#CEBRA2021 - Session 2: Banking, Imperfect Competition, and the Macroeconomy

#CEBRA2021 - Session 2: Banking, Imperfect Competition, and the Macroeconomy

hey rustom hi hi everybody hello nice to meet you so we're on limiting will [Music] okay i guess we're all here so um we'll just give it a couple minutes and start on time and uh yeah and we'll be following the order that's on the agenda that's okay with everyone and it's 20 minutes with 10 for questions or 25 and 5 or yeah we can do exactly so there's like 30 minutes for each paper so i'll kind of like let whoever let the presenters perfect allocate perfect uh shift the distribution yeah let's let's target i guess 25.5 but you get a little buffer also should we save comments until the end or should can we interrupt or might reference is interrupting uh yeah so you have 30 minutes uh but also because because we have limited time if someone is being extremely annoying feel free tell them shut up because it's your 30 minutes so the usual stuff basically exactly yeah i yeah i i like to keep it like relatively informal but there's a limited time so a fine balance we have to okay i guess uh i guess we can start because it's 11.

Um and uh we're very happy to have uh four very interesting papers here in this session that's sponsored by the federal reserve bank of saint louis banking the perfect competition than the mecca economy so i'll just follow the order in the agenda so we'll start with kyle dempsey from ohio state then we'll go to will diamond from wharton rustam jamilov from lbs and then pierre mabi from yeah so each paper will have we'll allocate 30 minutes for each paper ideally 25 plus 5 but uh yeah feel free to manage that however you like it so i'm going to ask kyle to uh start sharing his screen and i'll be keeping track of time okay great and everyone can see and all that yep feel free to go ahead okay great uh well thanks for having me um so this is joint work with our distinguished host miguel faria castro um so this is called the quantitative theory of relationship blending and so the starting point for this paper is really recognizing that um you know relationships are an important part of credit markets but they their macroeconomic consequences have historically been understudied and so we're trying to focus in this paper on a particular facet of the role of relationships in macro finance which is the propagation of shocks in the financial sector okay so by way of motivation and there's a lot more to unpack here potentially but this the standard financial accelerator type mechanism so think bernanke girdler gilchrist or kiyotaki more implies counter factual responses typically in settings with financial intermediaries right the idea here is that banks can recapitalize and therefore the fundamental shock dissipates too fast right this is true importantly for what we're going to set up here this is true in settings that have perfect competition and this is also true in settings where you have a lot of monopoly power right so we're going to set up a particular notion of competition in this paper okay so as a way of sort of dampening that sort of fast recapitalization mechanism this paper is going to look at a mechanism by which we close this through relationship lending okay in this setting what's going to be key is that banks are going to have excess profits right they have some sort of at least temporarily captive customer base right but they can't cash in on those excess profits all at once right so it's going to give rise to a very particular nature of competition okay the building blocks that give us this set up here is going to be an environment where we have heterogeneous banks subject to financial frictions and this imperfect competiti imperfectly competitive setup right well we're going to model using what we'll call customer capital sort of following um gory over rodengo and some other literature there we can also think of them as habits borrowers have habits from borrowing at each banks and there's going to be adjustment costs right i'm going to it's going to be costly for me to sort of shift my portfolio of lending decisions in the aggregate away from certain banks okay so as banks are hit with shocks that's going to add a layer of adjustment that's going to govern dynamics in this in this economy okay quantitatively what we're going to find in this paper is that this mechanism really reshapes sort of the aggregate recovery from a financial shock and in particular it's going to dampen this sort of top-line motivating feature of many models within this literature that we're um that we're sort of interested in in particular banks who have access to these sort of excess profits that they can use in principle to recapitalize will optimally choose not to do so all at once right that's going to mute the impact of a financial shock on the real economy on impact but it is going to make the recovery more sluggish it's going to add some persistence to sort of a transitory fundamental shock inside of the banking sector okay so that's going to be the main punch line okay um so let's just go over sort of more concretely what we do and what we find so i'm going to organize this talk into basically two parts one's going to focus on our model and the second is going to focus on our quantitative results and some empirical results um so in our model relationships uh affect equilibrium in two primary ways right first is going to be on the demand side right rather than just there's a sim a single aggregate price that the borrowers in this world take is given there's going to be a distribution of prices and sort of banking traits that are relevant to the borrower okay what's going to be really nice in this framework is that things are going to boil down pretty neatly into three primary moments of this distribution right so on the demand side of course the high level of interest rate spreads is going to be costly right that's going to make borrowing more expensive there's also going to be some variance in equilibrium in terms of interest rate spread some dispersion that's good for borrowers all else equal because it allows them to substitute right as long as there's some variance i can always go down the cost distribution for my financing but in the presence of these habits or this customer capital relationship capital i'll use all of these terms interchangeably there's going to be correlation between pricing and these habits right and if those if high spreads tend to be concentrated at banks which i have a low relationship intensity that can be problematic okay um because then i don't i can't substitute as readily into those banks and so we're going to capture this through those moments then on the supply side banks are going to internalize how these habits and how this customer capital is formed over time right so that's you can think of that sort of in two ways on the one hand you know a bank that's starting out has to build up this customer base that's going to have certain implications for how they choose to price loans the other one that's sort of more germane to our our um initial motivation here is that when faced with a shock to net worth right it's going to change the nature of how banks want to respond to that shock okay like i said quantitatively what we find in in terms of simulating a negative aggregate shock to bank net worth so sort of a pure financial shock right that in principle has nothing directly to do with uh low ending sector right we're going to show that in the aggregate this is really going to slow the recovery of net worth it's going to lead to sort of a more persistent drop in lending than a purely competitive case okay lastly what we're going to do since our model has these implications for sort of aggregate recoveries we're going to try and use data to validate our mechanism at the micro level right so what we're going to do along this front is construct empirical measures of customer capital and we're going to look what happened to a cross section of banks around the global financial crisis right so we're going to sort of have this event study type type design there where we're going to document that banks who experienced large drops in market value around the financial crisis so from before 2008 to right after 2008 they experienced also large drops in our measure of this customer capital object okay that's going to be the first point the second point is that banks who did experience these large initial drops in customer capital subsequently experienced faster recoveries in market value same same goes true if we if we hold this for net worth as well right so this speaks to customer capital this relationship object being sort of an expendable asset in times of of crisis that banks at the individual level are going to optimize and decide how to expend okay and that's going to be true both in our model and in the data just briefly in terms of time let me just go over the literature briefly we view ourselves as combining insights from three main branches first is sort of your standard financial frictions financial accelerator type literature like i mentioned before um second is going to be this customer capital or deep habits literature um goryo and rodanko ravin uh uribe mcgrowhay these are all um important sort of building blocks for our framework and then lastly on the bank side we're going to have a lot and a lot to borrow from sort of classic incomplete markets heterogeneous agent style setups okay putting all these things together the way we do we're able to deliver a tractable quantitative framework where we can analyze uh how lending relationships shape these aggregate credit outcomes okay there's an empirical literature on this um you know going back to a lot of empirical finance papers in the um in the mid 90s some more recent papers there's sort of less equilibrium macro work on this and so that's where you know an important paper there is yasser mulan's work but we're trying to sort of flesh that out a little bit more okay so let's dive in so i'm going to start you know the middle of this talk here i'll devote to the model and then i'll get into our quantitative results so i'm going to start with the borrowers then move to the banks then move to our equilibrium concept so we're going to try and collapse everything about the borrowers in this model into one slide here and the key thing to recognize in this model is that there's going to be a two-part loan demand system in equilibrium right we can think of there as being a representative borrower they operate some decreasing returns to scale technology subject to a working capital constraint but critically they're able to source their financing from multiple banks but they're subject to adjustment costs so this little stylized you know pseudo equation here we can think of as defining the nature of relationships in this economy right there's one product here it's it's a loan right so loans all else equal are the same right but there's multiple sources right i can get these from different banks i have a different sort of degree of relationships or a relationship intensity or a habit at each of these banks and then it's costly conditional on those habits for me to adjust my portfolio in response to changes in pricing okay so a world i want you to think about here is there's a set of banks who are being hit with idiosyncratic shocks responding to that by changing their loan prices and there's a borrower who is taking into account the loan prices i see and my relationship with these banks when making my individual and aggregate lending decisions okay um so let's start at the bank level basically this bank level demand system is going to specify the bank the bank's share of aggregate lending or aggregate borrowing right as a function of the habits and the prices they charge right so i've tried to sort of match uh you know by color here the different components of this demand equation to our relationship construct here right so first key object here is going to be that the state variable that's relevant to the borrower in this world is going to be this joint distribution of prices and habits okay that's this mu of q and s objects okay um q is going to be the loan price s is going to be the level of customer capital or the habit at that bank okay the next key object comes from this costly adjustment component and here's this parameter phi the way we model this is that there's a quadratic cost of adjustment in terms of loan shares relative to customer capital shares right so it's costly for me to get a greater share of my total lending from a bank at whom i have a relatively brief or relatively low relationship intensity okay so what's important here is that a high level of customer capital is going to shift up the level of loan demand but then there is a standard mechanism by which uh if i charge a higher interest rate spread so one over q is going to be that bank's interest rate capital r is going to be the equilibrium average interest rate if i charge a high spread over that i'm eating into my loan demand in the aggregate what's going to come out of this setting and we can sort of you know we don't need to go into too much detail on the loan demand equation this l prime equation what's important is that we replace the standard object which is the average interest rate or you know a common interest rate charge by all banks in this world with equilibrium