and that's why we're also in this session about data and tools what i will show you is a tool in the end now in crisis situations in particular the modeling of asset deleveraging requires a notion of price impact that means you need to know what is the change in price given a certain volume of assets sold and that is in particular important for end for the mbfi sector now moving to the next slide um how do we model price impact so what is specific to our paper um is basically a regression framework that is based on quantum regressions leveraging upon the work done by adrenaline bronermeier and the kovar paper but also angel and manganelli in the earlier caveat paper we basically tried to extend the standard models available for assessing price impact which are kyle's lambda which is a linear framework or a square root framework from bouchol we go a bit further we say we try to estimate a right a wider range of impact severity levels from the data that is available to us at eisen level introducing the converging nature of the exponential function now this is seen here briefly in the in this little equation and i don't want to spend too much time on that practically speaking i mean r is the return of an asset and the small v is the volume and we regress on this with beta0 basically the volume traded but also the system level return of that asset which means if you have a bond for instance and you know what is the return of that bond you would not take only into account in this regression the volumes traded for that bond historically but also the return of the market where this bond is located meaning the country and the sector and potentially the maturity depending on whether it's bonds or equities so this is like say the high level concept now if we move to the next slide um i show you what this means in practice you will see on the left-hand side a plot where we show you for a specific security or single security which is representative for many securities on the x-axis the volumes traded and on the left-hand side uh access you see in the scatter plot the changes in price and what you can observe there is a few outliers so these are the dots that are more on the right hand side but the bug of the changes are in the space that um where you see this is clustering and the green ones are the positive ones and the blue ones are the negative ones now we concentrate in our work on the negative impacts and on the right hand side panel you see again using empirical data different lines that represent different quantile quantiles of our regression framework that i showed you before and you can see depending on where you are in in this quantile spectrum which goes from the median which is 0.5 to the tail which is 0.05 so that's the the lower yellow bar we are able with this regression framework to capture [Music] also the more extreme movements and what we have in mind here is basically to go a step further by using very granular ice and level data and instead of having a homogeneous price impact parameter in the analysis of fire cells to come up with a solution that is a bit more heterogeneous and takes into account also the historical movements of individual assets while at the same time taking into account as well the market movement why this system level returned and if we move to the next slide i give you an example for corporate bonds with a focus on what we call price at risk so we borrowed this from this name from the discussion that we have on the growth at risk framework for instance and um in the end given that we use the same methods here as in in the context of growth at risk we internally we came up with this name price at risk and what does it mean i show you here two bars for different segments of the corporate bond market and the market price impact from a 50 million euro fire sale and the blue bars are the more extreme quantile which is which represent the fifth quantile and the yellow bars the tens quantile and what you can observe is as expected at the tails or the blue ones so at the further going down into the tail because both our tail parameters um you have stronger reactions but then also if you uh think about or look at the different classifications uh down to non-investment grade you see that there is also a non-linear relationship as documented also widely in the literature and you see that this non-linear relationship increases even further uh when you compare the fifth versus the tens quantile but this is just some numbers to give you an idea what does this mean and now we can talk about these results summarized at different for different asset classes and maturities but in the next slide i show you basically what happens if you use that in a fire sale simulation and what i show you here is um the results from a model that we are developing since almost four years at the european central bank which we uh call systemwide stress test model that is now in this paper also used in a version with two sectors banks and investment funds we have it now with three so also insurance corporations where we basically allow for liquidity shortfalls in the system following a stress to the system where banks and funds are allowed to cover the liquidity shortfalls by selling tradable assets this is done following a pro rata approach and the asset sales generate a new price equilibrium where these price impact parameters that we have been showing you before are used to calculate the impact on prices until there are no further changes in the market values of asset holdings of the individual agents in the system and if we move to the next slide i give you a brief overview how this system operates but i will not go into the details so practically we have uh here two panels on the left hand side there is um what is happening in the force first order uh effect so we have a stress scenario that leads to some exogenous shocks for the agents in the system covering market risk credit risk income of financial agents this leads them to default or distress of agents in the system the securities issues issued by agents that are under this or in a situation of distress or default will have changes in their price and you have basically an overall impact on the balance sheets of these agents then in the second step and that's the right hand side panel we iterate the system via a simulation framework until there are no further losses by the agents which have to adhere to realistic regulatory constraints meaning capital constraints or in the case of funds given that there is not so no strong regulation they have to adhere to at least at least cash constraints um they interact uh with the possibility of interbank withdrawals unsecured borrowing redemptions from funds and so on and until there are no more losses now if we move to the next slide i show you basically the result of this when you are using um either homogeneous price impact parameters as standard in the literature where you would just use an estimate done for a given asset class and maybe country and then you see what is the reaction in the system and we basically contrast this with our findings from the usage of heterogeneous price impact parameters and what you see here basically is for different quantiles in the chart on the right hand side what are the losses expressed as a percentage of total assets in our system of banks and funds given an initial redemption shock so here's no scenario with a full narrative implemented but rather a homogeneous shock for investment funds in terms of redemptions of minus five percent and what is the reaction and um basically this is based on the system that i explained before so the redemption shock for funds triggers fire sales uh of all securities uh uh or that are held by funds um banks and funds suffer fire sale losses and these losses these fire sale losses largely depend on the applied price impact parameters and what we can see is that the risk that you estimate by using heterogeneous price impact parameters is more limited than with homogeneous price impact parameters and if we move the next slide i show you one other uh sensitivity analysis that we have done which in short basically reveals that if you look at the chart so you see on the right hand side horizontal axis different levels of redemption shocks so before we had minus five now there is a full spectrum of zero to minus eight and then we have different quantiles on the left hand side horizontal actual axis which represents the different um price impact parameters estimated by the quantity regression approach that i showed you earlier and then the losses uh on the y-axis and practically what we see is basically that there is a sub-linear increase in system level losses uh with the increase in redemptions and yeah this is uh in short given the the short time um uh the the overall result and in the next slide i conclude um briefly yes exactly so what we have done was we estimated security level price impact parameters for different arbitrary amounts sold we have introduced the concept of price at risk by using our quantile regression which we think is a useful complement to the standard average price impact parameters that are used in literature taking into account the heterogeneity across securities alleviates some of the risks shown by fireside models that apply homogeneous price impact parameters historical data cannot explain the future but former crisis episodes can provide an indication of the severity of future price movements which are affecting the liquidity of all agents in the system and what we think is important is that scenario-based multi-sector so not only for one single sector like say banks only but several sectors at the same time multi-sector stress testing frameworks based on granular network models can shed light on pockets of vulnerability in the financial system and yeah this is my conclusion thank you thank you so much matthias for sharing this with us um great insights and you can see in the chat box that that we already have two questions um uh it's from christoph and sean um and and maybe we keep these questions later on for the discussion because they have also a wider bearing on on what we're looking at and this might be very useful in any case first of all uh we're looking forward to hearing the initial comments from from our discussant sylvia petzini from the hkma please start it over to you right away sylvia please ah thank you thank you stefan and thank you to the conference organizers for giving me the opportunity to make some remarks i enjoyed reading the paper it's a very interesting empirical contribution and it analyzes the fact of large-scale portfolio deleveraging on the price of bonds and equities and using actual data traded on the market and also models how the leveraging sharks would propagate through the financial system by modeling institutions holding similar portfolios actually i'm using the the incorrect word not by modeling but actually by using actual data on actual portfolios which are very much overlapping as the as the authors show and it matters a lot to any analysis nowadays trying to understand the vulnerabilities of mdfis first of all because the paper touches on the exposure to common assets which are a very powerful conduit for indirect interconnectedness and second because it models fire sale dynamics which is really about how the sale of securities affects uh the price depending on how much the vol of the volume is sold and we have seen in the great financial crisis for example that prior cell dynamics can really tip entities from liquidity stress into solvency stress now the the authors do an enormous job at combining granular data so they have access to the ecb securities holding statistics so they have access to the portfolio the actual portfolios of bonds and equities held by banks and investment funds in the euro area they also have access to the ecb centralized security database which is an eyes in level if uh has id level information on bonds equities and issuers and also by combining other databases both commercial and non-commercial they have they combine the daily traded prices and the volumes of the largest 10 000 securities by value for every trading day for three years and they also go very if you like granular in terms of methods so they estimate the price impact at security level they use quantile regression to understand the non-linear impact and i love the chart the rainbow chart that you showed in the at the beginning of your presentation um it enriches the model by adding this systematic component in order to model the fire sales and in the results it exploits several dimensions of heterogeneity and this is really to model what happens in real life now if we move on to the next slide the results are powerful because they show that equity portfolios are more prone to cascading effect than bonds because it also shows that towards the end how moving really in the equity space those cascading effects are potentially the largest in stocks issued by small cap non-financial corporations and this is really what speaks to the archagos episode in march 2021 in which archaegos had invested indeed in small cap non-financial corporations and so it makes it very relevant and it has some other results that the that matthias has taken us already through i think this is an important paper that makes an important contribution in a very um in a very empirical sense because it allows us to put numbers on effects that are analyzed and observed and it starts to model those in a more um perennial way and it exploits the rich heritage of real life distribution and impact and it's very important for practitioners for example in central banks and if i move on to the next slide i would like to give just a couple of suggestions that probably you have already thought about but the first one is why don't you try to merge it with give given the access to the various ecb data if you have access to the tr data it would be a great addition i think to start to incorporate the derivatives as well into the this price dynamics so by selling derivatives do you have the same impact on equities and on equities particularly and the second idea i was wondering is in order to validate your results it's true you say that there are not good historical data and there's not going to be anything like the data we have today but i wonder if you could um get hold of one of those commercial data that are available now for research purposes on uh central unit central limit over the books in order to see because in 2008 that's when you really had fire sales uh with a debt that we haven't experienced in the last few years so i wonder if you could get some external uh validation through that way and i know my time is running out so i'll be very quick i just wanted to take the opportunity to share a bit about the experience in the next slide about what we have done at the hkma this is really about trying to find an empirical framework in order to focus our attention on the mbfis that are most vulnerable so once you have access to data you have thousands of amplifies in your in your in your financial system so we needed a method to to have those most important and most vulnerable to emerge so over the last year we have worked on a framework that brings together various risk indicators and basically boils them down into an impact indicator and to a vulnerability indicator the vulnerability indicator tried models um the portfolio characteristics in derivatives so these amplifies their