Collateral and Monetary Policy: A Puzzle

A stylized fact about post-crisis economies is that asset markets have become segmented with “safe assets” trading differently from assets more generally. I have argued elsewhere that the collateralization of financial sector liabilities has played an important role in this segmentation of markets.

I believe that this creates a puzzle for the implementation of monetary policy that provides at least a partial explanation for why we are stuck at the zero lower bound. Consider the consequences of an increase in the policy rate by 25 bps. This has the effect of lowering the price of ultra-short-term Treasury debt, and particularly when combined with a general policy of raising the policy rate over a period of months or years this policy should have the effect of lowering the price of longer term Treasuries as well (due to the fact that long-term yields can be arbitraged by rolling over short-term debt).

A decline in the price of long-term Treasuries will have the effect of reducing the dollar value of the stock of outstanding Treasuries (as long as the Treasury does not have a policy of responding to the price effects of monetary policy by issuing more Treasuries). But now consider what happens in the –segmented — market for Treasury debt. Assuming that demand for Treasuries is downward sloping, then the fact that contractionary monetary policy tends to shrink the stock of Treasuries itself puts upward pressure on the price of Treasuries that, particularly when demand for Treasuries is inelastic, will tend to offset and may even entirely counteract the tendency for the yield on long-term Treasuries to rise. (Presumably in a world where markets aren’t segmented demand for Treasuries is fairly elastic and shifts into other financial assets quash this effect.)

In short, a world where safe assets trade in segmented markets may be one where implementing monetary policy using the interest rate as a policy tool is particularly difficult. Can short-term and long-term safe assets become segmented markets as well? Given arbitrage, it’s hard to imagine how this is possible.

These thoughts are, of course, motivated by the behavior of Treasury yields following the Federal Reserves 25 bp rate hike in December 2015.fredgraph


What Gorton and Holmstrom get right and get wrong

Mark Thoma directs us to David Warsh on Gorton and Holmstrom’s view of the role of banking. I’ve written about this view in several places. My own view of banking is very different and here is a quick summary of my key points.

The source of Gorton and Holmstrom’s errors: Taking U.S. banking history as a model

In my view Gorton and Holmstrom err by basing their view of what banking is on the pre-Fed U.S banking system. Nobody argues that the U.S. represented a “state-of-the-art” banking system in the late 19th century. In fact, in the late 19th century the U.S. banking system was still recovering from the reputational consequences of the combination of state and bank defaults in the 1840s that had led many Europeans to conclude that American institutions facilitated fraud. By the end of the 19th century, however, the U.S. did have access to European markets and there is evidence that the U.S. banking system relied heavily on the much more advanced European banking system for liquidity (e.g. the flow of European capital during seasonal fluctuations). Indeed, the crisis of 1907, during which the none-too-respected U.S. banking system was at least partially cut off from the London money market, was so severe, it led to the decision to emulate European banking by establishing the Federal Reserve.

What Gorton and Holmstrom get right: the fundamental difference between money market and capital market liabilities, or as Warsh puts it: “Two fundamentally different financial systems [are] at work in the world”

In particular, it is essential for the debt that circulates on the money market to be price stable or “safe.” This distinguishes money markets are from capital markets, where price discovery is essential. Holmstrom writes:

Among economists, the mistake is to apply to money markets the lessons and logic of stock markets. … Stock markets are … aimed at sharing and allocating aggregate risk … [and this] requires a market that is good at price discovery. … [By contrast,] The purpose of money markets is to provide liquidity for individuals and firms. The cheapest way to do so is by … obviat[ing] the need for price discovery.

What Gorton and Holmstrom get wrong:

1.  The historical mechanisms by which the banking system created “safe” money market assets.

Holmstrom writes: “Opacity is a natural feature of money markets and can in some instances enhance liquidity.” This is the basic thesis of Gorton and Holmstrom’s work.

A study of the early 20th century London money market indicates, however that the best way to create safe money market assets is to (i) offset the implications of “opacity” by aligning incentives: any bank originating or selling a money market asset is liable for its full value, and (ii) establish a central bank that (a) has the capacity to expand liquidity and thereby prevent a crisis of confidence from causing a shift to a “bad” equilibrium, and (b) controls the assets that are traded on the money market by (1) establishing a policy of providing central bank liquidity only against assets guaranteed by at least two banks, and (2) withdrawing support from assets guaranteed by low-quality originators. (ii)(b) plays a crucial role in making the money market safe: no bank can discount its own paper at the central bank, so it has to hold the paper of other banks; at the same time, no bank wants to hold paper that the central bank will reject. Thus, the London money market was designed to ensure that the banks police each other — and there is no American-style problem of competition causing the origination practices of banks to deteriorate.

The Gorton-Holmstrom approach is based on the historical U.S. banking system and sometimes assumes that deterioration of origination quality is inevitable — it is this deterioration that is “fixed” by financial crises, which have the effect of publicizing information and thereby resetting the financial system. In short, by showing us how a banking system can function in the presence of both opacity and misaligned incentives, Gorton and Holmstrom show us how a low-quality banking system, like that in the late 19th century U.S. which could only create opaque (not safe) assets, can be better than no banking system.

