Paper presented by Charles M. Kahn, Professor Emeritus, University of Illinois


Curt Hunter: Over the past 12 years, I've had the privilege of serving on the boards of the closed and open-end mutual funds of a very large U.S. asset manager; I managed about $180 billion in assets. As I look back over this involvement, I think it's generally agreed that prior to 2008, collateral management was not the "top of mind" topic in the asset management industry, nor among asset management companies or their fund boards more generally.

So as you can imagine, during that time, the intricacies of fund collateral management was typically the purview of securities lawyers, custodians, and other industry specialists. I think our collective experience with the financial crisis and the Lehman bankruptcy in 2008 really changed all of this. Many asset managers learned that much of the collateral that they had pledged to support their leverage programs for their funds, had actually been transferred to Lehman's U.K. subsidiary and then reused, or rehypothecated, several times over with firms that in many cases were not even known to be entities that had access to the collateral.

To say the least, this was a very painful period for the asset management industry, and I guess that probably is an understatement. But what we did learn from that episode is that sufficient and focused efforts on good collateral management need to be risen and has made it to the top of the list, in terms of things that asset management companies and their boards should be concerned about. We're all focused on exceptional collateral management today, and that's the top priority.

The crisis also made clear that rehypothecation chains are actually very fragile. They're subject to runs, withdrawals, failures, as result of defaults and insolvencies, and they're very costly to unwind. The crisis also raised some interesting questions about the nature of the costs and benefits of collateral rehypothecation, as well as the efficacy with which unregulated transactions in this area are wise.

This morning, we will hear from two distinguished experts, who have given the subject of collateral and rehypothecation much thought and will share their views on these topics, as well as hopefully answer some of your questions.

Their detailed bios are included in the conference program, so I will be brief in my introductions. Our first speaker is Charles Khan. Charlie is professor emeritus of finance at the University of Illinois and a leading scholar working in the areas of payment systems and liquidity. In addition to publishing his research in leading economics and finance journals, Charlie regularly consults with central banks throughout the world on issues related to the design of the financial infrastructure, institutions, and markets. Today, Charlie will discuss the results of his latest research project, which uses a game theoretic model to examine the efficacy of rehypothecation.

The research was conducted jointly with Professor Hay-Jen Park, also of the University of Illinois. Our discussant is Phil Prince. Phil is managing director and head of treasury at Pine River Capital in New York. Phil is uniquely qualified to serve as our discussant, given his vast experience on the street managing central funding functions, including financing, cash management, liquidity, and risk management. He began his career at Arthur Anderson Consulting and before assuming his current position at Pine River, held several senior positions at Goldman Sachs, including cotreasurer of the Investment Management division.

So let's get started and I'll remind you to use your Pigeonhole applications to submit questions, and I'll do my best in managing them. I'm not a quick as Paul Kupiec, so you have to bear with me as we get to that point in the program. So Charlie, I'll now turn the floor over to you.

Charlie Khan: Thank you all very much too for inviting me to be a part of the program. This is a very exciting conference to be in. As Curt mentioned, this is joint work with Hay-Jen Park. Hay-Jen's not a professor yet. She's still a graduate student but she's going to be out next year, and look for her. She's going to be good.

So, rehypothecation is the repledging of collateral. For example, when a prime broker uses a client's collateral to back up his own trading and borrowing; he does that provided that the client's given him permission to do that. Rehypothecation and other forms of collateral reuse are ways of economizing on scare collateral. That collateral is scarce is pretty clear. Persistent liquidity premiums that exist on safe and liquid assets are well documented in the literature.

Without rehypothecation, lenders are going to have to just keep that collateral idle, until they return it to the borrower. Rehypothecation is going to allow lenders to raise additional funds. That's good news for both lender and borrower, in principle. The lender is able to decrease the opportunity cost of holding that collateral. Meanwhile, that allows the borrower to get funds at a cheaper rate, so that's good news for him as well. In short, rehypothecation is a technique for providing more funding liquidity in the economy. If we used Pete's kind of logic as liquidity as having a time dimension, then we can think of rehypothecation as a bit of a time machine. That is to say, it increases the velocity of the collateral in the economy and allows it to do more work in the same amount of time, if you will.

However, rehypothecation does have a downside and the downside comes from the counterparty risk. The receiver of that collateral may go bankrupt, having repledged that borrower's collateral to a third party. As we've seen, that's not just a theoretical possibility. The collapse of Lehman Brothers in 2008—or, less dramatically, the effects of MF Global in 2011—shows that this is something that we really do need to be concerned about. Rehypothecation failure leads to the misallocation of those assets. The collateral can't be easily returned to the borrower, who is likely the one putting the highest value on it in the first place.