interconnectedness their leverage uh tries to capture the market news and sentiment score and has an element of macro environment and all of this is really a device to boil down the complexity and to sift through thousands of mbfis in order to identify the most systemic that then go on for more qualitative work and um at the point if i move on to the next slide that the point of contact here is really that i have admired you your herculean effort of putting together all those data and we have really um experienced something similarly similar in combining the data in in our framework for derivatives the otc derivatives that we see in hong kong the bank lending activities also granular data that we can combine we try to combine the supervisory data we have as well as some other public data sources and so this is really um yesterday we had a lot more theoretical papers if you like but i think this paper that you presented matthias is very important on the empirical front and i want to to conclude by basically inviting the attendees to read it to understand better the fire sales dynamics that's all from me thank you so much uh silvia uh for these comments um very relevant thoughts i'm looking right away at matthias was there is anything initial you want to respond or whether you want to keep it for the greater discussion later on no i mean let me maybe just thank you briefly for the very nice discussion and your kind words and yeah i think we have a few questions also and uh i'm happy to discuss uh in the format that you prefer stephanie okay supers and let's keep it uh for the remaining time in the meantime we will move on to our third paper that we've got the pleasure of looking at today it's on increasing corporate bond liquidity premium and post-crisis regulations uh by boto wu at the new york university um later on uh jung paoten will take a chance at discussing the results but first of all let's look at buddha's results thank you very much brutal for for joining us and for sharing this with us we can see the paper on screen we can not yet hear but how you're still muted okay can you hear me now now now it's there wonderful yes okay so let me uh share the slides uh okay um thank you so much for having me here so um this paper the motivation of this paper is really about this debate between regulators and practitioners about copper bound liquidity after the financial crisis and this debate has been around uh for for over a decade after the financial crisis so um so after the financial crisis um practitioners have constantly been complaining about the lack of the liquidity in the corporate bond market they blame the post-crisis banking regulations for disincentivizing the dealer banks from actively market making and providing liquidity to market participants but on the other hand such lack of liquidity fails to show up in the traditional transaction cost measures such as the bidar spread so the corporate bond realized that without spread has remained low and comparable to the pre-crisis level so regulators and practitioners now debate whether the post-crisis regulations truly harm the corporate bond market liquidity and if and if the bilat spread is no longer informative it is worth asking whether there are alternative measures that could truly reflect this rising illiquidity that practitioners have been complaining about so while while many existing studies have chosen to uh to to look at this issue via various forms of transaction costs my paper took an alternative perspective right if liquidity has really deteriorated after the financial crisis then investors should require a higher premium for holding illiquid bonds so the liquidity premium which i define as the liquidity component of the credit spread should become higher so how do i measure the liquidity premium i follow the standard asset pricing literature to extract the liquidity premium using this very simple cross-sectional regression so every month i regress the credit spread on the bidder spread on controlling for various bound characteristics and the credit risk controls then the liquidity premium is simply the product of this cross-sectional regression coefficient lambda and the beta spread now if you run this regression you can see the following result on the on the right on the left panel i plot the average uh corporate bond without spread and as you can see that the post crisis without spread is comparable to the pre-crisis level however despite the low bi spread well the cross-sectional variation of the credit spread has become more and more sensitive to the cross-sectional variation in the pita spread as you can see on the right panel this lambda coefficient now what's more ah the increase in this regression coefficient is more than the decline the beta spread so the product of the two the liquidity premium has also increased more than 20 percent of the credit spread uh is now due to illiquidity uh compared to 10 before the financial crisis so from the liquidity premiums perspective it is fully consistent with practitioners notion that okay the liquidity has gone uh has gone worse now what's more the liquidity premium also has interesting variations across different regulatory episodes so as you can see that liquidity premium is a fraction of the credit spread has significantly increased for bbb rated and speculative bonds and the liquidity premium of the speculative bank has significantly increased following the vocal rule but compared to the pre-crisis level the liquidity premium either as a fraction of the credit spread or in terms of the absolute magnitude has increased for at least the bb rated and speculative rated bounce now keep in mind that if you run the pull the regression of the copper bomb without spread you won't have a positive sign here if anything you'll probably get a negative sign now does the low bidder spread contradict with the higher liquidity premium that i document here so here i'll use the standard otc model to reconcile these two similarly contradictory patterns right in the model you have customers and also dealers dealers have this market making capacity v which is directly affected by the dealer cost decay now customers can trade directly with dealers as some poisson rate of alpha which is an increasing function of the dealer's marketing making capacity so if dealers do more market making customers meet dealers more often so dealers basically buy from sellers and sell and sell to the buyers and they charge a bidder spread for such liquidity provision services now alternatively customers can also trade with each other via other brokers now the brokers do not serve as counterparties they simply allow customers to meet and trade and because brokers do not serve as counterparties they charge a a low spread a low bill spread so that so simplicity let me just assume the spread is zero there and the customers meet with each other via the brokers at the poisson rate of beta so how do we map this into reality so basically dealers can use it computers can use their balance sheet to trade directly with customers that is the alpha route but alternatively dealers can also allow different market participants to match and trade among themselves instead of taking upon two to the dealer inventory and that is the beta route so let's look at how regulations can change this dynamic right regulations increase dealers market making costs forcing dealers to reduce their market and making capacity so customers meet dealers less often and dealers will charge a higher beta spread compensated for such uh rising dealer cost but that is not the whole the full story right because as dealers reduce their market making more fraction of the trades will be done on a brokered basis and those broker trades are involved with with zero spreads so the average bidax bread can remain flat now the contribution of my paper is that from the investors perspective when dealers reduce the market making inevitably investors will experience longer trading delays they will they need to wait longer time you know for for the trade to happen and because investors not experience longer trading delays they now need to uh they require a higher liquidity premium so the analogy one can think of is that suppose new york city levitt attacks on all real estate agencies and suppose the tax is so high that all agencies all agencies just shut down then you can still find apartments by contacting renters directly and when you do find apartment you don't need to pay any intermediation fees because there are no intermediaries in the market and so the intermediation fee is automatically zero but because there are no intermediaries you'll not need to spend a longer time searching for apartment so here i plot this time series on statistics and as you can see that as more trades are being matched and when dealer costs go up the average bidder spread can remain flat now of course the beta spread charged by the true dealers also known as the cost of immediacy in this literature will go up in the dealer cost but the problem is that the cost of immediacy is not a continuous measure of liquidity you cannot measure the cost of immediate cost of immediacy continuously in all instances of time right because when the dealer costs are high more fractions of trades are being brokered so that means as economic tradition it is more likely to end up measuring the average beta spread so that's why the existing studies have to rely on rely on various sorts of stressful events in order to measure the cost of immediacy but during normal times customers do not request for immediacy they would rather wait to get the trade stock so the illiquidity that practitioners have complained about is really about this longer trading delays and because they experience longer trading delays they require higher liquidity premium now the model also gave a theoretical justification of the cross-sectional regression right once i introduce the bound heterogeneity basically i can show that i can write the liquidity premium as a linear transformation of its beta spread with this linear coefficient independent of this asset heterogeneity okay so that's why that cross-sectional regression work in the sense that once you control for the credit risk the cross-sectional regression of the credit spread on the bidar spread will produce a positive regression coefficient and once you multiply the two you can uncover the liquidity premium and then the model implied the regression coefficient will goes out will go up in over time in the dealer cost so the results suggest that you know there is an alternative there is other dimension of liquidity that has not been captured by the transaction cost dimensional liquidity and that is the trading delays now the problem is that trading delays are not directly observable in the in the existing data the existing data like trace only requires a realized transaction level uh uh trades so this motivates me okay to estimate those unobserved trading delays that are implied by the magnitude of the liquidity premium so basically i want to know how long investors need to wait in order to support that level of liquidity premium that i observe in the data so i here i run a very simple structural estimation um basically it is in the similar spirit as banking out the implied volatility from option prices on you using a black source model same philosophy and once i've done one once i've finished this exercise i show that okay the implied trading delays in the market has increased as you can see in the picture here and i think the magnitude is within the range suggested by the practitioners so according to some anecdotal evidence trading delays have increased you know from less than a day before the crisis to over a week after the crisis and finally you know with this true measure of liquidity now let me revisit the ongoing debate between regulators and practitioners and understand the impact of regulations on corporate bond uh liquidity so i look at this puzzle 2.5 which introduce incremental risk charge and stress valued risk to account for the default and migration risk of a credit of the credit pro of the credit products and i use the yield volatility to proxy those risk charges because it is a standard input of the vr calculations now as you can see that as the regulation kicks in the liquidity premium of the of the more volatile bonds diverge from the liquidity premium of the less volatile bonds and to translate the impact into trading delays the impact is on a magnitude of almost 10 days in the interest of time i'll just uh skip the literature okay so so let me just conclude so i think you know over the past decade you know the illiquidity in the corporate bond market has been summarized as as as this notion of uh loss of immediacy and the existing studies have have relied on the transaction costs or the trading activity variables to proxy this loss of immediacy but if you think about it the loss of immediacy literally means refers to a time dimension of liquidity it is all about trading delays so my paper essentially provides a way to extract the trading delays to extract those unobserved trading delays through the length of the liquidity premium which is easily measured measured from the cross-sectional regression so trading delays and liquidity premium are essentially the same thing they are like uh implied volatility and option prices and using the liquidity premium or the trading delays allow me to to quantify the impact of the post-crisis regulations on corporate bond market liquidity thank you so much thank you so much potel for sharing this uh insight into corporate bond liquidity and in particular the time dimension we've got um as a discussant uh now uh waiting um um you i think who will take the opportunity at uh providing comments yum is joining us from the shanghai advanced institute of finance and i'll hand it right over to you for your thoughts thank you so much europe see my slides yes we can hear you and see the slides thank you so much okay um thanks for thank thanks uh mateo for inviting me to discuss this paper um it's um it's a really interesting paper and uh bottle uh photo has uh has done a good job in explaining it the paper was is also very nicely written so i agree to discuss this paper really because i was very much intrigued by this cross-sectional sensitivity measure which is lambda so i see that jan jenny is also in the audience i'm sure she that measure uh also is very close to home uh for her as well because the first time i saw that measure was at uh as uh as in her job market paper so i will have more to say later on so as bottle said that there is a tension or a puzzle between somewhat a decreasing which is on the left panel in the plot a decreasing bit ask spread meaning that liquidity in the corporate bond market is improving at the same time if you look at that uh cross-sectional sensitivity it's increasing okay so so the reason why we're so intrigued is that i really want to figure out what is this lambda is really measuring bottle use it as a measure of liquidity and i would argue that this might not be so so i would have have more to say about this particular measure of lambda what is really uh it plays a pretty central role in in this paper so and also in jane's paper so i wanted really to get to the bottom of that measure of lambda but before i do let me summarize of summarize what uh what about balto does in his paper so the key insight as bottle said dealer can function as a either as a broker in that