price dispersion we get this effective interest rate which depends on three things the first is the average interest rate so this is the typical object let's just some overall banks what's the average cost of credit this is sort of a level effect the second term here reflects who's charging what rates right i care about the covariance between spreads and customer capital if this term is high that drives up my cost of borrowing in the aggregate okay because it's costly for me to then borrow from who i want to borrow from the last term here is this uh pure dispersion term right in the presence of a given level and covariance of interest rates all else equal interest rate dispersion is good for me because i can borrow from someone cheaper okay so these three components now are going to combine to form this new effective interest rate which is really what's relevant for aggregate demand okay so another question here so quick question here so usually with relationships there's large numbers of state variables this is tractable because this covariance and variance summarizes everything you need to know about the distribution exactly it's a pure sufficient statistic basically these um yeah l s and r and r tilde which we see on the all in different points of the slide are everything that you need to summarize about this yeah um [Music] okay so let me dive now into the bank's side of this okay so i'm going to try and go quickly just lay out all the the facets of this problem and try and highlight the two key pieces of this analysis which are the financial friction sort of the standard pieces and our new uh modeling of this competitive structure okay so the bank problem in some sense is going to be fairly simple there's just two states i have a net worth and i have a level of customer capital right i need to choose my loan price my loan quantity my external financing think of that as deposits right and i need to choose how much dividends to issue or dividends to pay out or equity to issue right um and i'm going to be subject to a sequence of constraints right the first two constraints in black here are going to be standard budgetary and net worth dynamic equations right i have to finance my loans out of my net worth deposits in terms of dynamics i'm going to have some return on loans which is going to face some idiosyncratic shocks and i'm going to have to pay back my deposits okay i'm going to be subject to a leverage constraint or a capital requirement here that's just going to put a sort of ceiling on my external financing as a fund as a function of my total lending and then the really sort of novel parts of this framework are going to be these bottom two slides through the bottom two equations the second to last one is the bank specific demand function that comes from the uh bank specific demand uh curve on the last slide okay banks internalize that when they price that has an implication for how how much how many loans they'll actually be financing the second and this is key for our analysis is that banks internalize their the habit formation of borrowers right so they know that if i lend a lot today right then they're going to have a stronger relationship with me going forward if i lend not a lot today they're going to have a weaker relationship with me going forward that has implications for both the level of demand at my bank and the elasticity of demand for me okay so this is the key uh part of our setup here what does that deliver and in particular what do these last two pieces of our model deliver they deliver this sort of lifetime euler equation for for loan pricing so there's a lot more equations than i would typically like to do on a short talk here but let me try and unpack this one the left hand side is lifetime net profits of the bank the right hand side is going to be what i'll call excess return okay so let's start on the left side the lifetime net profits are going to combine two pieces here there's a static component that's captured by this pi t object that has three pieces there's the return on loans there's a shadow value associated with easing my financing constraint or my capital requirement and then there's a funding cost right every loan i make today is a dividend that i don't issue today okay so that's going to be true within any given period then dynamically there's this additional term right because my pricing decision today affects the demand curve that i face tomorrow right so future periods are going to be linked by these customer capital dynamics right and so what i want to highlight here is that in a world where there's no customer capital right or even in a world where there's permanent customer capital right so that rho would be equal to one that persistence this term goes away right there's no dynamic linkages right so we can nest those cases pretty easily in our setup okay now let's go on the right hand side we can think of this the sort of right most turn on the right hand side as the excess return sort of per or the return per unit of loan right but that's going to be scaled up in this type of setting by the inverse elasticity of loan demand okay again in the setting where we have no customer capital this is going to boil down to a perfectly competitive model this excess return is going to go to zero right but in a world with customer capital that gives each bank some degree of market power that's going to give banks some excess return that they can tap into in the face of shocks through time okay and that's the key insight of this of this framework okay we're going to define equilibrium in two steps just to say this quickly you know going back to sort of a wills question from before um you know we do boil down this sort of in general large state variable which is the joint distribution of prices and habits into a couple of key moments but that affects how we have to solve this first we have to sort of let banks take key to demand parameters is given and solve for a banking industry equilibrium that's going to be sort of our incomplete markets block of this model once we have that we then have to ensure that this loan market is actually in equilibrium right because each distribution of banks right implied in this banking industry equilibrium is going to imply different covariance terms variance terms level terms for interest rates so there's going to be sort of an outer loop that that involves making these things all consistent okay okay so that's all i wanted to present in terms of the model just a high level overview i'm going to take my last bit of time here and dive into the quantitative stuff okay so first i'm going to look at results in the cross section thinking about how the role of relationships shapes the banking industry equilibrium here then we're going to look at dynamics think about how this affects the response to an aggregate shock and then we're going to try and see if we can find evidence and support this mechanism in the micro data so here i'm just comparing two versions of our model in the steady state and i'm just looking at some cross-sectional moments okay so in the left column we have our baseline model that's the one i've laid out the the right column we have a competitive case again think of that as the limit as that adjustment cost parameter goes to zero in that world there's really no notion of customer capital the bank problem simplifies and there's no sort of market power at the individual bank level okay now we've set these two economies up so that they deliver the same aggregate loan volume right a couple things i want to point out what does that mean in terms of interest rates well the average interest rate sort of summing across banks in our baseline model has to be lower in order to get the same level of loan volume right because it's not just that level that matters anymore right in general there's sort of a positive covariance term where hot banks with higher um customer capital are charging higher spreads that makes borrowing costly on average right so we see this discrepancy in these pricing moments right there's pricing dispersion there's more dispersion in loan volume in our baseline model right what i want to stress here is that you also see different dynamics around net worth here right and in particular the banks in this model in this baseline model since it's sort of costly to build up customer capital right and since banks face these financial frictions there is some incentive here to maintain an additional buffer of net worth away from these constraints in order to cash in on this customer capital that you've built up over time and so you see actually about a 10 percent higher level of bank net worth on average here right so it's sort of a a buffer stock type motive okay the other thing i want to highlight here the last thing on the slide is that you see a pretty strong correlation between customer capital and net worth that further supports this mechanism right banks are sort of growing by lending over time and so as they build up this net worth they become more profitable they have more customer capital okay uh i have a quick question so do you reverse engineer the baseline such that the total loan volume is exactly equal to competitive or is it just by accident we calibrate it by adjusting these um these relationship parameters in order to deliver that same like there's a the persistence the adjustment cost parameter that's going to deliver this the you know we we can make it such that they match because as soon as you depart from perfect competition you must have the case that quantities are lower than in the perfect competition counterfactual so you do get that uh that your model does deliver drop in quantities yeah so you you would get a drop in in relative to the purely competitive case just because the level of spreads is higher yeah but there is this sort of um other force in this model that brings down the level of spread somewhat but again there's this subtlety where you care about not only the level but these also higher order moments but that's an over answer your your intuition is correct yeah that was a bit um that's the first thing that comes to mind when you talk about perfect versus competition and second so your correlation between customer capital net worth is it the cro in the cross-section or is it for the time series for because this is a cross-sectional moment but then that can i interpret it by saying that relationship based lending uh intensity increases with the size of the bank which would be the opposite in the data or how how is it so i i think it's a different measurement of intensity that you have in mind there um i think in the interest of time i might hold off on that question but we should talk more about that one that one after but i i i think there's a subtle difference there that that we should get into um okay let me move on to share our dynamic results i'm running out of time a little bit here okay so here's sort of our our big picture quantitative results so what we're doing here is we're simulating an aggregate shock to bank net worth okay blue line is going to be our baseline model think the analogy of the left column on the last slide orange line is our competitive case right column on the last slide all we're doing is we're taking our steady state distribution of banks knocking off 10 of their net worth okay so what we see across these two economies first and foremost in our top left panel is that indeed we sort of do get proof of concept on this first motivating fact here the net worth recovery is sort of you see a more persistent drop in net worth coming out of um our baseline model than the competitive case okay why does that happen right well there is sort of these the competitive forces in some sense cut against these banks on the impact of the shock and what you see is that the average the level of interest rates tends to go up less because banks don't want to perfectly sort of cash in on or erode this customer capital just because they're in distress right now so they raise interest rates less cut lending less on impact but then of course that saddles them with sort of less net worth going into the future and so you see um total lending sort of stays below for a couple of periods um relative to the competitive case and interest rates stay elevated for a couple of periods relative to the competitive case as well okay last couple of minutes here let me just wrap up with sort of how we're thinking about validating this mechanism in the data so the first uh thing to do is of course to find a measure of customer capital on the data right we can go into much more detail here fundamental issue we have this object s that is clearly defined in our model we do not have that object in the data as clearly so we're looking at a couple of ways of handling this group one we'll call proxy measures i'm going to show results on that today we can think of sort of a tobin's q on our loan portfolio that's going to be related to this fair value of loans from atkinson at all we can think of a learner index or a banking markup type measure related to the jimenez at all work um we could also think about doing something more detailed and diving into sort of loan level rather than bank level data and trying to estimate demand relationships we're not you know we're sort of working on that still we're going to just show the proxies today right but what we're going to do now is