So, rehypothecation had enormous effects or enormous consequences in the financial crisis. In 2007, there was about twice the amount of rehypothecatable collateral out there in the six largest U.S. investment banks as there was after the Lehman Brothers collapse. After the collapse, hedge funds stopped being as willing to allow their assets to be rehypothecated as they were beforehand. Details of that kind of information are available in some of the research by Manmohan Singh. You can find more details of the way that collateral has been changing over this period of time and some interesting work of his.

Anyway, because of these large changes, the debates about whether rehypothecation should be regulated or not does still continue. And so, as a result of that, it becomes rather useful to ask the questions, which are the main questions for us in this paper. Those questions are as follows: What are the benefits and costs that come from rehypothecation? Is rehypothecation simply an accounting tool with no economic consequences? Or does it have some social benefits or maybe even some social costs associated with it? If it does have some real economic consequences, then are the individual's preferences for rehypothecation versus nonrehypothecation aligned with social efficiency or not? From those questions, can we develop some idea as to when it's going to be that rehypothecation is more or less likely to occur?

So, in the model, we have three key assumptions. The first assumption that is needed to make the model work is that a borrower is subject to a moral hazard problem. Because there is a moral hazard problem, because we've got to induce some good behavior, it's necessary to post collateral, to ensure that borrowers make efforts to avoid default. The ability to post that collateral increases the borrower's credibility and his ability to raise funding for his productive investment.

The second key part of the model is the assumption that the collateral is actually transferred from the borrower to the lender at the time the contract begins. That is to say— as in a repurchase agreement, for example—borrower exchanges the asset for cash, and then effectively buys it back later on. However, financial distress can tie up that collateral.

The third key assumption for our story to work is that the collateral is more valuable to the initial owner than it is to the others in the model—that is, that it's less than perfectly liquid. Doesn't have to be very illiquid; the illiquidity can be quite small, but it needs to be there somewhere. It can come from things such as, for example, portfolio considerations—the initial borrower had it in his portfolio for a particular reason—or from a cost of resale, or from just the certainty of title—once the thing goes down the collateral chain, that title becomes a little more murky than it was before. Any of those will work to give us that thing.

What we do in the model is to start with a benchmark story. The benchmark story is one involving two parties. The two parties are basically the same as the model of Bolton and Oehmke in 2014, which is a model which explains how to understand the workings of collateral in an incentive problem.

In this model, there are two periods. There's risk neutrality; there's no discounting. I've got a firm, call it Firm A, that wants to borrow to finance an investment project, but the lenders to the firm are faced with a limited commitment from him to be able to repay the loans. If lenders can, after the fact, only attach a fraction of the cost of the investment, then there's going to be no willingness to lend. They won't be willing to give the money if they can't get it back.

That source of the inability to get the money back can come from, among other things, a problem of moral hazard. Suppose it's the case, for example, that the success of the investment depends upon actions taken by Firm A, which can't be observed by the lender. If that's the case, then the ability of creditors to extract value is going to be limited by the need to induce the effort from Firm A. If I make the terms too onerous, he's going to stop playing the game, he's going to stop working hard. In this case, borrowing that's backed only by a future gain from the project might not be a feasible thing to do.

What can you do in such a world? You can get collateral. If Firm A has another asset of value to him that he wants at a later date but not just yet, he can offer that asset in pledge for redemption at the repayment date. The greater the value of that asset in the future to Firm A, the greater the amount he can credibly borrow by pawning that asset, even if that asset, for example, had little value to the lender.

In general, the terms that can be arranged between A and B—how much collateral, how much borrowing—are going to depend upon both, the amount that you can pull out of the project and the details of the incentive problem that Mr. A is facing in the story. So depending on both of those details, you can get one of two possibilities in the theoretical model. You can get a loan that's overcollateralized or undercollateralized—that is, one in which the value of the collateral is greater than or less than the promised repayment. Haircuts can go either way in the theoretical story.

Even if that lender is less than perfectly safe, collateral could still be a useful thing to have. Provided the possibility of the lender's failure is not too great, then it could still be desirable to engage in collateralized borrowing. The lender will simply compensate for the risk of his failure by improving the terms at which the borrower is going to borrow.

So that's the two-person story. Our model is going to be one in which we take this two-person story and turn it into a chain of collateralized lendings. So, we've got three periods in our world and we've got three players. We've got an initial borrower, we've got the lender to that initial borrower (call him B) and we've got someone who is going to lend to the lender (we'll call him C). At date zero, A borrows funds from B for his investment. He does so by pledging some asset as collateral. At date one, B needs to borrow funds from C for his investment, and he does it by repledging A's asset. Then at date two, both of A's and B's investments mature. B recovers A's collateral from C by making the payment and then B returns the collateral to A in return for receiving his payment. So, the moving of the collateral down the line, it moves back down the line to the end of the story.