case they're matching trades or they can function as a market maker in that case they hold inventory and provide liquidity so the the tension between market a decreasing uh uh uh bid ask spreads and market participants complain about um loss of liquidity or loss of immediacy really happens around the post uh 2008 uh crisis the regulations that come after the 2008 crisis in particular basel 2.5 announced around 2012 it effectively increased dealers balance sheet cost so they are not willing to trade or to hold inventory of corporate bonds anymore because it's very expensive for them so that in that case that the immediacy or the liquidity is uh reduced and that's the main insight of bottles uh paper and the bottle then looks at a sequence of empirical results uh one first one is the liquidity premium increase which is the lambda increase the plot you just saw earlier and it takes longer to finish a trade the delays in trading due to uh dealer's unwillingness to provide uh immediacy and then balto actually he he didn't go through but there are some nice different diff tests using the basil or the vulcan rule timing uh interacted with the affected bonds either the bounced volatile or these bonds issued whose underwriters are the affected the valkyru affected dealers so i would just as i said earlier um all of these results really builds on this london measure so i want to say more about i want to discuss spend most of my discussion on this lambda measure so if you look at that lambda measure and this is the first intuition i had is that you group all these bonds together it's a cross-sectional variation of credit spreads regressed on cross-sectional variation of bid-ask spreads so so when you see large variation cross-sectional variation in credit spreads you're going to be able to pick up a larger lambda so it's important that you separate these bonds into different samples before you run this cross-sectional regression so if you look at that picture when bottle separates the sample into investment grade and the speculative grades you see that not surprisingly speculative grades whose credit spreads are much higher whose cross-sectional variations are much bigger picks up a much larger lambda so you you can see the order of magnitude investment grades from zero to one speculative from zero to four point five okay another very important observation here is that you don't really see too much of a trend in the investment grade was the increasing trend of lambda is really driven by the sample of speculated grades for investment grades if you look at when they pick up it's it's picking up when the cross-sectional variation of credit spreads are back or when the there is a credit crisis to 2008 2013 2016 okay so in other words the post 2012 increase of lambda t of this cross-sectional sensitivity is driven really by high yield another measure is trading delays well this is a very nice uh story and i like the model a lot in picking up the channels which this uh immediacy could a lot of immediacy could happen which is trading delays but if you look at the top two panels of trading delays and compared with against the bottom two panels which are the lambda you can see that these trading delays are really a product through the model of lambda measure so in a sense these trading delays are not directly measured these are not empirical measures these are output of a model and the key input again is the long-term measure another angle i want to say is that what what about the fraction of brokered trades because when dealers are not willing to behave as dealer as market maker they are behaving more like brokers then you will have a higher percentage of broker trades so i'm highlighting uh the the important timeline before and after 2012 june 2012. well on the left panel these are the lambda measure you can see again for uh for the increase in this with bottles as liquidity premium it's really driven by speculative grades you see in a jump from 0.98 to 2 for the speculative grids not so much for uh investment grade but if you move on to the right panel you look at the fraction of the broker's trade you see that through that transition actually the brokered percentage of broker trades actually decreased meaning that it's not as severe or the unwillingness of of these dealers to code inventory actually decrease somewhat so that's kind of counter to the intuition that you you like to pick through the model um so so let me so these are the empirical facts that i want to kind of uh point out uh in the paper so what are my my thoughts on this lumped-up measure it's not really a standard test of liquidity risk premium because if you really want to measure this liquidity premium there is a standard test you can do you get a risk factor you measure the beta then you look at cross-sectional variation in expected return and it's linked to beta and yet you see in the literature especially in the recent literature you started to see the emergence of that that cross-section of sensitivity or that london measure for example in this paper it's dealers reducing provision of liquidity in jane's job market paper it's investors increasing demand for liquidity if you look at unconditionally a positive or significant number is actually well established it's an indication that liquidity matters for credit pricing but when you take it as a time series variation of lambda then this could be driven by many factors for example whenever you have an increase in cross-sectional variation in credit spreads and that increase in cross-sectional variation could be driven by a market-wide credit condition or you could simply be driven by a group of distressed bonds with explosive credit spreads so they are making the cross-sectional variation of credit spreads much bigger at the same time if the desk spreads do not increase by the same proportion you're gonna pick up a lambda that's increasing but this is not if you don't have a timely control of um of credit risk then that cross-sectional regression would yield a higher lump but this is not really driven by liquidity per se it's actually increase in credit risk either market wide or by a few bonds so taking this measure interpreted as a liquidity i have my reservation here and also post uh 2008 crisis um bond samples are getting bigger so you you you're gonna have a bigger cross-section so a bigger cross-section means a bigger cross-section of variation credit spreads and it might not have the same effect on bid ask spreads so you might pick up a higher lambda okay so i was told to conclude so let me conclude interesting topic um i agree with the hypothesis it's more delay it's observed in a few other papers as well um central to that empirical analysis is the lambda measure i'm not yet convinced my suggestion is that i would like to see more direct evidence of trading delays or cost of higher cost of in immediacy or dealers retreat from market making so sorry for running over uh a few minutes thank you so much for your comments june uh i think we can uh go straight to in the last five minutes that uh we have uh maybe to a couple of questions that we see from the audience we've got 270 people excuse me in the audience um and you've engaged very actively in the chat box so thank you very much for that and um and let's take back to matthias's paper um on uh that we saw um on the non-bank financial intermediation um and uh the way this has um actually um interacted with a price impact and we're grouping up just highlighting a couple of the of the questions that all of us saw in in the chat box um so apparently one theme seems to be uh what is the non-linearity uh between past returns and asset price sales um and also uh cross-asset responsiveness uh those questions came from christoph at the at the ici um sean uh asks um what is the total asset weighted effect of a shock on financial market is uh in general terms also if we look at the broader financial markets um and and then also oh let's let's limit it to to these two perspectives um matthias do you have a rough and ready short answer on those i will try thank you stefan and thanks uh two colleagues who have been raising these questions i think they are important now let me start with the non-linearity i tried in the meantime quickly uh christoph to see uh to have a look at your paper on overlapping portfolios now what i can say on the method that we have so non-linearity is captured by two ways uh first of all we use quantum regressions as a framework that are at the tail capturing the non-linearities because we use all data that is available for the estimate you see it also when we fit these lines through the cloud we are capturing the tail but we are still converging because we have tried other methods like convex hull for instance which is also documented in the paper where we just pick the very extreme cases and there's no other distinction the other thing is we have uh the exponential uh function in uh embedded in our framework which is also capturing uh non-linear effects these are basically the two sides of non-linearity now regarding the cross-correlation of markets so there are different ways to handle this and we tested this as well so normally when we look at the system level return that is also included as a regressor in our framework we think that this captures best the overall market movement you have uh cross-asset correlation so between equity and bond market but that is not that clear and also under in crisis situations not always the same so after some tests we figured out that we get like say from our perspective most meaningful results by keeping the system level return to the market that is relevant for the individual eyes and not including something else we could do that as well but as i said so this uh at least in our first analysis short bit blurred results and not so critical clear results as the ones that we have right now um then uh as another comment on um uh the benchmarks that i saw so they don't play a role in our case new benchmarks maybe the two other things that were mentioned by silvia so on the derivatives market indeed that would be a nice extension data wise uh it's a complicated thing though and we have not been focusing on this yet in our analysis because our system-wide stress test model where basically this paper that i showed you today is like a spin-off paper uh required more or less the the results for bonds and equities and we don't have derivatives included yet there so this would go hand in hand and then the other suggestion you had for um [Music] what's that now again i forgot it um yeah i stopped here in the interest of time uh i will try to get back to that thanks matthias and of course all of us are encouraged to get back into touch bilaterally in order to settle some of the questions that we have and that we can't address on these panels um maybe uh to boto do you have one sentence on the question that matteo asked would would actually market participants be willing to pay an extra premium on getting first in the line in these transactions what do you think you need to mute sorry about that so yes i think i i agree so that's why i was saying in you know presentation that of course nowadays the cost of immediacy has increased you know the beta spread charged by those market making services has has increased the problem is that as a as an economic tradition it is not quite likely to observe those trades right because more trades are being brokered so this is less likely for economizations to observe a situation when customers do requests for immediate customer knowledge simply wait in order to gather trades down yeah okay thanks for that very brief reaction brutal time is running out uh i need to conclude this has been uh hugely interesting and and great fun to to listen uh to your thoughts and and the reactions and also see such an engaged response from the audience first of all many thanks to the presenters in this session uh on data and tools to enhance risk assessment monitoring in nbfi um peter matthias bertao for your presentations and by the way also to all of your co-authors for the very important work you shared with us today this is important to us like i said also many thanks to the discussions when young sylvia and june for sharing your first thoughts and the comments on the three papers and kicking off our discussion i think we've seen important insights on both on pro cyclicality the market and price impact and the risk impact and also the rising corporate bond liquidity um premier uh three very important topics that are right on the top of our minds from the public sector um uh both at the easter beers we enhance our work on uh on nbfi monitoring uh but also i guess in the wider community of public policy makers no matter what they're from central banks regulator and supervisors and many of us have been on this call today to listen to what he said so many thanks for sharing this to us um list of thanks also goes out of course to all of the attendants here on the call for joining us and your interests and your comments and finally big thanks also to mateo and his colleagues at the fsb for organizing their support and with these thanks i close this panel and have the pleasure to hand it right over to the fourth session today martina moretti from the imf is hosting that policy tools and approaches to address systemic risk in nbfi um and let me wish all of us a great pleasure in in listening to that last session of the conference today marina over to you thank you thank you uh thank you seth and thank you very much um and again welcome to the fourth and uh and last session of of this conference like others i would like to thank the financial stability board for uh for putting together this very stimulating and important and timely conference and of course the secretariat for giving me the opportunity to share uh this panel now in in the previous session we have we have heard uh many interesting and useful ideas on on the practical implication of research for policy and and this session is specifically dedicated to discussing policy that authorities could put in place to enhance the resilience of non-bank financial intermediation from a system-wide uh perspective and we have three excellent papers that that will be presented uh on um during these sessions and each of them focus on one of the key areas of policies that we need to reflect on uh first on how to smooth peaks in liquidity demand uh second on how to increase the resilience of liquidity supply during periods of stress and and third on the role and potential unintended consequences of official sector intervention to address market dysfunction so in the interest of time this is a relatively short session so let me move straight uh to the first uh to the first paper uh which is on on financial fragility in in opening the mutual funds the role of liquidity management tools we have falco fact um as a presenter and andrew metric as uh is discussing uh falco is a chair of financial economics at the franco frankfort school of uh finance and management and also research professor for the bundesbank um his research focuses on the theory of financial intermediation analysis of financial systems and crisis as well as the efficient design of monetary policy instruments banking regulation and bailouts uh welcome farco and andrew metric is professor of finance and management at the yale school of management and the director of the yale program on financial stability his research current research