Surely, however, what we want to understand is how to have a high-quality banking system. The kind of system represented by the London market is ruled out by assumption in the Gorton-Holmstrom framework which focuses on collateralized rather than unsecured debt. An alternative model for high-quality banking may be given by the 1930s reforms in the U.S. which improved the origination practices of U.S. banks and — temporarily at least — stopped the continuous lurching of the U.S. banking system from one crisis to another that is implied by opaque (rather than safe) money market assets.

2. Gorton and Holmstrom err by focusing on collateral rather than on overlapping guarantees.

Holmstrom writes: “Trading in debt that is sufficiently over-collateralised is a cheap way to avoid
adverse selection.” His error, however is to use both language and a model that emphasize collateral in the literal sense. The best form of “over-collateralization” for a $10,000 privately-issued bill is to add to the borrower’s liability the personal guarantee of Jamie Dimon — or even better both Jamie Dimon and Warren Buffett. This is the principle on which the London money market was built (and because both extended liability for bank shares and management ownership of shares was the norm until the 1950s in Britain, personal liability played a non-negligible role in the way the banking system worked). This is rather obviously an excellent mechanism for ensuring that money market debt is “safe.”

The fact that it may seem outlandish in 21st century America to require that a bank manager have some of his/her personal wealth at stake whenever a money market asset is originated, is really just evidence of the degree to which origination practices have deteriorated in the U.S.

Note also that there is no reason to believe that the high-quality money market I am describing will result in restricted credit. Nothing prevents banks from making the same loans they do now; the only issue is whether the loans are suitable for trade on the money market. Given that our current money market is very heavily reliant on government (including agency) assets and that these would continue to be suitable money market assets, there is little reason to believe that the high-quality money market I am describing will offer less liquidity that our current money market. On the other hand, it will offer less liquidity than, say, the 2006 money market — but I would argue that this characteristic is a plus, not a minus.

3. Holmstrom errs by focusing on debt vs. equity, rather than money markets vs. capital markets

Holmstrom claims that: “Equity is information-sensitive while debt is not.” He clearly was not holding GM bonds in the first decade of the current century. A more sensible statement (which is also consistent with the general theme of his essay) is that capital market assets including both equity and long-term debt are information sensitive, whereas it is desirable for money market assets not to be informationally sensitive.


In short, I argue that in a well-structured banking system money market assets are informationally insensitive because they are safe. For institutionally-challenged countries, a second-best banking system may well be that presented by Gorton and Holmstrom, where money markets assets are “safe” — at least temporarily — because they are informationally insensitive.

In my view, however, we should establish that a first-best banking system is unattainable, before settling on the second-best solution proposed by Gorton and Holmstrom.

The Shadow Banking System is an Unstable Funding System for Banks, Not Assets

There are many definitions of shadow banking. A New York Federal Reserve Bank monograph effectively equates shadow banking to securitization, or the process by which individual loans are packaged into bundles, used to issue a wide variety of collateralized assets, and sold to investors. The New York Fed monograph is often used to demonstrate how complicated and virtually incomprehensible the shadow banking system is – it includes a “map” of the shadow banking system that, for legibility, the authors recommend printing as a 36” by 48” poster.[1]

More commonly, however, the term shadow banking refers to the use of money market instruments to provide short-term finance to long-term assets,[2] and thus focuses attention on bank runs and on the fact that shadow banks can face such runs, just as traditional banks do. For this reason securitization should not be equated with shadow banking, because a significant portion of private sector securitized assets were financed on a long-term rather than on a short-term basis.[3] This post will limit its focus – as does most of the literature on shadow banking – to the role played by money markets in longer-term finance.

This post finds that our current money markets play only a very small role in the direct finance of private sector long-term assets and for the most part are used as a financing system for investment banks. In short, the “market-based” credit system that some equate with the shadow banking system,[4] is very small – and relies heavily on commercial bank guarantees. To the degree that a substantial shadow banking system continues to exist, it does not fund long-term assets directly, but instead provides wholesale funding for investment banks, and to a lesser degree commercial banks.

To be clear, the focus here is on finance of private sector banks and assets. Thus, although Fannie Mae and Freddie Mac played a very important historical role in the development of the shadow banking system, by pioneering the practice of financing long-term mortgage debt on money markets through the issue and roll over of short-term debt that was at least nominally a private-sector obligation,[5] they now officially have government support, and, for the purposes of this paper their debt is treated not as part of the shadow banking system, but as a government obligation.

This post provides a simple framework for understanding the shadow banking system that is organized around  the two instruments, commercial paper and repurchase agreements, that play an important role in money markets and that are, very roughly, comparable to deposits. Studying how these instruments are used not only allows a distinction to be drawn between the direct finance of assets and the finance of assets that sit on bank balance sheets, but also makes clear why the shadow banking system is unstable.