On the other hand, if B defaults, then the collateral is going to be seized by C. Collateral remains in the wrong hands because it's really worth less to C than it is to A. That's the model. We work out in this model what the optimal contracts are going to be. Basically, we're going to have a sequence of two contracts. We're going to have a contract between A and B at time zero, setting up the initial arrangement and we're going to have a contract between B and C at time one for a subsequent loan. We're going to solve these things by a backward induction and we're going to compare what happens in that world, assuming that the relending of the collateral is allowed, and compare that with a world in which that's forbidden. A world in which A says ahead of time, "No, you're not going to be able to pass the collateral on to someone else, it's got to stay with you."

So we look first at the contract that's there between B and C. That contract is pretty straightforward. If B is going to default, C is going to retain the collateral. If B doesn't default, then the maximum that he is going to be able to pay C is going to be basically the amount that he receives from A. B is going to borrow the maximum amount consistent with these constraints.

So the ability to make a new loan with that collateral gives a shadow value of that collateral to B himself. Even if B didn't value the collateral himself, he values it for his ability to get back his money from A at a later date. So that turns the collateral into something valuable for B as well, and that turns it into a new problem, in which we can, taking that collateral value into account, figure out what the best deal is between A and B and what kinds of arrangements they would want to make between the two of them.

That's the model that's solved and a comparison is made with what happens if we wouldn't allow the second level of the arrangement to go through.

In that story, there's a welfare tradeoff. On the one side, rehypothecation supplies more funding, liquidity to the economy, so that additional productive investments can take place, which otherwise wouldn't take place. And on the other hand, rehypothecation failure may incur costs by misallocating the assets. So far, though, that sounds like a perfectly reasonable cost-benefit analysis, one that all the parties should be engaging in and there shouldn't be any conflict at all between decisions about whether to rehypothecate or not, and what the values underlying the assets and the economy may be.

Where the possible inefficiencies arise come from the wedge that's in the story, between the value of the collateral to the borrower and the value of the repayment. Remember that that ratio (is it the haircut to the direction of the haircut), is determined by the incentive problem between A and B. That means that the shadow value of the collateral to the middle man is not the same as the value of the collateral to the borrower. That means that the middle man may not make the right decisions, efficient decisions about whether to rehypothecate. The terms at which he is willing to rehypothecate are going to different than the terms at that A would like him to use in the first place. As a result of that, there's going to be inefficiencies in that rehypothecation decision.

If the loan is undercollateralized, there's going to be an insufficient use of efficient rehypothecation. That is to say, B values the payments she receives from A, more than A values getting his collateral back, and so B is going to prefer not to rehypothecate when he should. In the more relevant case possibly, if there's overcollateralization in this story, then it's going to the case that B is going to be too eager to rehypothecate, even in times when that rehypothecation would be a questionable decision from A's point of view.

So that tells us how the incentives work from the point of view of Mr. B deciding whether to engage in rehypothecation. What does that say about Mr. A's decisions about permission to give the right to rehypothecate or not? If A has the right to decide whether rehypothecation should occur, A is going to be reluctant to permit rehypothecation, when the optimal contract between A and B involves increasingly overcollateralized lending. The more valuable the stuff is back to him compared with what he has to pay to get it, the more he's going to be afraid of casual decisions made by Mr. B.

In extensions of this story, Hay-Jen is working through what happens as variability over time in more risky environments—what happens to those decisions as the possibility that the overcollateralization may increase?

So, in summary, what we have here is an analysis of the economics underlying rehypothecation. The model that I put together for you highlights the tradeoffs between the costs and the benefits of rehypothecation—that is to say, supplying more funding liquidity to the economy but incurring the deadweight cost of misallocating the asset and the spread between the collateral value and the promised repayment, leads to the incentive conflicts between the parties and the possibility for inefficient decisions to be made in the story. Thanks.

Phil Prince: First of all, thank you very much for inviting me here to respond to Professor Khan's paper. As a practitioner crashing an academic conference, I kind of assume that my role here is to talk about the assumptions in the model, talk about whether they accord or conflict with my own experience in the markets and suggest the directions that those conflicts might push the results, and of course to call for more research. Before I get to that though, I'd like to level-set a bit by giving a thematic example of the kinds of collateral use and reuse that we're discussing here. I should pause for a second to say I'll use the term reuse, rather than rehypothecation. Rehypothecation has a very specific legal meaning and the model and the discussion in the paper are much broader than the case of rehypothecation. It applies to many other ways in which collateral is also passed and collateral is passed on. So, I'm going to stick with reuse, rather than rehypothecation.