and teaching is focused on financial stability including regulation of systemic risk activities of complex financial institutions and the causes and consequences of the global financial crisis and welcome andrew park on the floor is yours okay they are unmuted hello [Music] you can hear me you're fine okay sorry uh i lost my button so where are my slides now can you see my slides oh yeah there we are okay so thanks very much for putting our paper on the program it's really a pleasure to be involved in that very fantastic conference um the paper entitled financial fragility open and mutual funds the role of liquidity management tools it's john paper with loris enter and joanna paya um it is a very fresh paper so it's actually one of the first times that i'm presenting this so that also comes with apologizing with the discussions for having not really a very polished version available at that time um but it also means that all the comments that you might have for us are very very well received and will immediately find their way into into our paper ah where my slides now okay so i think with this audience i probably don't have to explain much about why open end mutual funds are uh are very fragile they are basically exposed to potentially self-fulfilling runs as the withdrawal of investors causes some trading costs but these trading costs are ultimately borne not by those investors that are withdrawing but they are born ultimately by those investors that remain invested so therefore it creates some first mover advantage and this first mover advantage might use investors if they expect that there will be many others withdrawing to withdraw two and creating some self-fulfilling prophecies and some some panic induced test so different liquidity management tools have been proposed and actually also been implemented to some extent already to contain these panics however there is only very limited evidence so far on whether these liquidity management tools have been very effective and in particular which of these which of these liquidity management tools are particularly helpful at the same time we saw during the covet crisis that there was a huge withdrawal of funds especially from bond funds and we are focusing here on the on the irish case and in particular islands with the second largest fund industry in the euro area experienced roughly 50 billion euros of withdrawal on march in march 2020 so a huge outflow of liquidity from those funds and that was particularly focused as you can see on the right panel on corporate bond funds uh in the in your area in in in in ireland now what we do in this paper we basically use this uh the crisis the covet crisis as a laboratory to understand which liquidity management tools were particularly effective in containing excessive outflows from on funds in particular we use a unique data set as we think on the availability of different liquidity management tools that is collected and provided to us by the irish central bank we show that funds with price liquidity management tools or software liquidity management tools actually saw significantly lower than outflows during the covet crisis so in march 2020 as compared to those funds that had only qualitative liquidity management tools or tougher liquidity management tools available we also see that this effect is particularly strong for those funds that had historically particularly severe outflows when they underperformed so that had a very high flow performance sensitivity since these are typically the funds that are considered to be relatively fragile our results also suggest that that these liquidity management tools are indeed helpful in mitigating mitigating this stress that is induced potentially by panics a few words on the the data that we have so we have a the investment fund statistics from the from the irish central bank that covers all funds all investment funds that are domiciled in in ireland that means in 2019 we had 1.1 000 bond funds to 2 000 equity funds and 900 mixed funds in our analysis we focus on a sample of 527 funds that are mixed and bond funds only but that held at that point in time also some corporate bonds on their portfolio the investment fund statistics of the irish central bank covers a variety of different fund characteristics that we use as control variables but most importantly what the statistic also covers is it reports the different liquidity management tools that the funds had available now what are these liquidity management tools that are covered in this in the statistics so first the price-based liquidity management tools there are two types that are reported here the anti-dilution levies and the redemption fees anti-dilution levies are here basically babies that cover the actual costs that are caused by the withdrawal of a investors or the transaction costs that are caused by the withdrawals they are imposed on the investor that is redeeming his shares in contrast to that the redemption fees they are simply a discount a pre-specified discount before any withdrawal actually occurs a pre-specified uh fee on the redemption uh the discount on the net asset value if an investor is withdrawing his funds uh from the fund um so these are the price-based liquidity management tools the tougher quantitative quantity based liquidity management tools that are also reported in the statistics are redemption gates so that means that the suspension is of of uh shares into cash is suspended if there is more than 10 of the net asset values supposed to be withdrawn on a particular day so that threshold can be set by the fund individually there are also temporary suspension of dealing and calculating of the net asset value and redemption in kind reported to that statistics as additional quantitative quantity based liquidity management tools but they are to a lesser extent used but we overall see is that most of these quantity most of the funds reported to that statistic use or report that they have at least one of those quantitative based tools available like the suspension like the gates but usually they are hardly ever used mostly because of reputational concerns presumably so therefore what we do is we focus with our analysis on the add-on effect of these price-based liquidity management tools so having these additional tools available what are the implications they are available on top of the quantity based measures so we define as a treatment group the funds in our data set that report that they also employ besides the tougher quantitative quantity based liquidity management tools also either redemption fees or levies whereas our control group is are those funds that have none of these price based measures available so need they use neither fees nor levies but have at least one of the tougher liquidity management tools the quantitative quality based liquidity management tools available so this is basically our treatment and control variable and then our treatment and control group and what we do then is simply we run a standard diff a diff regression where you estimate the net flows to a fund in a given month regress that on a variety of different characteristics of the fund and on fixed effects and time fixed effects and fund fixed effects but then we're trying to tease out what was the effect of the treatment of having the liquidity management tools the liquid the price-based liquidity management tools available in during the crisis in march 2020. we're estimating the coefficient of this interaction term of the treatment dummy and of the march 2020 dummy to identify or to see was there an additional effect that affected the net outflows of funds that had these uh price liquidity management tools available now what are our key results that we that we get from running this different diff regression we first of all basically split our sample into those funds that have a high uh a high performance sensitivity of their outflows and those that have a low performance sensitivity in their outflows the idea is here that those that have a high sensitivity in their outlooks are those that are particularly fragile and that are particularly susceptible therefore to a self-fulfilling liquidity crisis or to panic induced distress and if we particularly focus with our different diff regression only on those funds that have a high performance sensitivity then we do indeed find that during the period in march 2020 those funds that had also price-based liquidity management tools available they had larger net flows or put differently they experienced lower outflows right so the treated funds was had had also the liquidity management tools the price-based liquidity management tools available and a five percentage points higher net flows as compared to the average funds that had a three percent outflows all right so the outflows were muted due to these liquid to these price based liquidity management tools we do not see that the similar effect occurs also for the less fragile funds so for those funds that had a lower performance sensitivity in their flows there we do not see that uh the price-based liquidity management tools helped them to mute uh outflows but at the same time we also did not see that for those funds that had a lower performance sensitivity the effect of the crisis was hitting them as hard as the sensitive funds the the very sensitive the fragile funds experienced way more outflows than those funds that were historically less performance sensitive in their flows and the next step what we then also do is we split we dissect further and try to see to what extent the liquidity management tools affect the cross outflows and across inflows to different extents so we in the first column we focus as on the left-hand side and left-hand side variable on the cross-outflows and then the second column we focus here on the cross inflows and again use this interaction term to see whether there was a differential effect of having the liquidity management tools available during the covet crisis and what we see is that not too surprising the additional liquidity management tools the quantity-based the price-based liquidity management tools reduced the outflows during the crisis reduced across outflows during the crisis which is probably not too surprising given that for investors it was at those funds with price-based liquidity management tools more costly to redeem their shares other things equal they were probably less likely to withdraw their money from those funds so this is the first column is probably not too surprising but what is i think a little bit more surprising is that we also see a positive significant effect of those price price-based liquidity management tools on inflows so also funds that had these price-based liquidity management tools available experienced also compared to other funds a higher inflow so liquidity provision to those funds was higher and i think that can also be pretty intuitive because what it means is that those funds that had these liquidity management tools available also ensured new investors that they wouldn't suffer from the costs imposed on these funds by the massive withdrawals that they experienced by the massive cross withdrawals by the massive gross outflows that these fund funds experienced so apparently these liquidity management tools these price-based liquidity management tools can also help to to help funds in stress periods to attract more liquidity given that new investors do not have to bear the costs of cross outflows now as a robustness test what we then also do in the third column here is simply to estimate the propensity or the probability of negative flow so that a fund experienced an outflow just to see whether overall we have an extensive and intensive margin that we can document here and indeed what we see is that those funds that had also price-based liquidity management tools available they also were less likely significantly less likely to experience outflows during the crisis periods which also squares nicely with our other results that we have on the on the intensive margin that brings me already to the conclusion uh so i think what we can say will as with the first set of regressions that we have now documented in that in that paper and i think we need to do quite a bit more but what we can i think already claim is that we provide some first evidence that price-based liquidity management management tools mitigate the fragility particularly of funds that are susceptible to panic induced distress to those fragile funds that have a high performance sensitivity we also show that these fragile funds that had both price based and quality-based liquidity management tools available indeed experienced lower net outflows during the provide 19 crisis as compared to similar funds that only had these quantity-based liquidity management tools available and surprisingly also the net where the gross inflow of funds was positively affected by the availability of these price-based liquidity management tools so overall i think our results suggest that price-based liquidity management tools here seem to help to mitigate financial fragility of open-ended funds and seem to be a useful tool in increasing in particular the resilience of the most fragile funds with that i would be done and i think for thanks to you thank you for your attention thank you uh falco very much uh we already have several uh uh very uh very good questions in in the chat i'll i'll keep those for later when when we open up the debate so let me hand over to uh andrew thank you marina and thank you to the fsb and the organizers of this conference especially mateo has been working so hard to get us all together uh uh and ready for this um falco says that the that that the paper is not very polished i thought it was quite polished i wish my polished papers my finished papers looked this polished much less my early ones very nice job i thought that it was paper was quite well executed i'm not going to comment on the econometrics and make those types of suggestions but i've noticed lots of really good creative things in the chat uh there's a wonderful opportunity to have a hundred people referee your paper and give you ideas for that but i i was quite convinced and convinced of what uh and that what what is one convinced of looking at this paper which is it's really useful to have these price tools okay so there's price tools and quantity tools some of the quant many of the funds most of the funds have some form of the gates um or the suspension or both um but but only about half of the funds have some of the or maybe a little less have some of the price tools so being the either the anti-dilution type of of charge or just a fee and what is clear from the results is that it's quite useful in the type of crisis that we just saw the kovic 19 crisis to have exposed to have in place these price tools i it's quite convincing i think it's really would be hard to look at this and say oh no no no there's some strange endogeneity problem driving everything the authors have been as careful as you can be and uh it seems quite logical my priors would be that this would work and it's good to have empirical work that confirmed that so what i want to talk about is what what this means the questions that it raises these are not things that i think this paper needs to answer before it's publishable this paper is you know publishable now once falco polishes it to the degree he likes although i think it's fully polished um so one of the questions the first is if i were running