This analysis finds that the money market instruments have in the past played three roles: they have funded banks and non-financial firms directly, they have funded assets that lie off bank balance sheets, and in order to play these roles, they have created a need for commercial bank guarantees that induce lenders to lend off-balance-sheet or  in the case of tri-party repo to investment banks. In practice, the direct funding of assets now takes place only on a very small scale.

Because the two money market instruments, commercial paper and repurchase agreements (repos), are both short-term, it is easy for those who invest in them to “run,” or to decide that they no longer wish to invest their funds with a specific issuer or, indeed, in privately issued money market assets at all. Because these investors can always choose to put their money in Treasury bills or bank deposits, runs in the money markets are associated with unmanageably sudden shifts in investor preferences across short-term assets. In short, a fundamental attribute of the shadow banking system is that the decisions of money market investors can destabilize the money markets.

Money market mutual funds  and enhanced cash funds (that promise liquidity, but are less regulated than money market funds) are the most obvious money market investors, but the buy-side of the money market is composed of a huge array of institutional investment funds, corporations, and government bodies that have funds they wish to keep in liquid form. All of these entities can be part of a run in the shadow banking system. In addition, as will be explained in detail below, in the repo market it is possible for the recipients of funds, such as prime brokerage clients and banks in the interdealer market, to run.

Now that the basic instability of the money markets has been established, the next step in understanding the shadow banking system is to understand the different ways in which commercial paper and repo-based instruments are used; this is discussed in sub-part A. The following sub-parts evaluate what shadow banking does, and discuss why it is more unstable than traditional banking.

A.  Shadow Banking Instruments

1.  Commercial Paper

a.  Unsecured

Commercial paper is traditionally an unsecured obligation to make a payment that has a maturity of one year or less. It is analogous to the commercial bills that were used to finance economic activity in 19th c. Britain, and indeed has existed in one form or another for centuries.

i.  Issued by financial institutions

A little over half the commercial paper issued in the United States, or approximately $550 billion, is issued directly by financial institutions.[6] Because this market-based funding source is much less stable as a funding source than retail deposits, it is categorized along with other bank funding sources that are prone to runs as wholesale funding. The case of Lehman Bros. illustrates the instability of this form of funding. When Lehman declared bankruptcy, its commercial paper went into default, and set off a run by investors who feared money market mutual fund losses on money funds that invested in commercial paper; as a result the commercial paper market itself faced a run.

ii.  Issued by non-financial corporations

Approximately one quarter of commercial paper is unsecured and issued by non-financial corporations. Because non-financial corporations have less access to liquidity than banks, there is a risk that when their commercial paper is due they will be unable to roll it over into a new issue and will be unable to honor their commercial paper obligations due to this liquidity risk. For this reason, almost all non-financial commercial paper is protected by a liquidity facility provided by a bank, which promises to retire the commercial paper if the issuer is unable to do so. Observe that when Lehman failed, the run on commercial paper was not carefully targeted to financial commercial paper, and as a result non-financial commercial paper was subject to a run as well.

b.  Collateralized: Asset Backed Commercial Paper

In recent decades, sponsoring banks have moved assets that they originated into financing vehicles that are “bankruptcy-remote,” or not available to the sponsor’s creditors in the event that the sponsor declares bankruptcy. In addition, in theory any support that would be provided by the sponsor to the vehicle was defined in a contract, so the sponsor had contractually limited exposure to the vehicle’s liabilities.[7] Thus, these vehicles were designed as a means of removing assets from the sponsoring bank’s balance sheet.

The ABCP market was one of the key markets that collapsed in the early days of the financial crisis – from $1.2 trillion outstanding in early August 2007 to $905 billion three months later. Since then the market has continued to decline slowly, and it now hovers around $250 billion.

Because these vehicles finance long-term assets they face the same liquidity risk as non-financial issuers when issuing commercial paper. In addition these vehicles face credit risk in the event that the value of the assets falls below the value of the commercial paper, and the vehicle is no longer fully collateralized. Both liquidity and credit risk must be addressed before the vehicle can receive a credit rating that is high enough for it to issue asset-backed commercial paper (ABCP) that is secured by the assets in the vehicle. The three principal means by which liquidity and credit risk were resolved are discussed below.

i.  Bank supported ABCP:  Conduits

Prior to the financial crisis most ABCP was issued by ABCP conduits that were sponsored by banks. The banks typically provided both a liquidity facility, which guaranteed that the commercial paper would be retired even if it could not be rolled over, and a credit facility, which promised to honor some fraction of the commercial paper in the event that the value of the collateral fell too low to cover the costs of repaying the commercial paper.

In August 2007 when the crisis started there was a sudden loss in confidence in the ABCP market and many conduits could not roll over their commercial paper. The banks had to step in and honor the liquidity guarantees that had been made – and in order to do so they had to seek regulatory exemptions that are documented by the Federal Reserve.[8]

ii.  Liability structure supported ABCP: SIVs, LPFCs, etc.