look at the dynamics at the bank level of our proxy measures in the model or in the data and to this s object in our model around the financial crisis okay so the first thing i want to highlight is that what we find is that banks who had large changes in market value tended to also have large changes in customer capital right that's true both in the model uh and in the data regardless of our our proxy measure right so this is our interpretation where customer capital is sort of expended right i experience these drops of market value and then my measured customer capital goes down right so maybe it's a it's a dampening so it's a way to sort of dampen the impact of the fundamental shock that happened in the financial crisis secondly we can ask well then what happens after that you know much of our modeling work is devoted to thinking about recoveries okay here we estimate that banks that had large drops in customer capital on impact of the shock actually had large recoveries um in market value thereafter okay that that change in since they change in market value that's a pretty bad typo but the um and that comes through with these negative coefficients we see evidence for that of course that the scaling here is difficult to have an apples-to-apples comparison but we do see sort of negative coefficients across the board so these drops in customer capital implied accelerated recoveries okay i see them like right on time here so let me wrap up okay um so we presented a model with imperfect competition arising through customer capital or habits which have persistence that's internalized by banks and these banks are also facing sort of a standard suite of financial frictions right the key thing out of the customer capital block today's pricing decisions affect tomorrow's loan demand key thing of the frictions block means that banks need to expend something to try to smooth out these shocks right because their net worth is a key state variable for them okay that delivers this more subtle relationship between prices and aggregate demand those three moments i talked about in the cross section banks want to maintain customer capital that gives a precautionary motive dynamically it makes recoveries more sluggish and we find evidence of this mechanism in the bank level of the data at the bank level in the data okay let me wrap up there um hand it back over to miguel all right right on time uh yeah so uh we've uh we've done the 30 minutes so let's just like keep going and uh let's move now to will sharing the screen [Music] uh can you make it full screen uh yes what's up full screen can you hear me yes we can okay and the screen looks okay perfect great thank you so much for including the paper on the program uh and for your attention listening this is joint work with jin yang zhang and yiming mai i believe hebing is in the audience hopefully maybe she can help me answer questions and we're studying the reserve supply channel of unconventional monetary policy so qe has been one of the main responses of central banks to recessions in recent years uh just in the u.s the large 2008 and 2020 crises uh both were responded to with conventional monetary policy cutting interest rates but once rates got to zero quantitative easing was the main tool for additional stimulus beyond that other central banks have done similar things and there's been a lot of work studying the effect of the assets purchased in quantitative easing broadly speaking the idea being that if say treasury bonds are purchased the prices of treasury bonds might be affected but there's been less emphasis on another key feature of quantitative easing which is the way that the federal reserve pays for these bonds or mortgage-backed securities or whatever they happen to buy whatever they're buying they're always paying by creating bank reserves so they are swapping some kind of long-dated asset for an overnight asset effectively that has to live within the banking system only banks have reserve accounts at the central bank this is effectively like a special bank account only for certain connected banks which can often pay a higher interest rate that is available to other investors so what qe does that as effect is puts large quantities of these liquid assets into the banking system that must live there and we want to understand what is the effect of this injection of trillions of dollars of bank reserves into the banking system on bank borrowing bank lending all the classic questions of what a bank would be doing and the main answer we're going to find is that every dollar of reserves created by quantitative easing crowds out 13 cents of bank lending primarily to firms so we study a model with multiple dimensions of choices banks can make but empirically what we find is that lending to firms particularly syndicated loans is crowded out by the issuance of these reserves that have to be held on bank balance sheets we answer this question with a structural model which is effectively supply and demand for bank lending bank deposits mortgages all the classic things banks do integrate it into a joint framework uh and one of the key things we're trying to do in this paper is identify the model entirely off the cross section so we don't want to act like qe itself is some kind of an exogenous shock we're going to try to rely on variation we do think is exogenous or at least more so to identify a model in which we can then simulate a counter factual of injecting reserves into the banking system and with this model we reach the answer that we crowd out this lending to firms 13 cents per dollar of reserves injected so in principle if you look at the various banking theories that exist you might say that this would either increase or decrease lending quantities if you think that reserves are a scarce liquid assets that banks are really constrained by this liquidity mismatch between illiquid loans on the asset side liquid deposits on the liability side increasing the supply of liquid assets they can hold might allow them to ramp up their lending this is the classic money multiplier logic you know this is built into reserve requirement type ideas cash up and stein is one paper along this line but if you think that reserves are not quite so special if you think that they're similar to the other assets banks hold the benchmark there is if you have constraints that costlessly expanding your balance sheet holding more of one asset will naturally make you hold less of others and in this idea where we think of reserves and loans being not so terribly different from each other holding more reserves would naturally crowd out bank's desire to hold loans and therefore might reduce their lending to the private sector empirically this is what we find but because the theory could go either way naturally this is an empirical paper rather than an a priori result so here is some very low tech evidence consistent with our findings this is just raw time series plots which is the fraction of bank's assets that we're calling illiquid this is everything except reserves treasury bonds and agency mortgage-backed securities this crashes over time as the supply of reserves increases which is driven entirely by qe so every time there's an asset purchase of qe reserves are used to pay for that purchase so this is sort of a measure of how much has been purchased this aggregate size of the qe program it grew after 2008 peaked in 2015 some tightening of the balance sheet and then in the covent crisis another big spike is there is a another round of stimulus now this time series trend is not causal necessarily because qe is a response to severe economic crisis we might just be picking up on what happens to bank lending in crises not what happens when there is an intervention by the market uh similarly we don't want to say that you know covered vaccines cause people to die of covid but in a correlative sense there are more coveted vaccines given out when there are a lot of covet deaths if you look at the broad time series so same kind of issue we want causal effect when the aggregate is not so clean and that motivates this structural approach we're taking using micro data and we really need to answer two key questions one is how elastic is the demand for bank loans and deposits so if there's a change in the cost of providing loans and deposits how much does that pass through into equilibrium quantities and then the second thing on the supply side is how much is injecting reserves into the banking system change the cost of capital for banks to provide lows to provide deposits and so on once we've estimated these two key pieces that identifies the question that we're interested in this counterfactual of the reserve injection this is the supply and demand side of answering that all right so our main finding is this reserve supply channel of qe where every dollar reserves crowds at 13 cents of lending deposit and mortgage quantities are effectively unchanged and the main reason why we find this is that bank loan demand is considerably more rate elastic at least at the aggregate level that we're interested in than deposit demand and mortgage demand so the interest rates change here by comparable amounts but the quantity changes are almost entirely in the syndicated loan data that we look at so we write a counter factual where we pick the size to match that there was about a 15 basis point difference between the yield on reserves and discount rates for similar investments the fed funds rate available to non-bank investors uh this takes a 4.23 trillion uh reserve injection in our benchmark counter factual which is similar to the size of the fed's actual balance sheet how many reserves are truly injected it's a little bigger but similar that 15 basis point passes through to about a third of that a little bit more to deposit in loan rates and that leads to this relatively large reduction of about 550 billion of loans to firms reducing because the demand elasticity there is so high and the key mechanism here is banks have to hold reserves that's why we get the spread between the reserve rate and discount rates for investments available to everybody else and that there's this cost of adjusting bank balance sheets they just can't costlessly borrow and lend and pass it all through and there's reasons you might think that has something to do with the regulatory environment which i might get into all right so this is related to a lot of papers that have studied the asset side of qme the effect of the app is purchased what that does to the term structure of interest rates and so on uh this paper also relates to the literature and banking theory on asset liability synergies on bank balance sheets which is a crucial thing for our transmission mechanism what we're interested in is how does a bank's holding of liquid reserves affect its cost of providing loans and deposits which is a connection between an asset that the bank holds and for deposits at least a liability that they issue banks only exist because you want to bundle lending and deposit taking jointly and one of our estimates is the first to quantify how much the quantities on one side of the balance sheet affect the costs on the other and this also really so a large structural literature where our methodological contribution is we rely entirely on this cross-sectional micro data that lets us use what we think are relatively credible instruments where the same bank is active in different regions of the country uh instead of bank level variation where these instruments wouldn't be possible all right so here's the model it exists on one slide but it's a slide i should go kind of slowly on uh there's going to be a bunch of banks competing with each other which are each going to face residual demand curves in the various markets they they're active in so the quantity of loans i get to make depends on my interest rate if i'm bank i that's rli but it also depends on the vector minus i of everybody else's interest rates as well so i have some market power it's not a perfectly competitive market but also there's competition as well because other people can steal business from me too so there's going to be these residual demand curves each bank takes and then the banks are paying what we're going to call a liquidity cost which depends upon the composition of their entire balance sheet this is where we're quantifying this notion of asset liability synergies that for example securities s that includes reserves that's our liquid assets uh the quantity held there might affect the marginal cost of providing a deposit or providing a loan but flexibly we need to estimate this entire function of how each balance sheet quantity affects the cost of providing everything else and banks are going to maximize their profits which is going to be a spread between the interest rate they either pay or charge depending whether it's lending or deposit taking over a competitive rate times the quantity and then in addition to that we're going to subtract off this liquidity cost so let's say they have a 20 basis point markup on loans above marginal cost times you know a certain quantity you multiply that together and get profits banks are going to maximize their one period profit by picking all their interest rates and that's going to give us first order conditions of marginal revenue equals marginal cost that's the standard thing in either monopolistic competition or pure monopoly