a fund i would i would ask two things after i saw this i would say okay i'm convinced but number one um are there any anti costs to me putting in these these types of fees so why doesn't everyone have them i think they're quite logical that they would be useful i'm sure some fund managers would say well my investors don't want them now if i'm an investor who generally likes to just buy and hold in my funds i do want these things because i don't want people taking advantage of me but maybe there's a whole lot of other investors who think they're just so smart they'll get out fast and they avoid funds and so generally i think that this would be what they would want to ask you you've shown me there's a benefit is there a cost uh again this is not falco's problem but i think it's the next question for the literature the next one is should they have these quantity things lying in the background so what we've learned is that it's better to have price and quantity tools than just quantity tools well what about just price tools we don't really have a way to know that we don't really have the the data to look at that but i think the question people would now ask is can i get rid of my quantity tools do i need them because probably my investors don't like having a possibility of a gate or a suspension sometimes they're required by regulation i don't know the rules in in uh in ireland but uh uh sometimes there's kind of reasons that they do that but it seems like that would be a question then for us to ask uh which is should we keep the quantity tools if the price ones seem to work well at least in that crisis that brings me to i think a much larger question about policy these are issues where we've left the choice up to the individual funds and families about what to do in these cases but broadly speaking at the policy level we have far too little discussion in financial regulatory policy of this old old prices versus quantities question one of my dissertation advisors marty weitzman wrote just a brilliantly insightful paper published in 1974 just called prices versus quantities which has been applied very much in lots of areas of regulation most prominently in environmental regulation when we when we debate questions about whether we should have you know tradable permits for pollution and carbon or a pecuvian tax right and and the general conclusion that comes out of that literature that comes out of weitzman's paper and people understand is you know it's better to have price versus quantity tools can get you to the same place but if you're going to have one or the other what you really care about is the uncertainty about whether you know as a regulator whether the quant whether you want to have the right quantity whether you know what the right quantity is or whether you know what the right externality in price is and the uncertainty around those things that drive it in in financial regulation we don't seem to do this we we have you know from a from the policy perspective almost all quantity tools almost all quantity tools we tell everybody how much capital to hold how much liquidity they have to hold it's kind of a zero one thing uh if they go over and we have very little in the form of just kind of a tax and i don't really know why that is i mean obviously it's it's harder to do one or the other but i wouldn't mind raising some revenue from this large financial industry for all the externalities they create that might be a nice thing to do that's the kind of thing you can say when you don't have a disclaimer next to your name like that my opinion doesn't represent anybody else but me you can't say these things if you work for someone um but i wonder why we don't do that more and and this paper is a nice little thing showing how effective very straightforward uh uh prices uh price types of interventions can be rather than just the quantity ones that that's it i want to try to stick to my five minutes so i don't get in trouble but i thought it was great paper definitely add something uh to what we know on this topic and i look forward to the comments thank you thank you andrews oh these were all very very good points um we'll get back to them a little bit later in the discussion let me um let me move to the next uh to the next paper uh paper number two is inefficient uh repo markets we have a tobias a dealer as presenter and caroline sisoko as as discussant um tobias is a lecturer in finance at the university of bristol his research focuses on the economics of financial markets in particular institutional aspects such as the role of ccp ccps in the in many markets among others carolyn sisoko is a senior lecturer in economics at bristol business school not at the university of west england actually sorry about that her research focuses on macro finance and the economic history of financial regulation and current projects include work in this area in the u.s so with that um tobias the floor is yours thank you matteo can i ask you to um share my slides i seem to have problems to share sorry about that just give me one minute otherwise i have to exit the call and come back i think then i might be able but this might take longer than you okay thank you very much and apologies for this slight hiccup so um and first off it's a great pleasure and honor to be included in this fantastic program and this is joint work with lorianno manchini and norman sherhoff and i'm going to talk to you about um repo markets in inefficient repo markets and i guess to this audience i don't have to explain repos so let me straight go in with our main observations next slide so we start from the following four observations about um repo markets and um rico markets are an important short-term funding market with um daily outstanding repo volumes and in the us alone of regularly exceeding 2 trillion u.s dollars and of course the repos are collateralized loans so obviously collateral is is important but it's even more important and the liquidity of collateral is even more important during crisis times now what you can see in this figure are the repo operations by the new york fed and essentially there are three points which i would would like to highlight one is the 2008 ramping up of the repo operations of the fed this was um essentially post lehman um during the great financial crisis the the fed stepped in and provided indeed liquidity to this inter-bank market then again um in september 2019 again and the the market came to a complete still stand the fed had to step in and most recently at um in march 2020 we see the peak and highlight um this peaks in march 7 on march 17 you see actually the differences in and in collateral used at that point in time so there overnight and the the um the fed provided doubled to almost 500 billion us dollars of repos now um what is clear is that in in the lehman um occasion um this this was a a run on on on the otc market what is it less clear is which market segment was mostly affected during um september 19 and and during march 2020 at march 20 and i think the verdict on this is still out and i would actually be interested to know to see hear your views on this and but what i i think i want to highlight is that the different market segments at least in in 2008 responded differently to to this to this type of of a funding shock and um so this is our fourth observation and and then i would like to move on to the next slide so the natural questions to ask are what are the trade-offs between different repo market structures is there an optimal reaper market design and what does it look like and what is the role of collateral across different markets and what i'm going to show you is that these repo different repo market structures trade off the resilience to runs with the efficient allocation of resources and this trade-off is crucially impacted by the trading and clearing mechanisms in place which i show you on the next slide please so we broadly categorize markets across clearing and trading mechanisms we distinguish between direct and central clearing and between anonymous and non-anonymous trading and i'm going to focus mostly on otc repo markets in the left corner and ccp markets in the right corner and so otc markets you can think of the bilateral and tripartite the u.s customer repo market and ccp markets is it would be gcf repo the ficc the ux and frankfurt the lth in london now the big differences between these two is our [Music] anonymous trading in ccp markets through central order books so short cobs create more um asymmetric information between borrowers and lenders than not administrating in otc markets then what we understand by central clearing is essentially the existence of novation and default fund so innovation in the ccp market means that once um borah and lender have agreed on on a repo contract through the central order book anonymously the ccp steps in and innovates the contract which means essentially they're becoming the legal counterparty to both borah and lender and finally boris and participants of the ccp have to contribute um to uh what what is called a default fund ex-ante and before um based on their risk exposure to the market next slide please so what we do in this analysis is we um we consider a model in which we have heterogeneous boroughs and they're heterogeneous in terms of the acid quality what we are going to call long-term technologies and they show they finance these assets with short-term loans so there's a national maturity mismatch problem borrowers learn the quality of over about their technologies quality over time so you can think of lehman waking up on this in the september 2008 day and and and realizing that they had a lot of um poorly um poor quality and asset back security on there on their balance sheets at the same time there may be a subject and there may be a funding shock to the to the market and borrowers are also endowed with a risk-free asset which they can use as collateral what we prescribe here is or what we assume here is that a packing order between collateral and this long-term technology and essentially we are saying that and because collateral is is typically more liquid it is cheaper to liquidated than long-term technology and i think this is a very natural assumption next slide please so to summarize um at an initial stage um boris and lenders agree on a repo contract and then borrowers and use that repo to invest into uh long term into their long-term technology into their assets now in the second um around so at t1 what i'm going to call the rollover stage is when employers have to repay these initial loans and obviously they can do so by on by obtaining new loans by and that if that is not enough they have to start liquidating collateral and if that's still not enough then they have to start liquidating part of their long-term technology at the same time borders learn the quality of their long-term technology and there may be um the economy may be subject to a funding shock and what we mean by this funding shock is essentially that there's less funding available at this rollover stage than initially when borrowers invested so that there's a funding shortage and borrowers are actually forced to liquidate part of their assets and then finally sm the the payers of um and assets um realize next slide please so what we do here now is we are going to compare otc versus ccp markets and i'll i'll do that in the following figure and so on the vertical axis you can see the net welfare of this economy which is the joint profit of borrowers and lenders and on the horizontal axis you can see the size of the funding shock in this economy at this rollover stage now in an otc market there's a there's less less asymmetric information so that lenders can condition their repos on counterparty risk which means that lower quality borrowers are um subject or bear the cost of this funding shock and and therefore once there is they there they um they cannot fully roll over the the initial loans with new loans they have to start liquidating collateral so because of the pecking order that's the cheapest thing to do and once they run out of collateral that's the point kappa one half then they start liquidating um their long-term technology up to the point where where a run on on the low quality borrower occurs which we call fotc and that's the point where lenders correctly anticipate if they would continue lending to these low-quality borrowers then um they would not um obtain their their loan back and therefore they stopped providing funding and and borrows this low-quality borrower has to liquidate all their assets and that's where we see this deep drop in welfare at fotc now just um maybe a short side note is that for for for most of the part of our analysis we assume that um a priori and low quality balls have positive npv projects so next slide please so now we compare this to the ccp market and we we do this um step by step by introducing one feature at a time let's start with the anonymity feature which essentially provides an asymmetric information pro problem between borrower and lander so now lenders can no longer condition on borrower credit worthiness and therefore they essentially provide a one fits all loan to the entire cc and to the to the ccp um borrowing market that means that not only high a low quality borrowers have to start liquidating and collateral in case of a of a funding shock but also low quality high quality borrowers have to start liquidating meaning that the the funding shock is borne by by both high and low quality borrowers which which um um therefore um leads to this purple line where where we continue staying on this this high first best outcome now that that's the benefit of of of anonymity however the the cost of anonymity is that that then also high quality borders have to liquidate um their long-term technology which is the most valuable asset in this economy and that's why we see this steep decline in welfare starting from kappa one and um and so that's that's the that's the downside of of the insurance and that's that's you will tell me that's probably a well-known um effect but then we see that um the the anonymous market the purple line is is more resilient against runs than the otc market and indeed this fcob is larger than this fotc um where and at fcob however there is the the risk of having a complete market failure because lenders do not observe counterparty risk and therefore they might be hesitant to provide funding at all to this market and they run on the entire market next slide please this market failure can indeed be um prevented through innovation novation allows the cm and so i'm going i'm i'm going to focus now on this green line which is identical to the fcob line except for the last bit um starting from fcob so innovation allows um the ccp to effectively exclude low quality borrowers and thereby assuring lenders that the pool of borrowers they're lending to is of sufficiently high quality so that they no longer want to run on the entire market and essentially they're still only the same run on low quality boras just like in the otc market next slide please and finally to put the entire picture together we include the default fund which to which and i repeat um participants have to contribute ex-ante um based on their risk exposure to the market and the the the this this default fund kicks in in case of a default of a borrower and um therefore ensures lenders that even if they they might end up um having um facing a bad bad quality or low quality borrower