Some ABCP-issuing vehicles guaranteed the payment of ABCP by funding the assets with a combination of bonds, medium-term notes and ABCP. These vehicles took many forms; the most common were called  structured investment vehicles (SIVs).

The concept behind these vehicles was that, in the event that the commercial paper could not be rolled over or the value of the assets fell below a trigger point, assets would have to be sold to pay off the ABCP and any losses would fall to the longer term debt holders. In 2007 most SIVs hit their triggers and were unwound. Because of the losses that were incurred by both longer-term and commercial paper investors (after lawsuits determined the allocation of proceeds), they are no longer a popular investment product.

iii.  Repo Conduits – discussed below

2.  Repurchase Agreements

A repurchase agreement (repo) is a simultaneous agreement to sell an asset today and to repurchase it a specific date and time in the future. It has the same economic effect as a collateralized loan. Typically the amount lent is less than the value of the collateral;[9] the percentage difference is called a haircut.

There are two repo markets: the bilateral repo market and the tri-party repo market. In the bilateral repo market the lender must have the capacity to receive and manage the collateral, whereas in the tri-party repo market the tri-party clearing banks, JP Morgan Chase and Bank of New York Mellon, provide collateral management services for the lenders. Money market investors like mutual funds lend only on the tri-party repo market where the principal borrowers are the dealer banks (although a few hedge funds and private institutions are credit-worthy enough to be accepted as counterparties on this market).[10]

The clearing banks also provide bank guarantees of liquidity to the tri-party repo market. Because it is the broker-dealers that borrow heavily on this market and because every trade in the market is unwound at the start of each trading day giving the borrowers access to their assets during the day, the two tri-party clearing banks extend credit to the borrowers during the day until the trades are rewound in the late afternoon. Thus the tri-party clearing banks provide a guarantee to the market and bear the risk of a broker-dealer failure during the day.[11] While reform of the tri-party repo market has been high on the Federal Reserve’s agenda, five years after the financial crisis 70% of the market is still being financed by the clearing banks on an intraday basis.[12]

On the bilateral market, where the lender must manage the collateral, the dealer banks are the lenders. The borrowers are prime brokerage clients, such as hedge funds, and other dealers.

As a result of this structure, funding generally enters the repo market via tri-party repo and the dealer banks, then, distribute this funding more broadly to their prime brokerage clients on the bilateral repo market. Thus, when a hedge fund buys an asset on margin, it borrows a significant fraction of the purchase price from the dealer bank that is its broker and posts the asset as collateral for the loan in a repo transaction. The dealer bank can then repo the asset on the tri-party repo market so that the dealer bank is effectively intermediating lending from the tri-party market to its client and earning an interest rate spread for the intermediation services. When the asset is of a type that cannot be used as collateral in the tri-party repo market, the dealer may choose to use the asset to raise funds on the inter-dealer segment of the bilateral repo market.

The dealer banks also hold collateral that is posted against derivatives contracts by other dealers and by prime brokerage clients. Whereas the inter-dealer derivatives contracts may have symmetrical collateral posting requirements, prime brokerage clients have typically been required to post collateral without having the right to require that dealer bank follow the same rule when the balance on the derivatives contracts is in the brokerage client’s favor. As a result a dealer bank is almost certain to receive collateral from its prime brokerage services when its client accounts are aggregated. The collateral posted by prime brokerage clients can then be used by the dealer to borrow in the tri-party repo market. As a result of this structure collateral posting by prime brokerage clients on their derivatives liabilities is also a form of financing for the dealer banks.

Thus, dealers often finance their own inventories, their prime brokerage clients’ assets, and any collateral that is posted against derivatives liabilities by other dealers or prime brokerage clients on the tri-party repo market.

The repo market is very different from the ABCP market and from commercial paper markets in general, because a run in one of the latter markets can only be caused by end investors. In the repo market a run can be started either by end investors or by other dealers and/or prime brokerage clients. Darrell Duffie has explained the many channels by which funding can be withdrawn in a repo market. These include: brokerage clients can move their accounts – together with all the collateral they have posted – to another dealer; dealers or brokerage clients who are derivatives counterparties can seek a novation (i.e. transfer) of a derivatives contract in order to post collateral to or expect payment from a more creditworthy dealer; dealers or brokerage clients may seek to reduce new exposures by entering into derivatives contracts that will require a dealer to post collateral; or repo lenders may increase haircuts or stop lending entirely to the dealer.[13] In short, the repo market is subject to inter-dealer and brokerage client runs, as well as to runs by repo investors.

In 2008 it is very clear that both Bear Stearns and Lehman faced a withdrawal of funding from other dealers, from brokerage clients, and from end investors in the repo market.[14]

3.  Repo Conduits

A repo conduit is a bankruptcy remote financing vehicle. The vehicle issues commercial paper that is backed by a repo with a maturity that matches the commercial paper. Thus, a repo conduit is backed primarily by the credit of the repo counterparty. Only if the repo counterparty fails to pay, can the repo conduit foreclose on the repo collateral. Because the term of the repo matches the term of the commercial paper, rating agencies do not require that a repo conduit have a backup liquidity facility.