reserves are going to trade at a competitive market so there the reserve interest rate which is common across banks that's set by the fed is going to be set equal to marginal cost all right so here is the model and a picture so marginal cost equals marginal revenue is the standard first order condition when you have imperfect competition the intersection of the marginal costs and revenue curves determines the equilibrium interest rate that's what's chosen to maximize profits then once you have that you plug that into the demand curve and you get the equilibrium quantity so this is how prices and quantities are set for the lending market let's see what happens in our counter factual let's shock this with what we think the reserve injection is doing this is what we find empirically that more reserves on bank balance sheets raises their cost of providing loans i'll show you my estimates when we actually get to them and that is going to shift the marginal cost curve and drag it across the marginal revenue curve to a new equilibrium interest rate that gets plugged back into the demand curve to get a new equilibrium quantity so this tells us what we need to know to figure out how our counterfactual works how big is the shift in the marginal cost curve and what is the slope of the marginal revenue and demand curves if we knew that then this one picture simplification of the model we know how the counter factual works and that tells us really what we have to estimate these residual demand curves once you allow the banks to compete with each other that's a demand system rather than a demand curve we're going to estimate that industrial organization style and there's this broad intuition that if you want to estimate a demand curve you need an exogenous shock to supply i'll show you the one we use for that and then to estimate the supply side of the model to trace out the marginal cost curves you need shocks to demand and we're going to have multiple of those as well the reason why we need multiple is because the bank is a multi-product firm in some sense it provides deposits and loans together not separately so really the marginal cost curve isn't just one-dimensional it depends upon the entire composition of the bank's balance sheet that identification question is one of the key difficulties in this paper all right so the first step is to estimate the demand side we want to exogenously shock the supply of deposits and loans and so on we're going to have instruments to do that across regions of the country and we're going to therefore use micro data at the bank branch level so we have fdic for bank deposit quantities rate watch for interest rates humda for mortgage quantities rate watch for mortgage rates and then for lending to firms we're going to use deal scan which is a firm bank match data set and we're going to use that for state level markets definition there so that's still going to give us geographic variation based on where the depart the borrower's location is their headquarters and then we're going to use the call reports as a bank level skeleton to put this all together all right so to identify the demand side we needed exogenous shock to supply we needed to change the interest rate offered to borrowers or depositors depending upon which market we're looking at in a way that does not affect the desirability of the loan or the deposit taken so we can't have it be atm machines are dropped as well as an interest rate change because that'll change how desirable the deposit is as well the shock we're going to use follows a paper by quarters and strahan so our goal was to use shocks which were up to the standard of the reduced form literature in our structural model and this paper shows that when there's a natural disaster in a region that region itself people tend to say borrow to fix their house after it got hit by a hurricane but the banks that provide that credit are going to reallocate funds away from the rest of their bank branch network and this is going to lead to a negative loan supply shock not where the disaster happened but in regions that have bank branches of the same banks that are active where the disaster happened so there's reallocation within bank branch networks after a natural disaster gives us this indirect loan supply shock at the rest of the bank branch network the assumption for validity is disasters in one region do not directly affect loan demand and others we have to say that the changes which propagate through the network are tr are solely through the supply side that's the assumption of the instrument uh which we think is relatively credible so we're going to estimate a logit demand system which is the simplest sort of i o style demand system you can estimate it has the nice property that if we look at the log difference in two banks quantities in a market that is going to be linear in the interest rates they charge and whatever observable and unobservable characteristics are left over we don't want to run ols to estimate this because the unobservable characteristics might be correlated with the interest rate if you've got a great bank branch network people might desire to invest there and that might be reflected in the optimal price you choose so we're going to estimate this by two stage least squares using this reallocation after disaster instrument i just told you about and because this is indifferences we have to put market level fixed effects to identify this parameter alpha which is how much your desirability of a bank is affected by a change in its interest rate so here's our results and the top co the top row here is the interpretable things these have the interpretation that a 10 basis point increase in deposit rates at an individual bank branch gives it about 4.6 more deposit quantity but a little bit over 50 percent drop in mortgage or loan quantity so for an individual bank branch the deposit market is considerably less elastic than either the mortgage market or the loan market but this is identified only looking at cross-sectional differences between two banks we're also interested in the aggregate elasticity if we everybody changes their interest rate that would difference out here and we need to know how the whole market responds by an aggregate interest rate change if we're interested in macro type counter factuals so the last thing we're going to do and this is new econometrics which i don't have time to get into here is there's a way in these logit models where you can construct a single measure how desirable it is to borrow at the composite good of like any bank branch in the market and then we can take our instrument aggregate it also up to a measure of how much disaster damage propagated through everybody's bank branch networks to become a loan supply shock here and we can regress this market level quantity on an instrumented market level shock to loan supply to estimate the aggregate elasticity of demand and our interpretation of the results here these parameters say that when an individual bank changes its deposit rate 29 percent of the new business it either gains or loses is coming from outside of the market rather than being reallocated from the rest of the existing branches but for mortgages it's only eight percent so the mortgage was very elastic for the individual bank but at an aggregate level not so much for corporate loans uh we observe an outside option here we see the firms that do not borrow at all and following the standard io approaches this gives you uh given the size of that roughly a 40 for the number here so the loan demand system at the aggregate level that we're interested in is by far the most elastic that's why we find that the loan quantities are the most responsive to qe all right so that's the demand side of the model to summarize we estimated these demand systems first at the micro level comparing the difference in two bank branches then we did it at the aggregate level with market level shocks and we found that the loan demand system for syndicated loans and deal scan is much more rate elastic at the aggregate level than mortgages or deposits and that's why that's the area where we see the big quantity response the last step is to figure out if we inject all these reserves into the banking system what happens to banks cost of capital and that's where we're going to have to estimate the supply side of that cost function that quantifies how a bank's marginal cost of providing deposits or loans is impacted by its reserve holding quantities or the other quantities on its balance sheet that's what we have to estimate next so this is the side where we're going to throw demand shocks at the supply side of the model and see how we can trace out that supply or this sense marginal cost curve so we set marginal cost equal to marginal revenue as we always have that's the first order condition that in the model the banks choose we've estimated the demand side of the model so we can observe the marginal revenue now effectively that gives us realized data points of each bank's marginal cost so we now know it costs say 40 basis points to provide deposits in the given market what we don't know is how that is changed by an exogenous shock to the composition of the bank's balance sheet because we have multiple endogenous variables here we're going to need to use multiple instruments to identify this so this data of these realized marginal costs we've generated using our model we're going to see how that responds to exogenous shocks to the composition of the bank's balance sheet to identify this remaining function c so the way we're going to do this is the classic idea with a one-dimensional model is you shock demand it traces out the supply curve because multiple goods are supplied jointly we're going to see how moving all the quantities on the bank's balance sheet jointly when you shock them with an instrument change the bank's marginal cost of in this example providing deposits so we're going to use our disaster shock but not in the indirect way of reallocating across branches we're just going to say this bank got hit by a lot of natural disasters that raised its loan demand and then we're going to also use a bardic type instrument for deposit demand so if the bank is in a region that has a lot of deposit quantity growth at an aggregate level that probably is not driven by the bank that's probably the local economy and that's going to be a shock to the demand for that bank's deposits so for both of these two instruments we're going to regress all of the four bank level quantities on the instrument shock as well as the bank's marginal cost here of providing deposits and the interpretation of it is these four shocks of the quantities jointly measured by these gamma parameters cause this change in marginal cost caused by the kappa parameter what we don't know though is whether it was the change in deposits or the change in mortgages and so on individually which one specifically was responsible we just get a linear equation that jointly these four quantity changes cause this marginal cost change from our instrument we're going to do this for both instruments that'll give us a system of equations and the details are in the paper for a 30 minute talk i can't get into this but we write down the cost function then and we pick its parameters to match all of these iv regression results so that we're consistent with the micro data and we end up with the following cost function so this matrix here this is how a bank's quantities of deposits mortgages loans to firms and liquid securities affect its marginal cost in the the markets in which it's active so this has the interpretation just to give you a sense of the numbers let's put one trillion of reserves into the banking system and divide it equally across all the bank branches that exist uh we met we normalize everything by branch in our specification that's going to be 184 million per branch based on the data in 2007 and we multiply that by the parameters at the bottom of this matrix the bottom row that's how much reserve supply quantities affect the various marginal costs the 0.0203 multiplication gives you a 3.73 basis point change in the bank's required return on holding reserves think of that as a change in the spread between the reserve rate and the fed funds rate and leads to a 2.3 basis point change in the opposite direction and the cost of deposits another interesting thing is if you look at mortgages and loans their columns are very similar to each other so we didn't hard code this into the model but we get that loans either to households or to firms are relatively similar in terms of how banks consider them as an asset on their balance sheet and we also get this classic deposit lending synergy which is at the heart of banking theory that if you provide more loans it's cheaper to provide deposits if you provide more deposits it's cheaper to provide loans all right so once we've estimated all this this identifies what we need to run a counter factual of injecting reserves into the system so what we're going to do now is take a large quantity of reserves uh this 4.76 trillion quantity is what we're going to inject in our benchmark we do some stuff more calibrated to quantities over time in the paper and we pick that number because that's enough to raise the required return on reserves by 15 basis points roughly speaking that is this 3.73 per trillion scaled up to reach 15.