um that there will be a repaid from this insurance from this default fund which then further increases the run resilience of the um ccp market next slide please so i want to highlight here that all our results and highly depend on um collateral being liquid and and so especially the ccp is re reliant on liquid collateral and and all what i've shown you really is really um driven by by this next slide please so what what is our um policy suggestion so we derive an optimal market solution which um prescribes two types of transfers a profit transfer similar or identical to the default fund we already know from the ccp so let me not spend time on this we already know how that works the novelty really is this illiquidity fund which which um requires borrowers um to to transfer their collateral into an escrow account which in case um a borrower runs out of collateral can be used by the ccp to to substitute their collateral and keep that borer running and so this collateral transfer resembles a little bit the collateral upgrade which has been done by the ecb and the fed next slide please so to conclude what what do we do in this paper we show that repo markets trade off the efficient allocation of liquidity with resilience to runs we show that this trade of crucially impacted by the trading and clearing mechanisms in place none of the existing markets yields the optimal welfare and we show that this optimal welfare can be attained via our two-tiered guarantee fund and we highlight the importance of liquid collateral for um resource allocation and resilience to runs thank you so much for your attention and i'm very much looking to the forward to the discussion thank you uh thank you very much to bs and uh i hand straight over to caroline to get to there we go can you hear me yes all right sorry um okay great um yeah thank you very much for inviting me to discuss this paper i really enjoyed reading it um i'm very interested in repo markets and there's a very interesting model of repo markets um i think because i want to try and keep to my five minutes i'm gonna say hey tobias did a brilliant job of discussing you know what the details of the paper i'm not going to review what he just went over um it's very um it's very interesting i'm going to jump directly to my comments and what i really like about this is how tightly focused the problem is here it really does an excellent job of focusing very narrowly on this decision to roll over a secured financing transaction and you know one of the ways that it does this is it basically neutralizes all of our dynamic readable market effects right you've got a direct three period model and there's one dead rollover problem in the middle period so that's nice and crystal clear right um there's an exogenous shock to funding liquidity and then collateral's market liquidity at day one is also exogenous at k1 so basically kind of everything is like these variables these variables we can just look at and i think you know the advantage of this of course from a theoretic modeling framework is that it's really useful for thinking through and characterizing what is the nature of the rollover problem and you know when can it freeze the market so i you know i i i do applaud this really simple and direct framework i do think though that we need to be cautious when we start talking about what are the implications for financial stability regulation right and in part it's because you know we think of repo markets the first thing we think of are the dynamic effects of margin and all of this kind of thing and our concerns about how destabilizing they are and so when you create this model that's really designed to neutralize all of the dynamic effects you also at least for me you make me stand back and say okay well now i'm not sure how much you have to say about financial stability right so i think we need to be a little careful about how this model addresses financial stability questions um and so essentially instead of asking how do we avoid episodes with funding shocks and asset price of liquidity it says okay repo markets have funding shirts you know repo markets have these issues of asset prices um dropping and they're just asking how can we make the market mechanism more resilient um to these funding shots and i will also note that there's kind of this caveat well of course it will only work if funding shocks are not too big there's this whole kind of discussion the first best and of course if the funding shock is too big then suddenly you're outside what you can hope for markets to do right and of course one of the questions is what what kind of funding shocks do we actually face right and so i just i you know we do need to have some caveats about how we take this kind of a direct model and use it to think about um financial stability um so uh i'm gonna talk in a moment about like just this idea you know we've begun to hit liquidity black holes like in 2008 like in march 2020 i'm going to talk a moment about why we didn't have in the past and why we have them now but i have two more kind of comments where i want in my short period of time i'm probably using it up um about the paper and one of the things that i think is marked more needs to be noted about this paper is your borrower participation or constraint requires a negative repo haircut now i have to admit this is something that puzzled me for a moment because i was looking at the setup of the problem and i'm like wait a minute how does this work as a repo transaction and of course a securities lending transaction is the mirror image of a repo transaction where in fact what's actually happening is you have somebody who wants to borrow a security and they're doing the exact same thing as you're doing repo but you think of the cash as being the collateral that's posted against the security being borrowed right and of course that's why you have a haircut that goes the other direction because the focus is on borrowing the security this is actually structured as a securities lending transaction and not so much as a repo transaction so i think we need to think about hey i'd like to see this model but like actually tackling the repo market where maybe what you need to do is take your risky technology securitize it and directly finance it on repo um but anyhow i think that's an important thing it looks to me a lot more like aig and maidenlink2 type structures than like what i would think is a repost structure and i think the other thing to keep in mind when we look at what this model is doing is that there does seem to be a really strong bias in favor of otc markets because they're treated as having full information on types when you know to me if i were thinking about modeling this i would be thinking about oh let's have a somewhat um imperfect signal about types rather than actually giving them full information because then they look really good right because they've got this full information and i also thought it was a little bit too interesting that your ccp's the only type of equilibrium you're looking at is the pooling equilibrium and this of course has built in for inefficiencies precisely because you're pooling you're making both types you know have the same uh have the same contract um when you actually have a separating equilibrium it's in there in the appendix you know we could have a ccp which was using a separating equilibrium and i just you know the um the efficiency results would obviously be very different if you gave us that analysis so it really felt a little odd to me that we weren't being given the analysis and separating equilibrium for the ccps um but you know so i i mean i love the fact of working out modeling these kind of repo problems but i do think um to really start to integrate how repo markets work and how we should think about the room for market um infrastructure we you need to kind of go into a little bit more detail about some of some of these more specific issues to really get at some of that let me just talk for a moment about you know why do we have this situation where repo markets are causing these liquidity black holes today and repo markets are essentially money market financing and long-term assets we know this it needs to be understood that the traditional anglo-american financial regulation was designed to put firewalls up to very strictly limit this type of finance right if you're going to look at how british finance worked in the 19th century the real folks doctrine is completely targeting this issue we don't want this kind of finance if you look at what the glass-steagalls act did to stabilize markets in the united states in the 20th century one of the first things it does is take securities loans and says banks cannot be intermediate why do we have repurchase agreements so you can arbitrage the existing law right um so we should realize there were policies that were put in place in pursuit of financial stability that protected us from the kind of situations we have now from the 1980s on on these productions have been eroded so we've completely rewritten the laws governing financial market collateral and where are we now in 2005 i'm thinking the bankruptcy abuse protection consumer prevention and consumer protection act in the u.s there's an fsa in the uk as well as um the european law around the same time we integrated derivatives collateral securities lending and repo markets and um and this smooth flow of collateral means that today any stress on any one of these markets is going to show up in the repo market we've been talking a lot about how in 2020 um it went directly from derivatives collateral problems to the repo market um and so i just want to bring up this idea that if we really want to tackle the repo market problem we've got to shrink it right you've got to be thinking about how do we make this a smaller market than it is i know i'm already going over my time there are a lot of ways to do that but i did want to kind of put that in there all right thank you very much very much enjoyed reading the paper thank you uh thank you very much carol caroline very um you know very uh useful and interesting uh points and let me move straight to the third and last paper that we have uh today um it is on exorbitant privilege quantitative easing and the bond market subsidy of prospective fallen angels and we have matteo crozillani as presenter and jane lee as um as discussed mateo is a research economist at the financial intermediation policy research division at the new york fed and his research is focused on financial intermediation and the transmission of unconventional monetary policy and regulation on credit supply and jane lee is um assistant professor of finance at columbia business school mateo the floor is yours in fact thank you for the invitation to present this paper um i hope you can see my screen i'm sharing my slides this is joint work with viral acharya ryan banerjee team eizert and rennes peaked and the usual fat disclaimer applies so the motivation of this paper is really coming from the observation that non-financial corporate debt is now the largest type of private debt in the us global financial crisis we've seen a reduction of the debt of the financial sector and in the household sector however the non-financial business that has been increasing now to be the largest type of private debt if we look within the umbrella of non-financial corporate debt we find that corporate bonds are driving this entire increase post global financial crisis and somewhat interesting even within the corporate bond space has been the entire increase has been driven by one particular segment of the investment grade corporate bond market namely the bbb or triple b rated corporate bonds what's so special about this bonds well this is the rating category that leaves just above the investment grade threshold that we know matters for regulation among other things so uh the u.s corporate bond market doubled in size between 2009-2019 and as i mentioned driven by triple b corporate bond market during the same time we show in the paper that the triple b corporate bond market has been characterized by deteriorating quality however as you can see in the figure on the right uh the yields on triple b corporate bonds have been compressing and getting closer and closer to single a and double and double a post global financial crisis if we look within the triple b market to see which firms are driving this increase actually they are the lowest quality triple b rate at first which we call prospective fallen angels because exactly these are the ferns that are lived just on the left of the ig cutoff and they would become an industry jargon fallen angels should they fall off the cliff well this happened exposed during covet what i'm showing here is the billion dollar volume of total asset downgraded from triple b to sub investment grade or the total asset of fallen angels from 2007 to the covet period and you see this huge spike in just a few weeks in 2020 uh at the onset of the covet 19 crisis and what's interesting is that you can barely see the 2008 crisis here so really an unprecedented volume of uh fallen angels during the covet 19 crisis let me summarize the paper this paper we showed that these prospective fallen ages again these risky firms rated triple b enjoyed cheap funding in their corporate bond market funding post global financial crisis and this privilege has been driven by the demand by investment grade investors such as insurance companies and the sluggishness of credit rating agencies the idea is that central bank qe induces the demand for risky triple b bonds exactly by investors that focus on the investment grade corporate bond market and on the other hand prospective fallen ages namely these risky firms above the triple b rating cut off meet this demand by issuing bonds and as i will show mainly to finance risky m a activity this matters for the macroeconomy because it creates negative spillovers and misallocation effects that affect competing firms that happen to compete in a market where many of these prospective fallen angels operate let me be more precise in the data how we identify the prospective fallen ages well these are firms that as i mentioned before are rated triple b and they're vulnerable to a downgrade how do we find whether a firm is vulnerable to a downgrade well we look at the firm's fundamentals and we rely on combining the balance sheet characteristic and income statement using the almond z-score which is nothing but a measure widely used in industry to basically measure the credit risk of companies and the idea is that we classify a firm as quote-unquote vulnerable if its z-score is lower than the historical median of the z-score of the next lowest rating category well we do a series of validation tests to see that whether we are capturing really downgrade vulnerable firms and while vulnerable firms look worse along observable characteristics like sales growth investment ic ratio profitability leverage and so on compared to non-downgrade vulnerable firms and also they are more likely to be downgraded and have a negative credit watch and outlook so let me show you this funding privilege in the data let me start with raw data the figure on the left is showing during the entire period post global financial crisis for each rating category the spread between vulnerable firms and non-vulnerable firms well this spread is positive for all rating categories except triple b firms it's positive meaning that if you are a vulnerable