The credit rating of a repo conduit typically is based entirely on the credit of the repo counterparty.[15] For this reason, repo conduits can be used – by institutions with high credit ratings – to finance assets that would not be eligible for tri-party repo financing.

B.  What Does Shadow Banking Do?

1.  Shadow Banking is a Funding Mechanism for Banks

The most important role of the shadow banking system is to provide wholesale funding for banks. Unsecured wholesale funding is provided when a bank issues commercial paper. Secured wholesale funding is provided when a investment bank uses the tri-party repo market to finance inventories, the assets of brokerage clients, and any collateral posted by counterparties in derivatives transactions.

As of Dec. 31, 2013, financial institutions raised $550 billion unsecured on financial commercial paper markets and the dealer banks used the tri-party repo market to borrow on a secured basis close to $1.6 trillion. 80% of the collateral posted is Treasuries and Agencies. Only $330 billion of private sector assets are financed on this market.

2.  Shadow Banking is a Funding Mechanism for Assets

Before the crisis, the shadow banking system played an important role in funding assets with liabilities that were secured by assets that were held off of bank balance sheets in bankruptcy remote vehicles. When this secured asset funding relied on bank support, it was usually provided by ABCP conduits. When this secured asset funding was made possible by a tiered liability structure, it was provided by SIVs and similar vehicles. When this secured asset funding relied on a maturity-matched repo, it was provided by a repo conduit.

Before the crisis the ABCP market was the most important source of shadow bank funding of private sector assets. (Not only did the tri-party repo market fund private sector assets that were for the most part on dealer bank balance sheets, but it was dominated by Treasuries and Agencies and thus played a relatively small role in financing private sector assets even indirectly.[16]) In post-crisis markets vehicles like ABCP and repo conduits are financing far fewer assets than they did before the crisis. The ABCP market is continuing its slow but steady decline over time and now hovers in volume around $250 billion.

3.  Shadow Banking Allows Money Market Issuers to Rent Bank Credit and Allows Banks to Avoid Capital Requirements

When assets were directly financed by the shadow banking system, it was usually because financing vehicles paid a small fee to “rent” a commercial bank’s credit rating by purchasing a guarantee of the vehicle’s liabilities. Because these guarantees were off-balance sheet, the bank was able to avoid the capital requirements that would have been imposed if the bank had done the lending itself. The role played by the clearing banks in the tri-party repo market is similar: they provide intraday credit in order to give dealer banks access to their assets during the day, but face no capital charge for the credit. Thus, a key function played by shadow banking is the arbitrage of capital regulations.[17]

The liquidity and credit facilities provided by banks to ABCP conduits are examples of unsecured bank guarantees.[18] By contrast, the tri-party clearing banks provide secured guarantees. The intra-day credit that the clearing banks provide to the dealer banks is secured by the collateral that has been posted on the tri-party repo market. Banks may also issue guarantees in the form of swaps that offset the market risk of collateral; these guarantees may be secured or unsecured depending on the derivative contract.

The collapse of the ABCP market since regulators have become attuned to the problem of regulatory arbitrage of capital requirements is just another piece of evidence that the vast majority of financing on the ABCP market at its peak was not driven by economic efficiencies, but by regulatory arbitrage as banks used liquidity and credit facilities to take on credit risk, while avoiding capital requirements. Indeed, the industry reaction to the 2004 Final Regulation governing such liquidity facilities – which resulted in a “reinterpretation” of the regulation that effectively gutted it – is also evidence of the importance of regulatory arbitrage to this market.[19]

C.  Collateralized Money Markets Are More Unstable Than Traditional Banks

The use of collateral in repo markets makes them particularly unstable for two reasons: leverage and the fact that not just lenders, but borrowers, can start a run.

When the price of the collateral in a repo contract falls, the borrower is typically required to post more collateral within a day, and, in the event that the collateral call is not met, the collateral that was posted can be liquidated immediately. While this description shows how quickly market price changes can be reflected in the sale of collateral on repo markets, it does not take the leverage that is ubiquitous on repo markets into account. Because of leverage small changes in the market price of an assets can force the borrower to sell off a large fraction of the borrower’s holding of that asset.