Uh it's a little more complicated because we re-clear all of our markets but that tells you the benchmark size so we're going to inject that amount of reserves that's going to give us a 15 basis point spread between the reserve rate and the fed funds rate as we roughly see in the data and we're going to let banks reoptimize all of the interest rates they pay in charge for lending and deposit making as well as trade reserves with each other in a competitive market and clear it at a new equilibrium so this is relatively high dimensional happy to talk about the methods offline if you're interested but the benchmark we're going to find is that 15 basis point and change in the reserve rates passes through to six basis points five basis points four basis points for deposit loan and mortgage rates on average but because the loan demand system was so much more rate elastic than everything else we get about a 550 billion basis uh dollar reduction in loan quantities to firms in our syndicated loan data a little increase in deposit quantities a tiny decrease in mortgage quantities so similar rate pass through but large quantity difference in terms of 550 billion of loan supply reduction and not much elsewhere so to summarize this is this reserve supply channel of qe that we're studying in this paper there's also effects of qe on say the term structure of interest rates mortgage refinancing because of rate drops and so on we don't want to say that that's not also part of the transmission mechanism but our model which has been designed to study the assets company outside of bank balance sheets specifically is designed specifically for this reserve supply channel where we injected trillions of reserves into the banking system and calculated a new equilibrium we estimated the model using these cross-sectional instruments to have the supply and demand framework we quantified these cost synergies which told us how holding reserves affect deposit and loan cost of capital and we came to this finding that every dollar reserves crowds out 13 cents of lending to firms leading to about 550 billion reduction in bank loans to firms uh when you aggregate it up to the the size we look at uh if you think that this is an unintended consequence this is negative uh you might want lending to firms as well as qe you could change qe in a couple ways relaxing bank leverage regulation the supplementary leverage ratio would loosen some of the regulatory constraints that we think make it costly to hold reserves uh it might also be possible through either reverse repo facilities or central bank digital currencies to allow assets like reserves to flood out of the banking system so that they're not constrained to be held by a specific actor and these might be some potential changes to make qe still potent in the ways others have studied but that the channel worth studying this unintended consequence of reducing lending could be reduced thanks everyone for listening and we have a few minutes if you have any questions yeah thanks uh can i can i ask a question okay uh yeah no this is this is fascinating i was wondering if there's if you guys have thought i understand that it's it's very hard to do but if you have thought about the fact that there might be potential substitution from like traditional lending which is like kind of kind of what you're measuring with those general syndicated what's called traditional lending yeah and just like more indirect learning as in like banks just like switch from like direct lending to just like holding bonds and securities issued by firms yeah i think for the the substitution the banks are doing here is to holding reserves in the first place so i think the substitution would be more that non-banks would be hold it would be lending to firms would be the direction because we're already filling up the bank balance sheet with all these assets as the side effect of qe where it's the other people who might have sold treasuries to get bought in qe they're looking for something to hold and this might motivate a few different things either bank firm substituting towards the bond market potentially we could look at that in the data we haven't gotten into that this might motivate something like the rise of close this is sort of a structure for non-bank investors to enter yeah that's that's what i was having yeah yeah so we don't have any clean identification on which of these channels would be the natural way to substitute but you know given that the clo market boomed in a correlative sense at the same time qe was happening it's consistent with our results at least that this might have been one of the motivating factors any more questions looks like we have one minute well um i guess we can uh sorry move on that's good okay thanks everybody yeah thank you thanks a lot uh okay so let's move on to rustam um if you don't mind sharing your screen i'm doing it well yep looks good yeah so uh why don't you just take it away okay uh thank you uh everybody thank you miguel for organizing uh such a great team of like-minded people so today i will be presenting watch something that used to be my job market paper now just a paper a macroeconomic model with heterogeneous banks the basic theme of this work is very uh very simple the idea is to study uh the intersection of the macroeconomy and the banking sector in a way that is consistent with microdata when i say study i mean looking at positive normative and policy implications and when i say microconsistent i mean having aggregate implications that are in line with the distribution of banks that we observe at the micro level so the purpose of the stock would be to convince you that having environment having frameworks with bank heterogeneity is essential for essentially any kind of policy relevant question that we might ask but the ones that i will be pushing more a little bit today are three facets efficiency competition and stability each by itself i guess we can all agree or something that we really want to to study um by itself when it comes to the financial sector so we want the banks to be efficient uh we want to monitor uh competition and we want to prevent financial crisis so we want to maintain financial stability so it turns out when you put these three um uh factors into an environment with a realistic bank size distribution then it turns out that these three factors are not uh are mutually incompatible so you have a trilemma for bank regulation so comp again if efficiency competition and stability cannot improve in response to no single policy instrument that you can uh that the regulator being the fiscal authority or the central bank could could implement so this is a quantitative theory but i will be uh trying to prove this this theorem in a quantitative way with with some micro data as a backup so to just to lay down some um micro data for us to have a picture of what i'm trying to do here so four facts for you uh fact one the the banking sector is very concentrated that's fairly obvious uh well at least for us who started the topic up to 65 70 percent of the assets are concentrated in the top 10 largest institutions that nothing that doesn't necessarily mean anything bad per se it could be an efficient outcome but it's it's a symptom potentially of some market structure um issues that we should study so when you actually take the next step and say okay so is it true that concentration in banking is a symptom of low competition and you you take existing state-of-the-art approaches and you measure markups for banks you find that markups are a high and b they correlate with size so that's the first argument for having an environment with size heterogeneity is that if you want to study in perfect competition then um that the computational framework has to be uh related to the size uh uh uh to the size uh distribution so in this particular context um a markups will be in increasing uh with size and by size i mean assets or net worth of the intermediary second there is a large empirical literature in banking that claims that um their economies will scale so what i'm plotting here is that the the leverage growth of banks positively correlated with total asset growth but you could find a less pretty picture such as marginal cost fall um with the size of the intermediary and there are many other statistics that one can track so so but a lot of independent studies have pointed that intermediation efficiency is something that is concentrated in the right now um so again if you want to if you want to study efficiency we need to to have some notion of size pinned down in and some degree of concentration in order to analyze anything quantitatively as reasonable and finally i mean financial stability if there is anything that we have learned over the past 20 years that's the first order issue but it turns out and it's not surprising at all that if instability is concentrated in the left tail of the distribution and here i'm only focusing on insolvency risk or exit risk so for now i'm not making any too big to fail claims i will return to them later so merely from the point of view of exit risk being be that forced exit or being acquired uh in an aggressive way by a large institution those risks are all concentrated in the left tail so again something uh a factor that is very strongly and in fact um monotonically related to uh two sides of the of the banks so um yeah so so the uh so you motivated with the uh with them uh with these um four facts i will be developing a framework and discussing it uh with you in a moment the model will be able to replicate these facts and we can study some interesting policy contractuals so very quickly a related literature so there is an explosion in the past two years of important papers on heterogeneous intermediaries um there are in terms of imperfect competition and banking there are two maybe more but i see two branches on credit market power and deposit market power so but my focus today is on the market power on that on the asset side of bank balances but there is a a very important complementary channel uh as well the deposit franchise is a source of market power for banks as well and there is also some growing empirical literature that is providing evidence for existence of what i call uninsured idiosyncratic credit risk or essentially biuli algari huguette imrahoglu type of frameworks for financial intermediaries that but immediately implies a presence of some kind of precautionary lending motive and we can bring all the tools of those heterogeneous agents models and by extension hank the hack literature to the case of um financial frictions and when we put those two together you know beautiful things can happen so let me um dive in straight to the model so yeah should be on time so the meat of the model is the capital boost producer uh so uh so the the first major departure from uh sorry the first major departure from a standard setup is the way capital will be aggregated in the economy so the red the real continuance of uh you can call them financial varieties j so j is an index of heterogeneity that will be uh that has multiple interpretations uh one my favorite one is geographical so if you could think of local markets uh and there is one bank serving serving all markets you could think of jay as having a product based interpretation there are fixed income derivatives equity um uh exotic uh equity the routes and all kinds of financial varieties and the borrowers require all of them in order to construct the capitals though or jay could have the bank level interpretation that literally there are 5 000 banks in u.s they're all competing with each other but customers potentially have you know in kyle's and mcgill's word they have words they have preferences for particular banks so banks enjoy some kind of customer capital base so um the aggregator is uh non-ces so not only am i departing from the efficient aggregating aggregation benchmark i'm also imposing that um markups are potentially uh size dependent so there will be not there will not be a constant markup which will be invariant across time and across the cross-section so the kingdom aggregator is amazing at that it will deliver essentially all the cross-sectional properties that we that we want um subject to the do the cable aggregator the capital goods firm so do you solve the usual problem and we find um a familiar uh to all of us uh demand function for bank funds so the other implicit assumption here is that the firm wants to produce but it lacks funding so it has to borrow from bank from banks in terms of claim in terms of claims on the future capital stock uh and uh uh because of um the aggregation technology banks will be able to extract markups from from roms so far so far i haven't specified this uh capital gamma function uh yet so it's it's a very general so now let me specify uh take a stance on the on the formulation and i follow what is uh by by now um state of the art is the clino wheeler's specification um to spare you some of the technical steps that is needed to derive the markup function this will this is how it will look like this mu of um mu as a function of y where y is the is the relative size of the bank so the markup will depend um on the relative sides of the bank in a way that makes markup markups uh increase uh with relative size so remember that's our fact two uh uh competition in the banking sector is low and also markups increase with size so the aggregator will give us that property as long as the super elasticity is positive and the super elasticity here is epsilon over theta so the another cool thing about this approach is that the ces case is nested for epsilon equal to zero so immediately that shuts down the the idea that uh as the relative size goes up the elasticity falls then we kill that the channel and the elasticity is the same for all agents okay so we can nest the perfect competition competition case okay so now to the the to the banks aside of the problem uh so so the first equation is the standard balance sheet uh constraint so on the right hand side we have their um assets they invest into claims on on on on capital that will be produced by uh by the capital goods producers they they accumulate net worth um they can borrow from households and the deposit deposits will be priced competitively together with households i will describe how on the outside side uh of the balance sheet but the