First, I'm going to promise to not sell you anything today. I'm going with my early training here, so we'll draw a boxes-and-arrows chart. Professor Khan starts with two parties: A and B. We'll call them levered investor, or hedge fund for short, and global systemically important financial institution, or bank for short. In the model, the hedge fund starts with an asset, so let's add in a corporate issuer who wants to sell a bond to build a factory, a nice appropriate social use. The hedge fund buys the bond and borrows from the banks to fund it, putting up some of the hedge fund capital, some of the nav of the fund as a haircut. Of course, banks are by their nature levered entities, so the bank needs to borrow from a cash provider. That might be a money market fund, a corporate treasurer, a sovereign wealth fund, or any other short-term money market investor. In Professor Khan's model of course, that's party C. In the real world, we have an issue here. Those investors would prefer to be secured by high-quality liquid assets, even when dealing with a giant bank. In fact, they likely have certain investment guidelines that restrict them to taking high-quality liquid assets to secure themselves.

What we're talking about here is corporate bond collateral, which could be a high-yield issue let's say. So we're going to introduce one more player, a lender of U.S. treasuries. This might be a life insurance company or a pension fund. It's an entity that needs long credit-free duration but is happy to pick up a lending fee for the term of a short-term trade. It lends U.S. treasuries against a pledge of the corporate bonds to the bank, and the bank repos the U.S. treasuries to the cash provider. So this is the slightly more realistic picture of the transactions in the model. We've added one player, but we still have the bonds going from A to B and, with a little transformation, to C and cash going the other direction, right?

Now, one of the important underlying assumptions in the paper is that the collateral is more valuable to hedge fund A than to the rest of the market. This picture really doesn't do anything to help us understand why that might be. This picture is just credit and maturity intermediation taking place in the shadow banking system. This is money market funding of cap and markets lending. That's a thing that happens, yes, but that's not really what most hedge funds are about.

We're relative value traders, right? Our job is to find relationships in the market that are dislocated or mispriced and by trading against them, bring them back into line. To understand why this is important in the context of the model, let's turn this intermediation trade we've got up here into a capital structured trade, a relative value trade, by selling the underlying equity in the market. Now we'll assume for a moment that the bank is also the hedge fund's prime broker. So the bank will provide the securities to make delivery to the equity market, which the bank will in turn get from an equity securities lender. The equity lender now has extra cash to invest, of course, so he'll send that over to the money market mutual fund. So now we have a picture where it's a little clearer why the collateral is more valuable to the owner, the hedge fund, than it is to other market participants. It's not just a bond, it's a building block to a portfolio—portfolio effects, as Professor Khan said. The extent of the excess value is the relative value dislocation, less the financing costs of carrying the trade until that dislocation is realized.

Another of the important assumptions in the model is that A has an outside project to invest in that offers some degree of flexibility of cash flows and the collateral is used to make up the difference—either as blackmail, in the sense that A wants it back enough to do what he has to do to make the outside project work, or it's an insurance function and the bank can sell it if A fails.

But in a secured financing transaction as we see from the example, A's project tends to be the portfolio itself. So there are no cash flows outside of the collateral to pledge. In practice, that means the undercollateralization case, at least at inception, isn't really likely. The analysis and modeling of that case is still worthwhile of course because the trade may become undercollateralized between time zero and time two. The portfolio could effectively drop in value, and we should think about the incentives in that case. Not every hedge fund is right every day.

Another interesting set-up in the model is that B, the bank, lends to the hedge fund, A, at time zero and at time one, borrows from C for another investment. One of the important variables in the resulting equations is the probability that B's other investment will succeed. As you might expect, social efficiency tends to be higher when it does succeed. But if you look at this picture with that in mind, it provides an interesting insight into some of the regulatory structures around this business, in the U.S. in particular. Much of the U.S. regulatory structure, including the U.S. definitions of customer and counterparty, which are important in that prime brokerage relationship, are intended to cleave this business here of borrowing and lending money against securities A to C away from B's other projects of making markets or of principle investing.

Here in this picture, B's project is portfolio lending itself, much as A's project is the portfolio. As Professor Khan's calculations show, with portfolio lending as the only project, reuse of collateral is both required and much safer. That give some insight into the way U.S. regulations are structured to cleave this securities borrowing lending from other businesses of the bank holding company or the broker dealer. B's other project, by the way, introduces the final welfare equations in two ways—one is that probability of success and the other is, of course, its expected return.

As your paper summarizes algebraically, social efficiency under reuse consists of three components. There's the surplus generated from A's investment. Here in our picture, that's his capture of the market pricing dislocation. The second is the surplus generated from B's investment when it succeeds, which here in our picture is simply the return to the intermediation process. And three is the cost generated by the misallocation of resources whenever C, the cash provider, is stuck with the asset, when B and his investment fail. Collateral use is socially efficient if the expected benefit, number 2, is greater than the expected cost, number 3. So let's focus on that for a minute and unpack it just a bit.