An example (drawn from a Fitch Ratings report) will make the instability inherent in repo market finance more clear.[20] Consider a borrower with a $5 million equity stake, which uses repo markets to finance the purchase of a $105 million portfolio of corporate bonds on which the lender imposes a 5% haircut, so that $1 can be borrowed for every $1.05 in collateral repo’d. The borrower will therefore have a leverage ratio of 21 to 1. A 2% decline in the value of the portfolio would reduce the total portfolio value to $102.9 million, reducing the equity in the portfolio to $2.9 million. If we assume that the borrower has no additional equity to contribute, the borrower can now only finance a $60.9 million portfolio at a 5% haircut. In short, because of the leverage inherent in using repo markets to finance assets, a 2% drop in portfolio value can force a sale of 42% of the assets held. Note that this example doesn’t take into account the possibility that the lender increases the haircut on the repo, which would mean that even more of the assets had to be sold. In short, once a borrower has maximized the use of leverage on repo markets – whether the borrower does this intentionally in order to “maximize” returns or simply ends up in this situation after the collateral has declined in price – very small declines in price can force the borrower to sell a significant fraction of the assets. If the borrower is a large market participant, such as an investment bank, this is likely to be the first step in a liquidity spiral, where asset sales further reduce the value of the collateral and trigger additional assets sales.

Not only does leverage make repo markets inherently unstable, but, in addition, a key characteristic distinguishing the repo market from unsecured credit markets generally is that not only the lenders, but also the borrowers, can start a run. The use of collateral in bilateral repo markets makes a borrower run possible, because the collateral can be rehypothecated, or posted as collateral in a subsequent loan by the recipient of the collateral. In short, the collateral posted by borrowers in the bilateral repo market is a source of liquidity for the lender.

When borrowers decide that they don’t want to be exposed to a troubled lender that may not be able to return the borrowers’ collateral in the event that it fails, the borrowers may seek to transfer their accounts to a lender who is not troubled. When the borrowers’ accounts are transferred, the collateral they have posted it transferred with the accounts, and the troubled lender loses the liquidity that was provided by that collateral.

As a result of this property of the repo market, the dealer bank failures of 2008 were characterized by “runs” by both prime brokerage clients and other dealers, none of whom wanted to be exposed to a failing bank. In fact, Krishnamurthy, Nagel, & Orlov conclude that the evidence supports the view that the 2008 crisis looks more like an inter-dealer credit crunch than a run by end investors on the two firms.[21] For these authors one factor distinguishing the two types of runs is the fact that the dealers are well-informed market participants, whereas end investors typically must decide whether to pull out of the market based on very limited information.[22] In short, it is possible that, far from being comparable to bank runs, the runs that took place in 2008 were runs that started with the most informed participants in financial markets.

Thus, there are two very important differences that make the repo market more unstable than unsecured funding markets. Not only does leverage mean that a small decline in price can easily force a large sale of asste, but in the bilateral repo market a run can be started not only by lenders, but also by borrowers.

In conclusion, it is misleading to describe the shadow banking system that exists today as “money market funding of capital market lending” and to focus on it as a means of financing assets,[23] because at present by far the most important use of shadow banking instruments is to provide wholesale funding for dealer banks and through them indirect financing of assets that sit on their balance sheets. Although the view that shadow banking finances assets directly may have held some truth prior to the crisis when $1.2 trillion of ABCP financed bankruptcy remote vehicles, today, to the degree that shadow banking disintermediates commercial banks, it does so by reintermediating investment banks – using a form of funding that is even more unstable than deposits.

The key question that regulators have yet to answer is whether this collateralized wholesale funding market is a valuable addition to the financial system or whether the risk of instability that accompanies it is so great that lending on this wholesale market should be curtailed.

[1] Zoltan Pozsar, Tobias Adrian, Adam Ashcraft, & Hayley Boesky, Author’s Note in Shadow Banking, NYFRB Staff Rep. No. 458 (July 2010).

[2] Perry Mehrling, Zoltan Pozsar, James Sweeney, and Daniel Neilson, Bagehot was a Shadow Banker (Nov. 2013).

[3] For example, although only $35 billion of private label residential mortgage-backed securities have been issued since 2008, at the end of 2013 more than $1 trillion of such securities remained outstanding. Data from SIFMA:

[4] Perry Mehrling, Zoltan Pozsar, James Sweeney, and Daniel Neilson, Bagehot was a Shadow Banker 2 (Nov. 2013).

[5] See Frank Fabozzi & Michael Fleming, U.S. Treasury and Agency Securities 11 (April 2004), available at

[6] Federal Reserve Commercial Paper Release, Outstanding

[7] In practice, banks sometimes supported these vehicles even in the absence of a contractual obligation to do so, and sometimes did not.

[8] See the letters granting JPMorgan Chase & Co., Citigroup Inc., and Bank of America Corp. Regulation W exemptions that are dated August 20, 2007, available at the Federal Reserve website:

[9] Note that in securities lending, where institutional investors provide high-quality, high-demand collateral like Treasuries to the market, haircuts frequently go in the reverse direction. That is, more money must be lent than the value of the collateral in order to induce the securities lenders to lend.

[10] Tobias Adrian, Brian Begalle , Adam Copeland , Antoine Martin, Repo and Securities Lending, Federal Res. Bank of NY Staff Report No. 529, Feb. 2013 at 5-6.

[11] Adam Copeland, Darrell Duffie, Antoine Martin, and Susan McLaughlin, Key Mechanics of The U.S. Tri-Party Repo Market, 18 FRBNY Economic Policy Review 17, 22, 24 (2012).