banks um receive two types of returns so um their total portfolio return is a weighted average of the systematic return on the neoclassical capital stock and the idiosyncratic component um side j so x i j is the uninsurable idiosyncratic rate of return risk which is uh you know has the j bracket meaning that each bank or each local market or each product receives its own um j-specific return draw every every period and as long as um we have a little bit of persistence a distribution of bank size will emerge in equilibrium so uh there is a reason there is there is a historical reason why this process is air run in fact in a parallel paper using norwegian data we have estimated that process exactly that process for the norwegian bank firm administrative data so there is no reason for this particular formulation except for that you have some idea of what uh raw and stigma are how the correlation and volatility are but in principle it could be your your favorite specification so in the spirit of um get your kia duck and get karate banks face a leverage constraint which has a moral hazard um intuition law of motion of net worth is what you see in the in the middle here uh next period uh banks receive the return on their portfolio investment they pay off the depo the depositors and they also pay a non-interest cost and then it's going to be important for me that um that they are uh non-linear so non-linearity will break scale variance and pin down the size of the intermediary if you take away non-interest cause the the model will be linear with respect to net worth and there will be no notion of size to speak of um finally there will be endogenous default again it's here it's only in solvency risk so i don't have bank runs for instance and the influence for those could be introduced but it it's not not it's not in today's paper so the the this is the the bank's uh dynamic problem so they they that's their um value function so they they they maximize the stream of future discounted flows of network they pay off the the households in terms of dividends and the dividend structure is very simple for tractability so it's a it's a fixed and time invariant fraction of their of their profits goes to the household and the remaining fraction is is kept in the program and here the marginal rate of substitution this is the sdf of the pro of the household with the which the banks borrow and then augment with uh the dividend uh payout rule and the constraints that banks face are the six equations which are listed in the previous two slides plus the demand for financial varieties which comes from the cable aggregator so and banks internalize the idea that that uh that they could exert market power but of course they don't internalize their impact on aggregate prices so this is where we will have the aggregate um externalities so just just to clarify that kind of the main deviations from from garth lake otaki for example here would be the the non-ces demand the market power part and that uh cost that comes from your size right yes so there are three deviations you're right so so the first deviation is because uh the beauty of ghetto kyotaki world is linearity so i want to break it i want to make the model intractable on purpose so that's why we need these costs second imperfect competition that's the aggregator and third is idiosyncratic return which is persistent they also have idiosyncratic shots but for them they are id persistence make forces me to track uh you know the distribution over time thank you is it important to think of your notion of default here as negative net worth as default like how how would your sort of trilemma results go through if i were to say you could you know issue equity your way out of that at some cost is that first order great great question so uh i i don't know so i i so they don't have the banks here don't have um the option to issue equity just like they don't the the households equally don't have an option to extract higher dividends from bigger banks so sigma could be potentially also js so uh yeah so but so i don't know if it's first or second order i don't like to speculate i love to do it by brute force to actually solve and see but i yeah the quick answer is i have no clue so um the remaining parts uh in the interest of time i will skip so there are some useful but not essential components so there is enter and exit um similar to mellits but the mellitt's structure is a one-shot entry for me because the model is dynamic there is entry every period uh and again the way it works attention aspiring investment bankers they draw their uh uh the startup realization of um of their idiosyncratic profitability they receive some net worth transfer from the household if they like they would what they received they they become a bank and they and the mass of intermediaries hd which will grow up so that that slightly complicates the the model because now i have to keep track of the the mass of entry but it it's it's it's fine so uh the household problem is a fairly standard uh and the first order condition of the household problem pins down the deposit interest rate which is importantly bank specific so every bank type will will have its own risk profile dictated by the by the history of idiosyncratic draws and the the size the network network so they're just idiosyncratic state variables perfect information perfect visibility and households know exactly who they are dealing with the and they price competitively all the risks that they see an important assumption here is that there is no deposit market power so there is perfect competition on the deposit side which potentially is also something that could affect my main results but it's just the models which are at the at this at today's stage it would become a bit too much for me to handle but maybe in future work that could be introduced the final good producer is is very standard capital depreciates uh every period it it doesn't affect the results it just makes my life simpler because i'd otherwise i would have to create to artificially create a new price of capital so so and to achieve a stationary industry equilibrium all markets must clear okay now to the do the main uh result and i will walk you through results in the following way so so the most important equation that comes out from the model is the decomposition of relative prices uh relative prices on on on firm claims into markups and marginal costs so um remember as long as um the super elasticity in the kimbo aggregator is positive we will have one of the four facts that markups increase with size so we would get the endogenous competition channel right thanks to the uh the uh the flexible aggregation technology and calibrating the sign correctly marginal cost for as long as there is some presence of incomplete markets as long as um the idiosyncratic shock is at least slightly uh persistent and cup is positive i mean copper is them is the weight only the socratic risk in the total portfolio return formula so this will always give me the um the second fact that marginal costs fall with size and here the intuition is a little bit more complicated as banks accumulate more net worth they become more efficient because the only reason why they are big is because they have a good history of idiosyncratic draws so their marginal cost will be will be low because um they have accumulated a lot of net worth which is an endogenous outcome of their luck so they're they are efficient because they have been lucky for a long period of time and there is also the another component that falls into marginal cost it at that's the default risk that's teens or the as correctly corrected me it's the really the exit risk uh of intermediaries so it will it will fall with uh size and will get financial stability right the intuition here is straightforward the the the probability probability of exit is mechanically linked to the size of the balance sheet so the smaller uh net worth gets the the more likely you are that you will draw a negative enough in your socratic shock such that you you uh you you get um you exit next period and you may or may not be replaced because remember entry is endogenous as well so the main result is really this picture so by by introducing it using chronic shocks and having scale invariance we get a concentrated stationary distribution of bank size as part of the solution concept so this idea goes back to you know decades ago so it it's essentially operationalization of the building agar environment for bank so as long as you have an insurable idiosyncratic risk some notion of scale variance and you get this right skewed distribution of uh of of agents so you we get we have some degree of concentration we don't get i don't get concentration completely right in the sense that it is hard for the model only with idiosyncratic shocks to get the hairphendel index of the banking sector that we see in the data to get 65 of assets concentrated in 10 banks you need you need more tools and that's something that i'm working on right now so on top of this um on top of the core skeleton of the model that just gives me size heterogeneity now we introduce the kimbo aggregator we have the imperfect competition uh idea so here uh overlaid on the on the histogram are the absolute credit markups so just like in the in the empirical motivation before we see that markups will be increasing um in in bank size so that's the endogenous competition uh channel so the economies of scale channel can be displayed by by simply plotting them the model implied relative marginal costs for all banks uh so the marginal cost is the object that i showed you on the previous slide where i decomposed prices into markups and marginal cost so markups fall with the size of the intermediate and you we already see some interesting uh potential problems here because as the banks get larger they're more efficient so marginal costs are lower so potentially we can get a bigger bank for the buck from them they can they might have a higher marginal propensity to lend but higher concentration immediately implies lower competition as the average markup increases with the right skewness of the distribution and that's already not a good sign for us from a policy perspective and finally we throw in the the endogenous exit so the financial stability channel here the black square the probability of default is event essentially zero for larger intermediaries that's not surprising but it can be quite uh the smaller the band gets so now we have this the size distribution and the three channels competition efficiency and stability so this is where the trial the trilemma arises so it it happens to be the case that the same banks that are stable and efficient they have low probability of default low uh black uh squares they are very efficient they have low diamonds they also have a greater credit market power because their credit the amount elasticity is low endogenously so that means so whatever i know that's a generalization that you have to sort of argue for but no single reallocative shock can simultaneously improve find uh computational stability and efficiency at the same time so any any policy shock that affects the third moment of the size distribution will will will impose a change in the allocation of credit in the economy if you somehow manage to come up with a shock that that just is just shift everything symmetrically and that the net impact on skin skewness will be zero then yes you might you might get something but um any taxation scheme any um a monetary policy expansion would not have the zero skewness result why because marginal propensities to length marginal utilities marginal values of net worth all of these marginal objects are they are very heterogeneous across the distribution so any aggregate shock that you can possibly think of would have um would have to go through the trilemma um uh the trilemma situation and on the on this slide i'm listing some of the examples that i have actually tried by brute force in studying them in in the paper so you might argue that well you might if the the social planner might i undo the trilemma but in fact it doesn't work so the social planner actually reinforces it because it will reallocate more more assets to to bigger banks they have a higher marginal propensity to land in fact the social planner will lead to a cascade of defaults of smaller intermediaries it really wants to have some few but big and efficient and stable intermediate results so again the trilemma holds you can impose size dependent capital requirements or deposit insurance schemes or the too big to fail externality again the trilemma halls and when i say the trailer i mean that if you calculate the aggregate levels of markups investment consumption and the average probability of default you will never have all of these objects improve uh at the same time so you will have a deterioration in uh in some dimension and you will improve some other dimension but you you will have to make a sacrifice um somewhere so and and there are other some there are other quantitative exercises i do in the in the paper which are now in a very long lengthy appendix so i look at the rise of banking concentration uh you know emergence of fintech credit so a lot of very um creative ideas but it's um really it's it's simply an interaction of three frictions which happen to match the data so um in conclusion so it's really um if just the framework to think about interesting stuff in in macro banking uh it it it's calibrated um operationalized to match some key cross-sectional patterns and um it provides us with a novel trilateral trade-off to think about classic and new uh policy relevant issues so it it's also not that difficult to work with so in in progress we have an extension with monetary policy and nominal rigidity that we uh very ambitiously called h bank and um an important uh you know probably first order uh limitation of this theory is competitive deposit pricing that's something that should be worked on thank you very much uh we have like a minute for if in case anyone has a question um [Music] yeah what makes you think that deposit market power would be first order here because my intuition is that that would point exactly in the same direction that you have that in that is correlated with loan market power it would just be a different manifestation of the big banks in your model