firms say rated double b you have to pay more to finance yourself in the corporate bond market compared to another double b firm that is non-downgrade vulnerable well this relationship is flipped for the triple b rated firms and on the right hand side you see a similar figure but now expanding in the time series dimension and you see that this spread for the triple b firms has been hovering around zero during the sample period plus post global financial crisis and actually negative from 2013 to 2016. we can do a more parametric approach or robust and try to explain the spread of a bond b issued by a fermi in year t and this is the spread over the treasuries with a similar maturity and try to explain it with various bond level characteristics industry or fixed effect and crucially a dummy that captures the rating bucket of the firm and a dummy that captures whether the firm is vulnerable to a downgrade let me unpack the estimation results well first of all somewhat reassuring we see that secondary market spreads increase as the ratings of the firm deteriorates we know that rating capture meaningfully credit risk and so this is somewhat reassuring however within each rating category we see that downgrade vulnerable firms pay more to finance themselves in the corporate bond market compared to non-downgrade vulnerable firms and this is true for a for each rating bucket except the triple b rating bucket where again we observe this funding privilege if we look in the sub-sample in sub-sample periods this funding privilege is particularly large from 2013 to 2016. what's interesting is that this funding privilege seems to be unique to the corporate bond market for example if we have as a on the left-hand side edfs of two years and five years we don't observe this higher default risk we lower the fault risk of triple b rated firms based on these measures we also observe no privilege in the corporate bond market pre-global financial crisis that's the first column or in this in this indicated loan market that's the second column so what's the story of this privilege well the story is the role of qe in driving investors demand exactly for these bonds that are rated just above the triple b rating cutoff how do we measure this in the data well we have security level holdings by investors and we match the security level holdings with security level holdings by the fed in particular for each investor k at time t we can measure the exposure of these investors to qe measured as the share of investor case holdings that at that particular time t are held by the federal reserve well we can use this meaningfully in a regression specification in particular we can use issuer time fixed effect so we are taking the bonds issued by the same issuer and comparing these holdings by two or more investors that have a different exposure to qe that's what this regression specification is doing and what we find is that investors that have more that are more exposed to qe demand more bonds by downgrade vulnerable firms and in particular as you can see in the second to last and third to last column this is particularly driven by issuers that are rated triple b and by those investors that tend to focus on the i in the ig corporate bond market we can think of this as investment grade mutual funds and insurance companies in our data so then the natural question is what do what do these firms what do these firms do with this cheap bond financing and we see m a as an equilibrium response to investor demand in particular we see that prospective fallen ages supply bonds and these bonds are largely to farm dna this is something that we can see in the data what's so special of about m a on the one hand we know that there is sluggishness of credit ratings uh especially when it comes to downgraded firm downgrading firms from the investment grade to the sub investment grade area and in particular we document that this luggageness is particularly large post-m a activity now namely there is a uh the tax credit rating agencies take time to downgrade firms especially post m a in other words m a is a very cheap technology for firms to increase their market share while delaying the downgrades in the interest of time let me show you the increase in market share uh in the left-hand side you see market shares measured by sales by first by ratings and you see this huge increase in market share in triple b firms well on the right we divide each rating bucket in its vulnerable and non-downgrade vulnerable component and you see that prospective fallen ages namely those triple b rated firms in red that lived just above the investment grade cut of driving the entire increase in market share pro post global financial crisis and i mentioned before that m a exacerbates the sluggishness of credit rating agencies here you have transition mattresses that look at how first transition from each rating bucket to another rating bucket in the left to see the full sample and for example you see that a triple b firm has a a three percent probability to be downgraded to double b firms in the full sample however post mna activity and i'm moving here on the transition matrix to the right this probability basically falls to around zero we have several tests in the paper that document this luggageness post m a and interestingly i showed you before the wave of fallen angels post global during covet this entire wave of fallen angels has been driven by those downgrade vulnerable triple b firms that engaged in m a activity during the period post global financial crisis that's a figure on the left we don't see barely any fallen angels among those triple b fairs that didn't engage in m a activity plus global financial crisis why does this matter in the last 30 seconds or so this matters before the macroeconomy because non downgrade vulnerable first in an industry with a larger share of these prospective fall ranges we document have lower investment levels lower sales growth rates and lower markups compared to non-downgrade vulnerable firms in an industry with a lower share of prospective fallen ages namely this beloved effect suggests that this funding privilege has important macroeconomic effect in other words we acknowledge that the growth of the triple b market may have been a desired effect of quantitative easing but there are costs the these firms that are subsidized by qe grow disproportionately large and become more fragile consistent with the exposed evidence during the corporate crisis and the resulting spillover effects negatively affect their competitors potentially affecting the macroeconomy as well thank you thank you uh thank you matteo very interesting and uh you know one must wonder whether there are there is anything that actually prudential regulation can do to uh uh to address the uh the uh the behavior and the incentives of investors and and credit rating agencies in in this space um but let me move over to uh to jane for her uh discussion thank you so much well thank you very much for inviting me to discuss this very interesting paper um i really enjoyed reading it so uh let me jump directly into it um yeah so the paper is motivated by the huge issuance of triple b rated bonds since 2009. um the triple b bonds now accounts for more than 50 percent of all investment grade bonds this is true both in the u.s as well as globally the bonds are also issued as very low offering spreads suggesting that this is driven by a shift in the credit supply side um furthermore we see during the clovis shock a lot of these bonds are downgraded to high yields and the paper explores what are the driving forces behind this phenomenon unique to triple b rated bonds and what are the implications on capital misallocation and risk buildups so in particular the paper documents a large subsidy for a risky triple b rated bonds that have high chance of being downgraded to high yields this is mainly driven by those investors who are exposed to fat qe program and they switch to other risky assets in search for higher yields what i really like about this paper is that it then takes the next step and investigates the real impact it shows that the special demand for risky triple b bonds decrease the capital for those firms increase their leverage and encourage risk taking activities such as m a as a result exposed we see more of these downgrades being realized during the negative shock and furthermore as these firms expand other firms in the same industry are negatively impacted so they have lower investments and employment growth this paper is of course very important for understanding the unintended consequences of unconventional monetary policy which in turn would inform us about future monetary policy and natural potential policy design so let me put this paper in a broader context what this paper really makes me think hard about is how does qe differ from conventional monetary policy in terms of the transmission mechanism while conventional monetary policy mostly works through banks balance sheets which now are heavily regulated and closely monitored complicated easing takes effect through a various market-based financing channels these market-based financing channels are dominated by a heterogeneous set of non-bank financial intermediaries that have different objectives and regulatory constraints and furthermore as these markets are far from being perfect the specific preferences and constraints from these investors matter for prices and firms funding cost what complicates the issue even more is that the investor base is different for different segments for example in the investment grade segment there are 20 mutual funds and 60 life insurance companies in terms of folding in the high yield segments is almost the reverse so while mutual funds are concerned about meeting redemption needs the insurance companies are concerned about risk-weighted capital and because of these different features they're going to affect firms in different segments differently so this paper shows the friction in credit rating plus regulatory constraints based on rating generates a subsidy for these risky uh triple b terms uh this logic applies actually more broadly another example could be that the lengthening of corporate bond maturities might be potentially driven by life insurance companies preference for duration matching in a lower interest rate environment a consequence of this is that as the interest rate increases bond value now are more sensitive they could potentially drop more significantly than when the maturity was shorter and this could pose risk to other market participants and finally another way of seeing the contribution of the paper is that it basically showed us a very problematic example of non-market mutual monetary policy in principle we want the monetary policy to be neutral in the sense that it affects all firms funding costs equally but here given the particular market structure it is subsidizing a particular type of firms distorting a capital allocation which is presumably not the goal of any monetary policy program what i wish the authors can dig deeper into is the origin of the special demand for the risky triple b bonds um it does have a disc uh discussion but i think there's more can be done here um so usually this special demand is driven by what is called reaching for yields meaning that investors chasing for return ignoring or not accounting for the risks properly but in this case the yield for these riskier bonds are actually lower so it can't be just chasing for for return or for yields what i did notice is that the vulnerable bonds that are receiving these subsidies seem to be slightly larger in terms of issuance size compared with non-vulnerable bonds typically larger size bonds have better liquidity in the secondary markets so it might be some other features such as liquidity size or maturity that is driving the special demand for response and i thought it'd be very interesting um to get to the bottom of this to understand what characteristic of the bonds is receiving this special demand finally as the covet chalk hits we see another new round of liquidity being injected into the market it's almost like history repeating itself in another form and this time we're seeing a record number of first-time junk bond issuers in the market in the summer of 2021. i think the overall lesson here is uh to really track where the liquidity being provided by the central bank is being directed to given the investor base and particularly what are the frictions or constraints faced by these investors in the market it's much more challenging than regulating banks whose business model is somewhat homogeneous the set of players in the market can be can be quite heterogeneous but only after understanding the uh players in the market can we discuss uh what kind of monetary policy is market neutral and does not induce misallocation it also means that our macro prudential policy should be more market-based rather than institution-based um so with that let me end here i really recommend the paper to everyone who's interested in understanding qe and the bond market thank you very much and i look forward to the discussion thank you very much uh jane we have we have five minutes left so let me turn back to uh to the speaker and and give give them two minutes each to uh uh to uh with some sort of key reflection on on the points raised by the discussion discussions as well as uh you know there's been a lively exchange of questions and reflections in in the in the chat uh falco why don't you start do do so well first of all i wanted to say thank andrew once again for for this uh very nice uh discussion i think there were a lot of helpful points that we should further probably also try to even elaborate on within this paper um i thought um i mean in particular on his point on whether there are costs of imposing these fees um i think at least in that regards also the results um that we have on the net inflows seem to be quite interesting as at least in crisis periods uh apparently investors are not that worried about these fees that might be imposed on them also in the future so it might still attract also a lot of inflows i would suspect even from institutional contrarian traders that are in the market so i don't think that it is a big issue for most of these funds so but that's just a guess or base what we have uh certainly it's worthwhile to to to dig into that somewhat more um yeah i think um i mean certainly we cannot speak uh about whether these are only add-on effects or whether price-based liquidity management tools would also serve standalone as a very useful instrument but i think given that the quality based tools were not really used mostly i mean they stay and stand ready in the background to be used as an option but mostly even during the covet crisis they have not been applied at least not in the in the irish case so i think um they're probably not that important in order to contain uh liquidity outflows but again this is speculation this is not something that we can really uh really show with the data that we have but this is indeed also i think impossible with data and is available um i think there was one other comment also from the from the chat which i'd like to pick up on and that is um i mean we have um indeed an issue with the fact that the um those funds are particularly fragile that were most sensitive to uh past performance in their flows they might indeed uh be also those that do not have the price-based liquidity management tools available