Regulation as currently conceived is geared towards increasing this efficiency by lowering the probability that B's project fails, which would create the misallocation of collateral and all the chaos. But the model here is a single play, what is in reality a recurring set of plays. We do this every day. One of the first things I learned when I started out on the repo desk in 1983 was, we all come back again tomorrow, right? It's a recurring set of plays, many repeated trials. Even if the probability of failure is very, very low, the probability of no failure after many, many repeated plays starts to become vanishingly small.

So, where in this model here can we focus on reducing the cost of misallocation, in case of failure? The driving assumption in that piece of the model, for those deadweight losses, is the trading frictions between C, the cash provider, and A, the collateral provider. If we turn to the picture, we can see that those trading frictions, the fire sale costs as we'll call them, multiply because we've exploded the number of players involved. Instead of A, B and C, we've got A, B, C, you've got D, E, F, we're all over the place.

What's needed, though, is very simply a stock record. If the levered investor, the U.S. treasury securities lender, the cash provider, and the equity securities lender could all be introduced, with a balance sheet to intermediate, they would all re-establish their positions basically in a heartbeat. Now, this is one of the big selling points of a clearinghouse. It provides that industry-wide stock record by being the counterparty to every trade. This is also, conversely, one of the big selling points of the distributed ledger chain, a block chain, in finance these days because it provides everyone with access to the stock record of the transactions involved in this collateral framework.

The model puts a framework around those ideas and extending it to cover those costs—and think about the difference between those costs—should be pretty straightforward, I would think. I'd also like to see the research extended to include longer collateral chains. In our simple picture here, we have one bank between each of the lender-borrower pairs. How would the calculations change if the intermediation chain goes through two banks with separate customers? So, a customer comes in, bank–bank–customer goes out, or, customer comes in–bank–clearinghouse–bank–customer goes out. In default, would it matter if the failing bank's customers could only be introduced with other customers of that bank and not to the second bank's customers? In a longer chain, would the results differ, depending on which end of the chain fails? If we have customer–bank–clearinghouse–bank–customer, is it bank 1 that matters the most, is it "bank 2" that matters the most?

The paper's an interesting start to that question but we do need to extend the research in those ways.

The other interesting extension I would think of is the question of B borrowing more from C than he has loaned to A, which, of course, is a common occurrence in the market. How does that fit into the model? Where do we generate the ability for B to do that, particularly in a chain where we are really talking about cleaving the collateral, the money lending business, the money-dealing business from the risk business of the bank? That's my response. Thank you very much.

Hunter: Well, thank you, Phil. We've got several questions here, and I think you touched on one. I'm going to start with the most obvious one, the one that has the most popular votes. OK, how reasonable is it to assume that the original borrower puts a higher value to the collateral than down the chain, and what would happen in the model if you took that assumption away? If you think of asset managers, most of the securities are very liquid. They're treasuries, they're triple-A rated, but in some case they're not. How would the model change if you took that assumption away? Would it fail, or collapse?

Khan: In the straight story we've got here, things become less and less important as that's the case. To the extent that you're going to have even some possibilities, then I'm going to go back to the defenses. Yes, this is small. As those costs go down, then the differences between the ability to get it from the market make the problem go away.

Prince: I think that's particularly important when you're talking about, under IZO with pledge and rehypothecation of cash collateral, what is the deadweight loss? Essentially, everyone is valuing it in the same way. If C ends up with it instead of A, well, it's just a client, who cares? There is a portfolio effect in treasuries. Several times this morning already, and last night as well, we talked about the (unintelligible) on trade. The bounds of those arbitrage opportunities are widening as a balance sheet becomes more difficult to obtain. So that actually increases the potential deadweight loss of just a straight treasury portfolio.

Hunter: OK, I'm going to consider that one answered, and I will move on to the one here. Was collateral too often reused in raising leverage in the economy, including synthetically? Did leverage drive the chain?

Prince: You refer to Manmohan Singh's work on the amount of reuse in the economy. I'll go back to Pozar and Singh from 2010 or 2011, and make the claim that one of the problems with the paper is that it deals with the issue really from the point of view of A and B, when in fact what drives much of the collateralization and reuse is the demand of C, the money provider, for on-demand deposit alternative, which then drives the need to increase leverage in the financial system in order to provide those money alternatives.

Khan: The model can be modified to handle that by flipping who it is that's actually doing that investment at the middle part of the story. So it's probably worthwhile going back there and making the change, to make that more apparent as well.

Hunter: So, Charlie, does the model say anything about the optimal length of the credit chain?