have lower funding costs through lower default risk it would just be they have higher markups or you know higher markdowns i guess um that's in fact that's exactly what happens and that's also true for your framework from firming if you have um say um deposit client capital that would also be correlated with networks in fact that would only reinforce the results but but it is a a thing that people really want to see this aspect all right uh let's move to the next speaker pfm um okay we can see your screen all right can you guys see my screen all right yes we can see when we can hear you so you can take it away okay well great so thanks very much for being here thanks miguel for including our paper in the program this is a joint work with uh olivier wang from nyu stern on intermediary loan pricing which as you will see is an analogy with intermediary based asset pricing so the goal of this paper is really to understand how bank loans react to different types of shocks that hit bank balance sheets you can think of credit supply shocks like a credit expansion or a credit contraction you can think of demand shocks borrowers suddenly becoming riskier you can think of monetary policy shocks acting through banks cost of funds or through uh some for instance channels that uh well described or you know any type of risk shocks so what i want to do here is sort of break down this question into a three parts which i will motivate with the aggregate and micro data and then i'll sort of try to write down a model to jointly explain these these facts which we think is is the new contribution of the paper first fact that we want to draw your attention on is the strongly negative correlation between banks losses this charge of rates here for mortgages for households mortgages and loan growth so in other words when banks are making losses they are lending less so there's a strong negative correlation between if you will the health of bank balance sheets and the volume of loans that they are making the second fact which we want to draw your attention to is that this might be happening through different loan terms if you look at the bottom panel we've plotted the maximum loan to value ratio on these mortgages and the mortgage spread so what what might be happening is that loan volume is just decreasing because spreads are increasing which makes it less appealing for households to borrow or they this volume might decrease simply because credit constraints ltv constraints might might become tighter and indeed if you look at the maximum ltv constraint and the mortgage spread there's a positive correlation between them in other words there is a negative correlation between mortgage spreads the price terms on mortgages and the down payment requirements which are one minus the ltd constraint if you will so in other words there's some there seems to be some sort of substitution margin going on between between these two terms so for instance after the great financial crisis when maximum ltv constraints start being relaxed at the same time you see the mortgage spread increasing in as if banks were wanting wanted to sort of compensate for greater borrowing through the non-price margin by restricting this borrowing through the price margin and you see the opposite happening uh before the great financial crisis now this this is true for household mortgages this is also true for households credit cards and here in this talk i'm just going to focus on household data we have firm data in the paper for credit cards as well we do see this strongly negative correlation between banks losses this dashed line and banks loans and and these and these credit card loans that banks are making however here the dynamics of credit standards this blue line which is from the sluice that's the net percentage of banks tightening standards and the spread are somewhat different from from mortgages with uh the spread increase being a lot stickier uh sorry i can plug my left with this increase in in in the spread being a lot stickier for credit cards than it was for mortgages and on the other hand the increase the tightening in non-price terms being a lot a lot more short-lived than it was for mortgages so this is sort of this second fact so there's a cross-sect so first there's a negative correlation between the health of bank's balance sheets or so if you will in other words there's a positive correlation between the health of bank's balance sheets and the volume of loans that they are making second there is a cross-section of loan terms not everything is going through a price adjustment something is going on through non-price terms it can be ltv constraints can be pti constraints for mortgages it can be credit standards if you look at credit cards if you look at firms it can be governance and and and that's in the paper so these are the first two facts third fact is that these these these these relationships if you will are strongly different when you look within a given category of loans that's strongly different between borrower types so here what we've done is we've looked at risky households with low fico scores in red and safe households if you will with high fico scores in blue what we see is that after the the great financial crisis there's a strong decrease in the in in the typical loan that is made to a risky borrower while it's more or less unchanged for a safe borrower it might be puzzling if you just look at price terms on on on the upper right panel if you look at loan rates they have very similar dynamics across these risk categories so what might explain this is more more probably this very strong decrease in the ltv constraint of risky borrowers while we don't see that for for safe borrowers in other words these dynamics the dynamics of these price terms and non-price terms are very different across risk categories so that's this fact that we want to document here for mortgages and the same fact holds for credit cards so that's not that there's not a lot of easily publicly available data for credit cards so what we can do is is look at honest trouble and aggro wallet a qj paper and on credit cards where they find that for credit expansion now so for a one percentage point reduction in banks cost of funds so for this credit expansion it leads to an increase in the credit limit of risky borrowers of only 253 dollars while on the other hand it leads to a large increase in the credit limit of save borrowers in other words credit expansions are passed through to safe borrowers who do not need to borrow so much if you think about it that exactly the opposite story of what happened with the mortgage expansion of the 2000s if you if you think of of the mian and sufi paper where the where the credit expansion on the mortgage market primarily went to risky borrowers to these to these sub-prime borrowers so the goal of this paper is to sort of try to make sense of of these of these three facts to jointly explain them in the simplest framework as possible so just to recap what we've seen from aggregate and micro data is that first there's a cross-section of multiple loan terms so there are both price interest rate and non-price terms like covenants like ltv pti ratios and there's a cross-section of them meaning they differ across borrowers with different risk second this cross-section of multiple loan terms responds to shocks to bank balance sheets and to policy we've seen this strongly positive relationship relationship between banks health and the volume of loan that goes through these price and non-priced margins and finally these loan terms respond very heterogeneously across borrowers with with different so here again what we want to emphasize in this paper is that both price and non-price terms adjust so what we're going to do is we're going to build a framework which which we think of as intermediary loan pricing in an analogy with intermediary asset pricing where bank balance sheets affect the terms of the loans that banks are issuing to different types of borrowers on on different markets what we're going to do is think in this framework about the heterogeneous incidence of shocks to banks balance sheets so here in this talk i'm going to focus on credit supply shock so an expansion or a contraction in bank's lending capacity and time permitting on on a change in bank's cost of fund through a typical monetary policy shock we're going to see how they affect differently mortgage and credit card markets risky versus say borrowers and in the paper we sort of try to speak to the evidence on firm versus households these two these these two um approaches these two points if you will have implications for the dynamics of of credit crisis and here we're going to show how very simple elasticities have uh can capture the impact versus the persistence of other shock and how they affect the entire dynamics of of a credit crisis so what we're going to do in this paper is we're going to try to speak to both the micro and and the macro evidence we're going to add to a frameworks uh like uh stiglitz wise on at the micro economic level in which you know there's credit rationing but interest rates or spreads are fixed so you have on you have some credit rationing on the quantity side but you have nothing happening on on the price side and on the other hand you have home strong zero where that's the opposite the spread is endogenous but there is no credit rationing so we're going to try to uh have both credit rationing and endogenous spreads within a model so in other words we're going to try to endogenize both this price and non-price terms of credit contract to speak to the rich microeconomic evidence that both these terms are important on the macro side we're going to contribute to a some of the literature like many papers that were presented today for instance by um by kyle or or by will to these sort of macro asset pricing literature which focuses on the dynamics of credit crisis and and the role of of banks uh balance sheets uh for the for this credit crisis so without further ado let me uh jump into into the model so the goal here is going to be to build a model that's as simple and as general as possible which potentially connects different quantitative frameworks key ingredients of the model are going to be first information asymmetries which are going to generate endogenous default risk you can think of you know adverse selection on loan markets you can think of moral hazard as well we have simple examples in the paper where we show that our framework nests both second loan contracts are going to be multi-dimensional they're going to have both price terms interest rates but also non-price terms so quantity restrictions and potentially non-price non-quantity restrictions such as covenant restrictions or documentation restrictions third banks are going to compete for borrowers subject to our capacity constraints on lending so what what's the problem so again we want to think of us we'll think of this problem in in the most general framework as as possible so borrowers have an expected indirect utility function over a loan contract so ci for borrower i with a rate or i some quantity limits l i and some you know non-price terms like covenants that i to to the contract and they have some expected indirect utility over that contract on the other side of of this trade you have banks with an expected profit over a contract to borrower i conditional on on take up banks are subject to a landing capacity constraints on their total loans to cross-section of borrowers which must be less than some lending capacity constraint elbar you can think of it as regulatory constraints like you know bazel three kind of uh restrictions you can think of that as standing from you know deposit market powers it's hard to attract deposit to fund loans there can be many explanations for for this constraint we're not taking a sense on on what it is importantly the different loans for different borrowers have different weights in these capacity constraints which will restrict more or less the loans to these students borrower class so you can think of these risk weights row i as risk weights simply in a in a bazel uh type of uh regulatory constraint you can also think about them as the ability of the bank to securitize the loans take them off its balance sheets and therefore you know so so that they do not matter so much in in the lending capacity constraint what's the problem because banks compete to attract borrowers that's going to be a bertrand nash equilibrium with capacity constraints what's the problem of each bank b so each bank takes as given borrowers outside options which we call vi bar and they try to maximize their expected profits to all borrowers by offering different types of contracts subject to their lending capacity constraint and subject to a participation constraint on borrower's side for borrower i we want to give that guy at least their best outside option to attract them so what's the equilibrium well an equilibrium is for a given bank to associate to a given borrower some contract cib and whether these borrowers is being lent to which is going to be the case in equilibrium so such that borrowers optimize that is the best the outside option of borrowers is really the best option that they can get when they competitively search across banks if you will and second markets clear in other words every every borrower is is getting is getting their loans that's a question here so you have the bank choosing loan terms and a price and offering it is this similar to they offer a menu of contracts and that's correct choose if these are competitive is that equivalent sure that's correct it's an interpretation of our framework i'm going to come back to that yes all right so what are the equilibrium conditions uh that we want to focus on here so here i want to focus on a symmetric equilibrium across banks and the equilibrium is described by two conditions which are actually quite quite crucial to understand what's going on between between these facts that that i've shown you so first condition relates to whether borrowers are credit constrained why remember v is the indirect utility from borrower i so if the derivative of v with respect to the loan size is positive it means that borrowers would want to get more loans to increase their utility in other words they are credit constrained they are credit ration if you will if borrowers