[12] William C. Dudley, speech, Introductory Remarks at Workshop on “Fire Sales” as a Driver of Systemic Risk in Tri-Party Repo and Other Secured Funding Markets, Oct. 4, 2013.

[13] Darrell Duffie, How Big Banks Fail 23 – 42 (2011). See also William Dudley, More Lessons From the Crisis, Remarks at the Ctr. for Econ. Policy Studies Symposium, (Nov. 13, 2009), available at; Adam Copeland, Antoine Martin & Michael Walker, The Tri-Party Repo Market before the 2010 Reforms 56-58 (Fed. Res. Bank of N.Y. Staff Rep. No. 477, 2010).
Duffie observes that when there is a repo market run, the coup de grace is almost always given by a clearing bank when it responds to concerns about a firm’s financial position by exercising its right to offset aggressively, by for example demanding collateral for intraday exposures or refusing to give access to deposits. Duffie, supra note 9, at 41¬-42.  See also Tobias Adrian & Adam Ashcraft, Shadow Banking Regulation 17 (Fed. Res. Bank of N.Y. Staff Report No. 559, 2012).

[14] Duffie, at 23-42.

[15] Moody’s Revises Approach To Counterparty Rating Actions In Repo ABCP Conduits, Oct. 21, 2009, available at

[16] Arvind Krishnamurthy, Stefan Nagel & Dmitry Orlov, Sizing Up Repo 22 (NBER Working Paper No. w17768, 2012).

[17] Carolyn Sissoko, Note, Is financial regulation structurally biased to favor deregulation, 86 Southern California Law Review 365 (2013). Sissoko also has a discussion of the broader literature on the role of regulatory arbitrage in the ABCP market.

[18] See id. for details.

[19] See Sissoko, Deregulatory Bias at.

[20] Fitch Ratings, at 8.

[21] Arvind Krishnamurthy, Stefan Nagel & Dmitry Orlov, Sizing Up Repo 19,22 (NBER Working Paper No. w17768, 2012).

[22] Id. at 6.

[23] Perry Mehrling, Zoltan Pozsar, James Sweeney, and Daniel Neilson, Bagehot was a Shadow Banker (Nov. 2013).

Is modern finance the source of secular stagnation?

The recent discussion of secular stagnation has once again brought up the question of whether there is a “savings glut” that is aggravating our problems. To the degree that a savings glut exists, it generally has the property that it is focused on the safest assets. That is, for reasons that remain unclear, the collapse of returns on safe assets has not be sufficient to turn this “savings glut” into a vast flow of funds into real-economy risky assets.

I believe that there has been too little discussion of the possibility that the marginal “investors” who have created the savings glut are to be found in the financial industry itself. Although it is certainly true that after the Asian crisis developing countries became net savers — and I do not discount this factor in the flow of savings — at the same time there was a significant transformation of the financial industry. The growth of derivatives was accompanied by the growth of the collateralization of derivatives and the latter phenomenon accelerated after the LTCM crisis which took place one year after the Asian crisis. Thus a non-trivial component of the “savings glut” is likely to be the demand for collateral of the financial industry itself. This source of demand can also explain the strong preference for “safe” assets, since risky assets can easily become worthless as collateral in a liquidity crisis.

If my thesis is right, then Basel III is probably aggravating the “savings glut” problem by increasing the demand for collateral on the part of financial institutions. Thus there has recently been discussion of the existence of a collateral shortage, which sounds to me like the mirror view of a savings glut. (Note that the question of a collateral shortage is complicated by the fact that collateral circulates just like deposits in a banking system, but this issue goes beyond what I want to address in this post.)

One problem with a “savings glut” that is generated in significant part by a demand for assets to be used as collateral is that it is likely to create a segmented markets effect: that is, a significant demand for highly rated assets can coexist with very tepid demand for typical, real-economy, somewhat risky assets that don’t have good characteristics as collateral. This kind of demand for assets is unlikely to play a part in economic recovery by supporting an increase in lending.

The basic problem is this: If the role of the banking system in the economy is to manage and to bear risk for the rest of the economy, then trying to make the banking system “safe” by requiring it to hold vast amounts of collateral and by making it distribute to others the risk that it is supposed to be bearing may actually prevent it from performing its role in the economy. If our banking system is no longer capable of bearing good old-fashioned credit risk, but must find others upon whom to lay that risk, then we should not be surprised that the outcome is low levels of lending to the real economy, low investment, and poor growth. In short, we cannot make the financial system “safe,” by discouraging it from carrying real economy risk, because that undermines economic growth and the performance of all assets.

How the 2005 bankruptcy reforms guarantee unfair returns to derivatives counterparties (and our largest banks)

The effects of the derivatives safe harbors in bankruptcy that protect financially sophisticated creditors at the expense of the bankrupt company’s other creditors (and were greatly expanded by the no-derivative-left-behind act of 2005 — a.k.a. BAPCPA) are demonstrated by this Lehman Brothers lawsuit.