Khan: Like, can we get the velocity in the story? We could, at the moment, there's only three guys and they're not even the six guys that you have in there, but yeah. I agree that a primary thing to do is to go and see how far we can push that.

Prince: We know that velocity was much higher before the crisis. We know that velocity is much lower now. What we don't know, and we know that there are things that, in fact, regulators are doing to lower velocity, not just in terms of regulation, but also in things like the reverse RP and opening up accounts for the systemically important clearinghouses, but we don't know what the optimal level of velocity is.

Khan: It's going to be dependent so much on information along the line in that chain. That is to say, how much does A know about B, versus C, versus D, versus E? There's not going to be a clean answer independent of that question, unfortunately.

Hunter: Another question is, how should rehypothecation be regulated, since you claim in your model of one of the outcomes is that you could have overuse of the facilities or you could have under-use. So, how would you regulate it? What, in your model, would be a reasonable thing to do, to limit the risk that you outline?

Khan: As it stands in the story, there isn't any room for a regulator to improve on the decisions that Mr. A makes in the world about whether to permit or refuse to allow rehypothecation. If you add into the structure the normal kinds of systemic risk problems that would arise in this story, then you're going to get a bias in favor of limiting of rehypothecation. It's not coming out of the rehypothecation itself, it's coming out of the standard kinds of systemic risk.

Prince: And Curt, one of the points you made in your introduction was that in 2008, many market participants "discovered" that their collateral at Lehman Brothers had been rehypothecated into the U. K., from where it was rehypothecated into English law. The fact that that was a discovery for many market participants is a large part of the problem. That the issue of rehypothecation was apparently not so nearly well understood as many in the industry thought it was and that many participants who were playing the part of A had made those decisions about whether to permit or not to permit, without fully understanding B's relationship with C. That is, I think, a problem. But that's solved by the scars we all feel from 2008.

Hunter: One of the questions that came in early had to do with the importance of collateral versus creditworthiness and how do you look at that, as a practitioner? Is it more important to have the right amount of collateral or is the creditworthiness of the counterparty a more important thing to look at.

Khan: In the model, it's a tradeoff between the two. In the model, if the borrower is absolutely safe, you can always get the money back from him, you don't need collateral in the first place. The collateral is there as a substitute or a partial substitute or threat to keep the borrower trustworthy. So, if I can look into your eyes and know you're truly reliable, then I don't need your collateral.

Prince: And I think partly that's because this is a single-play model. You have, again, a series of repeated plays of the same process. The probability that there will be no failure over an infinite time horizon approaches zero, because if there is any probability of failure, it begins to multiply.

Khan: Which is interesting because most theory models say, "Oh, well, repeated makes it less of a problem."

Prince: Well, repeated solves some of the moral hazard problem because you want to come back tomorrow, but the underlying risks begin to come back to you, which explains why C, the cash provider, is looking for a secured par on-demand deposit alternative, rather than just a deposit at the bank. He is not going to be insured. There is some failure probability of the bank; he would prefer a secured collateralized deposit, which means that collateral will then be required all the way back up the chain to generate that in the first place.

Now, as the collateral provider, it's important to recognize that you have credit risk to your cash provider, so it is important to do the credit risk analysis on that credit provider. Your credit risk is not to the full notional, it's to your haircut, plus Professor Kyle's potential for change over the period in which you can recover your position, which means of course you have a larger credit exposure when you're short something than when you're long something.

Hunter: Given the differences—and this comes right out of the Lehman experience—given the differences in legal restrictions in different jurisdictions, what might be done to prevent migrations of these securities to markets where they're not so regulated and you have the problem that we had in 2008? I can answer that from the perspective of one U.S. mutual fund company. We restrict the ability of the prime broker to transfer these securities to any subsidiaries. We have visibility directly into the prime broker's rehypothecation program. We get daily reports on all positions, our securities that are pledged. We also have a custodian that actually monitors with the risk management team at the fund on a daily basis the transactions in and out of our collateral pools, as well as at the prime broker. There are things that are being done as a result of the experience in 2008 to prevent this from happening again. Now that doesn't say it couldn't happen again, but what's your prospective on that particular question?

Khan: But rationalizing the differences between the jurisdictions really seems like a crime.

Prince: That would mean rationalizing U.S. and U.K. law and that's going to be a real nightmare.

Khan: But the consequences of not doing that are very expensive.

Prince: Yes, but you have to go in with your eyes open, is basically the answer. You need to understand the difference between U.S. law and U. K. law, U.S. regulation, particularly in the prime brokerage market as opposed to the repo market, and U. K. regulation, and make a reasoned trade-off between the increased funding access of allowing full rehypothecation to the cheapest of all possible places versus restricting that. If you allow rehypothecation, basically ad infinitum, into the cheapest possible market, you better get value for that for your investors and do so with your eyes open to the risks and monitor the risks with daily reportings, so that you know whether you have a money claim or collateral claim and so on and so forth. But there's nothing wrong with that if that's an important part of the portfolio construction process. Just don't go in blindly. It's a risk. You take risk, you get return, if you do it right.