were on their loan demand where on their loan demand then they would then they would borrow as much as they want and this term would be equal to zero so when borrowers are rationed in other words when this term is positive what happens well it must be the fact that this term is positive as well which means that banks sorry that this term is negative given the sign of this term the fact that this term is negative means that when banks lend more to some borrowers they are actually decreasing their profits why can it be well it can be because there is default risk and i'm gonna come back to that in more detail on the next slide in other words to recap borrower credit rationing is related to the fact that there is some default risk which is increasing in the loan size that that a given borrower gets now the second equilibrium condition is that banks must equalize expected profits risk weighted across borrowers well if it were not the case then banks could increase their total profits by lending more to the guys with a higher expected profits those can be zero all those can be positive when are they positive when the multiplier new on the bank lending capacity constraint is positive so in other words new is the profit per risk weighted dollar when banks are constrained banks are making profit on their loans when banks are not constrained banks are competing and virtual competition drives their profits down to zero so opening up this expected profit uh condition we can we can really understand what's going on with with this endogenous endogenous default uh probability so we can write expected profit again so as what banks are lending to borrowers minus their expected loss minus their cost of funds minus some costs potentially zero this can be the cost associated with with non-price terms if there is more say documentation requirements that that's costly for the bank what we can do is rewrite these expected profits with by using what we call an effective default probability mu what this effective default probability does is it encapsulates the loss given default of the bank right so if if if i was really to you know break down this expected loss i could factor this r times l term and make this endogenous default probability appear this framework nests liquidity strategic default which stem from adverse selection or or debt overhang that's in the paper i'm not going to spend too too much time on it but what's really crucial here is that using this default probability mu we can rewrite the first equilibrium condition remember the one on credit rationing we can rewrite it in a way that makes this term appear what's this term this term is the elasticity of the repayment probability to the loans to to the loan phase value what does it mean well if i'm a bank and i lend you more by how much do i increase your default probability why would i increase your default probability well if default risk is endogenous the fact that i lend you more i make it also more difficult for you to repay a bigger loan size in other words what what's what what this condition is really telling us is that borrowers are credit constrained in other words this style here is positive whenever there is endogenous endogenous default risk whenever the elasticity of the repayment probability to the face value of the loan is positive all right and so together these terms are going to define what we call a virtual loan demand what's the virtual loan demand gonna be it's gonna be as as will was was saying sort of a menu of contract that is gonna say well if i give you some interest rate or i how much am i going to lend you that's going to include the effect of bank's capacity constrained again just to recap this style here is going to measure how the how constrained a borrower is going to be and here what's going to be crucial when when we come to the effect of shocks is gonna be how this how this how the loan quantity that's given to borrower when banks are constrained changes in response to shocks we're gonna see that it it depends on two elasticities it's gonna depend on two elasticities which will be sufficient statistics for the response of these loan terms to shocks the first one is gonna be the elasticity of the unconstrained loan demand imagine that there are no capacity constrained by how much does lending react when the interest rate changes if you are a borrower and i increase the interest rate by how much is your loan demand going to decrease the second elasticity is going to be this repayment probability whereby default is endogenous so again here what what this slide is really telling us is is the reason why there is a cross-section of loan terms and not just interest rates the reason that there is also non-price terms is that if banks were just lending more to borrowers at a higher interest rate with endogenous default risk it would also lead to a higher default risk on borrower side and it would reduce banks profits to control that banks need quantity limits not just price terms in other words the reason why we have non-price terms here again is because of endogenous default risk in the interest of time i'm going to skip the intuitive construction of the virtual loan demand the idea is that this loan demand comes from the intersection of the capacity constraint of the bank which may change because of a shock a credit supply shock a monetary policy shock and the indifference curve of the borrower and the expected the zero profit condition of of of the bank so this is where this virtual loan demand is coming from this framework has an important implication which is that if you look at the second equilibrium condition which tells you that banks equalized profits across borrowers and that those must be equal to the multiplier on the capacity constraint we can derive that there is in equilibrium an excess loan premium what's the excess loan premium is the interest rate that banks are charging to borrowers in excess of what is justified by the risk-free rate and by their default risk in other words this excess loan premium is the difference between the rate that banks are charging to borrowers and what is justified by this default part here and this risk-free rate part here what is the excess loan premium equal to it's equal to the multiplier on the capacity constraint in other words if banks are very capacity constraint the excess loan premium is going to be higher and it's going to be even higher if borrower i is important in your capacity constraint that's the first implication of this framework the second implication as will was hinting is that this framework gives rise to something that can be interpreted as a credit surface what's a credit surface well it's a menu of contracts that the bank is offering to borrowers with different risks so here what we do is we sort of specialize the model with a pareto income risk so the higher alpha is the riskier the borrower with liquidity default and we look at the menu of contract that arises from solving the bank's problem what do we get well for riskier borrowers if these guys want to borrow more if we want to increase the loan size they must also pay a higher rate what's the equilibrium that we get from here well we look at the tangent between borrowers indifference curves and these menu of contracts that banks are offering we can look at different risks and we can trace out this contract curve this control this contract curve is just telling us look we study borrowers with different risks what are the rates and what are the loan sizes that these guys are gonna get and here there's a super important implication which is that rates interest rates do not fully capture credit conditions as you see here as risk increases so does the interest rate to sort of compensate for the default risk but at some point the interest rate stops increasing why because if i am the bank and i further increase the interest rates i'm gonna make my borrower more likely to default at some point with endogenous default risk so what's going to happen is then i'm going to choose a non-price margin i'm going to impose a quantity limit on my borrower to limit this default probability as my borrower gets riskier all right we can do that for non-price terms i'm not going to do that in the interest of time but in the the five minutes or so that uh remaining uh time that i have left i'm going to look at the transmission of of shocks within this model so usually in in sort of macro finance models we we think of you know credit supply shocks as just shocks for instance to a maximum loan to value or maximum payment to income ratios in mortgage contracts so here what we do is we really endogenize the variation in in these quantity limits in response to a shock so what's the shock here well think of it as a negative yes can i just ask you a very a potentially very stupid question but in your context um what what you call uh what you call a non-price term the z it's it's essentially uh some type of monitoring effort no can i interpret it that way you can you can definitely think of it as a monitoring effort as long as it reduces borrower default risk and it is costly for the bank right exactly okay but we also think as as you know as this quantity limits l which we had on on the previous slides as well as as non-price top so basically in our model everything that's not interest rate is is non-price terms all right so what happens here in in response to a negative credit supply shock so what what does such a shock do well it increases the the excess loan premium by five percent in other words the bank lending capacity constraint becomes more binding the multiplier increases here by five percent in our calibration so what does it do to uh the menu of contract to the credit surface there's a level shift in other words borrowing becomes costlier for everybody but not just a level shift there's also a steepening of this of this menu of contract which is that for the for these guys for the riskiest guys borrowing becomes extra costlier if if you will so that's really what what's happening and and this is essentially what what what this gra what what this uh proposition is is telling us so to summarize the response of the loan of the loan size of given borrower i to the to a change in the lending capacity constraint is given by the elasticity of the virtual loan demand the elasticity of the virtual loan demand depends on the unconstrained loan loan elasticity and on the elasticity of the repayment probability i'm not going to go into my in too much detail here just state that these two are actually sufficient statistics in our model to uh for to understand the response of the of the credit um of of the credit surface to uh to these uh to the to these uh to these credit supply chains but what i want to do here is really speak i i want to conclude by by speaking to the evidence that i mentioned at the beginning and the extent to which credit expansions differed on mortgage markets and on credit card markets in 2000 which we don't really have an explanation for so think about think about these these markets so first think about the the mortgage market which we think of as a less elastic market if you really go to the data try to measure these elasticities what happens in response to a positive credit supply shock well interest rate decreases and loan size increases that's really what a positive supply shock does if the market is not so elastic then the rate is going to decrease more then the quantity will increase so the face value at the end of the day will decrease this means that the default risk will decrease because these guys are responsible for a lower loan size and it's going to decrease even more for risky borrowers who are very sensitive to to a change in the face value of their debts because banks equalize expected profits across borrowers and default risk decreases more for risky borrowers they're going to lend more to the risky borrowers in response to this positive credit supply shock which is the subprime story now if we think about credit cards and and the aggro while at all qje evidence let's revisit this channel positive credit supply shock interest rate decreases size increases however the size increases more than the interest rate decreases because it's a more elastic market so the face value of the debt increases it means that default risk increases but it increases by less for safe borrowers who are not super sensitive to these changes and then because banks again equalize profits across borrowers it must mean that the loan volume increases more for safe borrowers so the credit expansion is passed more to this to these save borrowers in terms of policy implications it's important because it's really telling us on elastic markets credit expansions are not so much passed through to borrowers who need to borrow these risky borrowers but rather to the save guys with very high credit limits who do not need to borrow so much i think i'm out of time so let me wrap up and conclude here so what what we do here is really to uh to come up with this trying to try to come up with this framework which we think of as intermediary loan pricing where bank's balance sheets affect both the price and non-price terms of loan what's key here is that credit risk is endogenous due to micro frictions and that's really what differentiates our framework from the intermediary based asset pricing which speaks more to the to the dynamics of equity markets we show in the paper that and and here i've tried to give the intuition of it that sufficient statistics the elasticity of the loan demand and the elasticity of the repayment probability to the debt phase value drive the incidence of credit supply sharks monetary shocks and when you connect these successive periods they also have an impact on on on credit crisis through impact and persistence thanks very much thank you pf uh we're out of time for the session so i'd like to thank all the other presenters in this session and also all participants who came here in spite of some heavy competition i hope they uh learned as much as i did and uh yeah thanks again to everyone and i hope you have a great summer and enjoy the rest of the concert thanks all thank you everyone bye thank you you

As found on YouTube

Looking to see what kind of mortgage you can get? Click here to see

Leave a reply

Your email address will not be published. Required fields are marked *