so there might be also a selection between the uh fragile and less fragile i know the high sensitive low sensitive funds but i think what that creates is rather a bias that works against us rather than in favor of us so i'm relatively confident that our results uh are not affected by this um but certainly um there are this would be something to to investigate somewhat further um one other concern that has been also raised is this selection effect um that potentially the funds apply based on some characteristics observed or unobserved characteristics whether they choose to to use a liquid a price based liquidity management tool but actually as far as our data shows the decision is mostly made at the asset management company level so it is not something that the individual fund decides about but it is rather something that the asset management company decides about and most of those companies are really huge so that really does not boil down to the characteristics of the individual fund i think we have no strong correlation here we also tested for this uh that does not seem to play a big role thank you i thank you falco 30 seconds each for uh for to bsm and mateo sorry about that okay thank you um quickly yeah thank you so much carolyn for this um insightful discussion um i agree with you i would love to um the study dynamic effects of repo markets but i think we first have to understand the static effects before we can go there and already with this very as you pointed out with this very simple static model we obtained very highly non-linear um welfare effects just looking at the basic market structures i think the other point i would like to comment on is that you you mentioned that there that you felt okay there's no asymmetric information in otc markets in our model where there is in ccp markets i think what we really need for our results to go through is that there is an informational wedge between these two markets and i think we will all agree agree that anonymity increases asymmetric information further than uh than in the otc market um i i was kind of happily surprised that you thought that we were favoring the otc market we were coming from the empirical paper from l'oreal in the rfs 2017 where he actually showed that the ccp market was fairly liquid and the volume continued working fine during the great financial crisis and so we were always trying to find negative things about the ccp and actually the pooling equilibrium um i mean it has hits its downsides because it's putting these bad borers with the good borrowers but it also ha i mean that has also its upsides right to the insurance effect so i think um yeah it's not so clear whether we are only picking the the the negative equilibrium here for the for the ccp and in fact we can show that um from an example point of view boris would actually prefer the pooling equilibrium but i'll stop here i think i'm more over time thank you so much for this insightful discussion and uh yeah thank you matteo if that's okay with you because we are really out of time i'll suggest that you get back uh bilaterally to uh to jane and and others uh thank you sorry about that and and let me let me conclude by by thanking very much uh you know the the presenters and the finalists this was a very stimulating discussion i think this is obviously going to continue and uh and perhaps i will just close by suggesting that um you know the excellent question that manfred put on on the chat about the time of time horizons in the digital age and what they mean for regulation is the subject of the next fsb conference so with that thank you again i close section 4 and hand it over to d3 many thanks marina and many thanks for this uh proposal for the next conference um now for the for the moment we are approaching the end of this one and i would like to thank everyone for your input uh special thanks of course to the chairs presenters and discussions for the time they have dedicated to make this an interesting and stimulating event i think we've learned a lot about how a systemic risk perspective can be developed for the nbfis sector i will ask martin maloney in a minute for his reflections on what has been discussed but before doing that let me briefly tell you about the way forward in the fsb the goal of the current stage of our work is to develop a set of proposals to be included in the november report to the g20 on policies to address systemic risk in nbfi to get to that point the fsb and namely our steering group on nbfi which brings together senior market regulators and central bankers will carefully consider a range of inputs including the contents and discussions that made at today's conference the findings of the various work streams underway in the fsb experiences made by our members with different policy tools so far and so on and so forth in any case the november report would not be the end of the road its proposals would form the basis for follow-up more detailed work by the fsb and standard setters as appropriate including potential adjustments too or additional guidance on existing standards or more guidance on existing standards as as needed um let me now ask martin maloney uh iosco secretary general to share with us his main takeaways and i will then conclude with some thoughts from my side martin the floor is yours dietrich many thanks and uh first i have to share with you my thanks to all the presenters and all the respondents for what i thought was quite an extraordinarily rich two days and anyone who has uh made it through both sections of this this conference will have a huge amount of data and information and analysis and discussion topics to take away and and and think about and we all know that turning academic analysis into policy conclusions is itself a fraught process so we will all have to deliberate very carefully on what we have heard on the journey towards a more systemic uh uh approach and i was very glad to hear you dietrich emphasize uh that while uh our report to the g20 your report the fsb report the g20 in october is important the work will continue uh uh thereafter and in many ways uh we are hostages to our own success because the work that the fsb has done with that school over the last 10 years has has made huge progress but as you make huge progress it gets harder and harder to fill in the gaps and do the last bits or the remaining bits correctly and you have to really calibrate your policy proposals really carefully i was very struck in the last session and that's not to take away from the uh from the other papers by a falco's paper uh it it would be very nice wouldn't it if we could uh ourselves with such detailed precise papers that create very clear measurement of of of individual policy tools but of course even as you listen to falco and his respondent you begin to ask questions around well you know how long would that affect last as stress got got more intense is it a linear effect does how precisely does it depend on how well the price uh effect is is calculated and so on and all sorts of questions really interesting questions arise um i i thought session three again raised a whole broad set of questions around around data and data analysis and they really illustrate how challenging the data issues are for us as we move to a sort more systemic approach i guess on the one hand there's what data we collect from industry and what data they need to provide us with and on the other hand that's what we focus on within that data and how we get effective good analysis a good red flags in order not to create lots and lots of false positives for ourselves making it impractical to devote regulatory resources to this and also keeping in mind of course uh uh the um the cost of data collection so we we probably actually need more uh academic input i suspect and reflect reflection among regulators around developing a a more systemic monitoring approach and we need to make sure this is really quite difficult that each step along the way we're adding value because it's quite clear that this is a hugely challenging issue for us and i've no doubt our debates will in future focus a lot on that session two it was very interesting that uh money market funds were so critical in relation to session two and uh the interconnectedness of markets and the interconnectors of jurisdictions and the interconnectedness of regulatory frameworks that were illustrated in in uh session two really show again and i'm sorry to go on about risks but it really is true uh the risks of unintended consequences that arise from how interconnected our markets uh uh can be who would have thought frankly that the behavior of insurance companies would have been affected in the way that we saw uh laid out for us uh in that in that section i thought it was extremely interesting and i come finally back to uh uh session one where the importance of self self insurance was a strongly emphasized and i think it really is important and we've done a huge amount of work on that and i would just add the thought there's an analogy here with driving cars that maybe is worth mentioning we require car drivers to self-insure but we also spend a good deal of money on ensuring good quality roads rules of the roles and proper policing of the roads in order to maybe make that self insurance at a reasonable price and perhaps uh one of the things that came out for me in that first session was the importance of looking at supply as well as demand and i think that was also illustrated actually by some of the comments and papers in in in part four as well we need to look at the supply and bond markets uh in order in a way to ensure that our policy is balanced and we've done everything we can to promote maximal supply in in in bond markets from the private sector and there i think the the the contributions are really interesting what we still don't have and what we will have to reflect on i think is what is the path forward so we had mention of the israeli experience which is very interesting but how do you get in global bond markets from where we are towards a central limit order book approach and towards overall trading if that is going to be helpful or perhaps towards central clearing if that is going to be helpful keeping in mind the comments and repo markets that we heard in in in part four uh are repo markets part of the solution there this is some really interesting questions here's some very good data analysis and i must say it was uh very interesting to listen to it dietrich i'll hand it back to you having laid out for us uh possibly an impossibly long list of follow-on questions and reflections that we have to make as regulators and policymakers but it was extremely useful thank you very much martin i agree with your points not least with with what you said about the critical importance of of data um let me perhaps add a few thoughts from my side on how to use martin's words to turn how to turn academic work that we've been discussing uh the past two days into policy the discussion particularly in session one but also subsequently underlined that the framework developed in our nbfi progress report last year last october provides a fairly solid basis for thinking about systemic risk in a quite heterogeneous nbfis sector namely that market-based intermediation can result in liquidity imbalances that need to be managed and when i say to need to be managed it is important to distinguish the normal fluctuations in liquidity demand that are part of price formation and portfolio management in capital markets from those large shifts that crystallize financial vulnerabilities in nbfi and that run the risk of jeopardizing the supply of finance to the real economy and it's the latter that are of concern to financial stability authorities many interesting points were made on different approaches that can be taken different roads that can be taken to manage such imbalances reducing the risk of spikes in liquidity demand has been one focus including in the session that we just had i think this is for a reason and important parts of the fsb's work on addressing specific vulnerabilities are about that what the discussion also brought up brought out clearly martin also mentioned it was the role of interconnectedness that may give rise to spikes in liquidity demand not least through margin calls steps to better understand and manage liquidity risks from interconnectedness including liquidity preparedness to meet margin calls seems important for containing systemic risk in nbs and bfi and in a similar way it will be important to think of leverage as a potential source of abrupt pro-cyclical changes in liquidity demands more systematically there was also a fascinating discussion about the determinants of liquidity supply including the impact of changes in market structure from otc markets to all to all platforms and of different clearing arrangements on liquidity provision under stress yesterday somebody said something to the effect of well market structures are what they are and while this may be true i would infer from that that policy makers should not take a careful look at market structures markets that remain liquid under all conditions may remain elusive nobody will catch the falling knife issue but promoting market structures that support resilient liquidity provision liquidity supply would still enhance the capacity of markets to absorb shocks that otherwise might require interventions so this brings me to the dog that didn't bark or at least didn't bark loudly whether and under what conditions central banks should act as lender or market maker of last resort in nbfi i think the pros and cons are well understood and we've heard some of them reference to some of them during the conference yes they should because of the relevance that nbfi has gained for the financing of the real economy yes they should because markets may not supply sufficient liquidity in times of stress again nobody will catch a falling knife argument but no they shouldn't because of concerns that central bank intervention gives rise to moral hazard and may lay the ground for for future problems now these are important arguments that require careful consideration and discussion and i think this includes the question well coming back to your car driving metaphor the question about what stringent preventative measures to address systemic risk in nbfi would imply for for the lender of last resort function um so with these big questions um let me stop and again thank you once again all for your participation in the conference goodbye well i would have said no yesterday today maybe a little bit more wise it was some interesting today yeah but then again you know how many of the audience read our stuff and already they are aware of all the papers so also yesterday i think it was interesting people that they know something but not everything yeah yeah i think it's a whole problem it's not meant to say that you shouldn't try to work on it but i think it's true the academics they just don't have an incentive to consider this kind of stuff it's about moments looks like you're still yeah but i mean maybe you should publish papers by saying if you read this you will get access to unique dataset i mean i know when i published the longest uh history at the time which was the original compendium of standards it was 210 pages back in 2012 when we did an update for the first time in 10 years it had a standard the number of standards just three people quite little i was curious to put the footnote on hey