Khan: The worry becomes if you think that there is a temptation for one jurisdiction to start, in response to this, changing its standards.

Prince: Are you worried more about one jurisdiction easing its standards to gather more collateral or are you worried about the other jurisdiction tightening its standards? The problem there of course is the social efficiency may go down because funding availability goes down.

Hunter: Here's another question for both of you. We saw data that you provided, Charlie, that rehypothecation really has been impacted by the crisis—having 4.7 trillion, down to two-plus trillion. So the question would be, why hasn't it come back in terms of the volume of use of rehypothecation?

Khan: Pure theory would argue that interest rates down have just made things less expensive at the basic level. The real test is going to be when interest rates go back up. Is this going to be a shortage? That's a pure theorist talking.

Prince: What has increased is other ways of satisfying the demand for money by that pool of wealth. You've got the reverse RP from the Fed. You've got...all of the central bank's balance sheets have ballooned, such that there are by nature more deposits in the system. You can't get rid of deposits to go to a nondeposit alternative. Is it sustainable? Depends what you mean by sustainable. It may, in fact, not be socially efficient. As we said before, there was higher velocity, now there's lower velocity, but we don't know what the optimal velocity is. Velocity may be lower than optimal now, where it was likely to be higher than optimal before. Lower than optimal is sustainable for longer because all that happens is that everything just gets worse, worse, and worse and our children are poorer. But they can be sustained and poorer. Higher than is optimal may not be sustainable; it could lead to a financial crisis.

Why hasn't reuse increased? Well, you can't get the balance sheet for it. If you look at the picture that I put up there, all those risk transfers can take place effectively, with very little balance sheet. But that money transfer from C to B, out and back to one lender, out and back to the other lender. Each of those is on balance sheet. So, given the leverage ratio, there is a limit to how much that can shrink and grow. So, to the extent that you need to increase risk transfer, you have to increase that money-dealing balance sheet. Reuse hasn't increased because that balance sheet can't be flexible anymore, so instead you see prices change more rapidly. The portfolio effects become greater, the arbitrage bounds on treasures get wider, CDS versus CDX gets wider. All the kinds of arbitrage bounds just widen up and get undertaken under with smaller position size on wider bounds instead of bigger position size on smaller bounds because the balance sheet is not available and reuse is not possible.

Hunter: Here's a question. Liquidity: help to the economy? Given low real rates indicating a surplus of capital, why is rehypothecation necessary, except to fund the trading of prime brokers? Does it really have real value to the economy or is it just a way for prime brokers to carry out their business? We know that funds are benefited by lower costs and we share in the revenue, under rebates from securities lending, so there is a benefit to the funds that you can directly measure. What about this whole question about is it really a way for prime brokers just to fund their business without using their own capital?

Prince: I would argue that without the funding liquidity created by the rehypothecation chains, or the reuse chains, that markets become less efficient—all the things we've talked about already, with on run, off run, which applies to corporate as well and applies to everywhere else as well—that trading becomes more expensive. Investors will, over time, require a higher yield in their investments and that central banks will be forced to keep lower rates for longer in order to try to keep the same quality of growth over time, the same rate of growth over time. With rehypothecation, long reuse chains, and funding liquidity, we can have the same level of an investment without the risk of bumping up against the zero lower bound, which may not be a lower bound anymore, as often.

Hunter: There's a follow up to that. How should the lender of last resort consider rehypothecation and what does it mean for the effective conduct of monetary policy? Some claim that these chains allow unregulated entities to create money. As you show in your model, you can create money, you can have the reverse happen. So what should the central bank do? How should we think about that, for the central bankers in the audience?

Prince: The Bundesbank actually just came out with a paper last week dealing with that very topic. They showed that most of the actual credit creation for the boom had taken place on the balance sheets of banks, but then in turn had been financed through the shadow banking system. So, you think of it as creating the money, the money escapes out into the economy, and then through the shadow banking system, everything comes out of the banks and into the capital markets financing and money market funding. The question here, would the lender of last resort consider rehypothecation? That's a very complicated question because it gets into some tough legal areas, but reuse with clear title transfer, I don't see why it's a problem for lender of last resort.

Khan: On the money supply side, it becomes one only in the sense that it becomes harder to figure out what's going on. That's not different in this respect than it is for any innovations in payments systems or anything else.

Prince: Right. We choose to say that an industry-wide stock record would have great value to the regulators.

Hunter: How do you price rehypothecation as an option and how should that be taken into account? When I got your paper Charlie, I talked to our portfolio managers and asked them how deep into the transaction do they look and do they consider some of the things that you outline in the paper. For the most part, the answer is yes, we do consider that and we try to work it not into the pricing, which usually comes from the prime broker, but we work it into our restrictions on what they allow and what they prevent.

Khan: In the story, the more variable the situation is going to be in that small period of time that it's in the power of Mr. B, that option is a particularly valuable thing for Mr. B and a cost for Mr. A. An option pricing model would be a kind of cool thing to try.

Prince: Right, and for a fund that's Reg X–eligible, there is a price. If you assume a fund that is using you as prime broker, a relative value fund, particularly equity heavy, so that most of his business is going through a prime broker, there is a different price for portfolio margining than there is for arranged financing. That price is the price of allowing rehypothecation out of the U.S. protections 15c3 into the U.K. rules. You have to evaluate that price versus the risk of it.

Hunter: OK, how about the FSB proposal for haircuts? What do you think about the proposal? If you have any opinions, let them be known.

Prince: It's basically minimum haircut proposals on securities financing transactions. As we've said a couple of times, we don't know what the optimal velocity is. These proposals have two effects. One is to constrain that velocity to some extent and the other is to ensure that in that collateral chain, there is enough capital being posted somewhere. If you think about all the stuff, if all that business had been done within a single bank deal or on the bank's balance sheet—say, here is the capital that that bank needs to hold against that business—that business of buying a corporate bond and selling the equity and doing all those transactions on its own. We split that off; there are now six people involved, but there's still an amount of capital involved.

You have a different capital for the risk-dealing part, the guy who's holding the portfolio, and the money-dealing part, the bank passing money up and down the chain. What's the right amount of capital to be holding in the money-dealing chain? What's the right amount of capital to be holding in the risk-dealing chain? I don't know, but the FSB thinks they know about what the right amount of capital is to hold in the money-dealing chain, and so they're going to enforce it. They're going to set a number and say this is the appropriate number for this kind of collateral and money transfer.

From the first session, there are some reasonable ways to sort of think about what are those right numbers? Where you can gather enough market data about an assumed market and variance and so forth, and come up with some probably pretty good numbers for what is the right amount of capital to be held in that chain in the form of haircuts.

The risk of course as pointed out, sudden regime shifts in volatility or in the cost of bearing a unit of trading risk, of market-making risk.

Hunter: The question is, with mandatory margining coming to the bilateral, over-the-counter derivatives markets and rehypothecation being prohibited by regulators, what impact do you think this will have on the OTC derivatives market? What are the connections between the OTC derivatives markets, regulation on rehypothecation, whether it be internally driven by those that are borrowing, with restrictions on the prime broker and regulatory restrictions. What impact would that have on the OTC derivatives market? That's the question.

Prince: It raises the price. As shown in the model, the price to A of borrowing is higher if B cannot rehypothecate the collateral. That's the case if you have segregated initial margin in the OTC space. An OTC derivative has two legs. It's got a risk leg and a funding leg. That funding leg is more expensive if the initial margin is segregated. That's fairly straightforward. So, OTC derivatives become a less efficient way of passing risk, relative to underlying securities. And OTC bilateral becomes less efficient, relative to CCP, where you're at least economized once you've transferred out as a dealer.

Hunter: OK, one final question. Would you consider a dynamic model that could incorporate the liquidity pricing factors from session one today? If you could, Could this model address the risk and the value of the collateral changing over the loan duration?

Khan: That actually makes some of the issues actually more relevant as well, because then your questions about how this is such a small margin we're talking about here, that margin becomes much larger, much more important as the dynamics come in. Particularly, if you can't adjust as quickly on the terms of your arrangement with Mr. A.

Prince: You need to tie Professor Kyle's model and the FSB thinking on SFT together. When you think about social efficiency of these collateral chains.

Hunter: OK, one last comment from you, Charlie. When I was reading your paper, I noticed that it seems to be a sequential model from A to B to C. As Phil's diagram shows, it's really a more complicated scenario, where you have simultaneous transactions taking place. Is it possible to model that kind of market?

Khan: Probably the best way of doing it is to have repetition of the story. Then you can finesse whether it's dynamic or simultaneously going on.

Hunter: I would note that in the paper, the model is generalized to simultaneous transactions and a couple other wrinkles that most people would probably say this is a step forward, so I encourage you to keep working in that area. I think this brings us to the end of our session. I'd like to thank both Charlie and Phil for their insightful comments and the audience for submitting some interesting questions. I'm sorry we didn't get to all of them but certainly during the rest of the conference, you'll be able to ask our presenters the questions and I'm sure they'll be able to answer them. Join me in thanking them for a nice session.