The former purpose of the bankruptcy process was to guarantee that all creditors receive no more than they are due under the law given that the bankrupt debtor is unable to meet all of its obligations.  The modern bankruptcy process allows derivative and repo counterparties to foreclose on any collateral posted to them. While theoretically they must demonstrate that the collateral they have taken was no more than was owed to them, the imprecision inherent in the process of marking complex assets to market clearly gives these counterparties the upper hand. In the event that discussion fails to result in the return of improperly seized collateral, the bankrupt debtor must sue the counterparty to get what is due the other creditors, as the Lehman lawsuit aptly illustrates.

It is a trivial matter to show using economic analysis that the costs to the bankrupt debtor of suing will guarantee that the vast majority of derivative and repo counterparties will be able to keep more collateral than they are due (unless they were undercollateralized at the date of default). In short the bankruptcy process now favors financially sophisticated creditors over trade creditors and debt-holders, not only because the financially sophisticated are able to negotiate more favorable contracts before bankruptcy, but also because they are able to take more than they are due under the law once a company enters bankruptcy.

Update 5-8-13: Matt Levine has done the yeoman’s job of reading some Lehman-related legal complaints and appears to reach similar conclusions.

Update 5-15-13: This article leaves the impression that that best targets for a lawsuit regarding the closing of a derivative contract are those who can’t afford expensive legal advice.

The Problem of Collateral

There are two major problems with collateralized interbank lending.  The first is that there’s no reason to believe that collateral will function to protect the lender in the event that a major bank fails and the second is that the shift from unsecured interbank lending to collateralized interbank lending is likely to have a contractionary effect on the money supply.

At least since Keynes, there has been general recognition that the analysis of aggregate economic activity and the analysis of an individual’s economic behavior require different tools.  The reason for this is simple, individuals can often be viewed as price-takers, who have no effect on the aggregate economy.  Effectively micro-economic analysis abstracts from the problem of liquidity, whereas macro-economic analysis must confront this problem directly (which is not to claim that the dynamic stochastic general equilibrium models that dominate the field of “macroeconomics” today  typically do confront the problem of liquidity, but that’s a different debate).

Alea points out (see comment here) that Basel II discourages banks from lending to each other on an unsecured basis.  The fallacy that regulators appear to be engaging in when they favor collateralized interbank transactions is precisely the fallacy that Keynes criticized forcefully in Chapter XII of the General Theory:

Of the maxims of orthodox finance none, surely, is more anti-social than the fetish of liquidity, the doctrine that it is a positive virtue on the part of investment institutions to concentrate their resources upon the holding of “liquid” securities. It forgets that there is no such thing as liquidity of investment for the community as a whole.

Regulators are failing to distinguish between what is optimal for the individual bank and what is optimal for society.   Liquid assets are supposedly “safe” – but for the problem that liquidity itself is inherently ephemeral.  How precisely do the regulators imagine that collateral posted by a systemically important financial institution (SIFI) is going to protect the lenders?  If the SIFI goes down, there is, in the absence of central bank intervention, a fire sale.  And if they’re counting on central bank intervention to make it possible for collateral to function to protect the borrowers from losses (e.g. via a PDCF or TSLF), why not just rely on traditional central bank lending to banks in a crisis?  What precisely does collateral posting by a SIFI add to the existing system of central bank crisis support for regulated financial institutions?

As discussed in my previous post, the biggest problem with allowing SIFIs to post collateral to one another is that it discourages them from restricting credit to banks that are poorly managed.  By discouraging normal market forces from working to limit the growth and interconnectedness of bad banks with the rest of the financial system, a collateralized interbank lending regime places an enormous burden on regulators to both identify and shrink a bank that has deep connections with the rest of the financial system.  Arguably collateralized interbank lending places an impossible burden on regulators.

The second major problem with shifting from a system of unsecured interbank lending to a collateralized banking system is that in the process of purging the money supply of unsecured debt, the money supply may well have to shrink to the size of the collateral base. Precisely because the shift to collateralized interbank lending creates strong contractionary pressure on the money supply, there is a call for governments to create “safe” assets – that is to increase the size of the collateral base to accommodate the money supply.

Why not call on the banks to create safe assets by underwriting loans carefully?  Such loans after all have historically been all that is necessary to back the money supply.  Perhaps the answer to the question is that “safe” privately issued loans aren’t part of the economic model being used?  I sometimes feel that macroeconomic models that treat government as the social planner’s deus ex machina have so infiltrated some economists’ thought processes that they actually expect a real world government to successfully play the role of a benevolent deity.

If the financial system is so fundamentally unsound that the banks should not be extending unsecured interbank credit each other, the government is not going to be able to do anything to save it.

Related Posts:
What banks do:  monetize human capital
What is capital?
Comparing bankers past and present 

Could Goldman have survived financial collapse?

John Gapper parses Goldman Sachs’ view of the government’s role in saving both the financial system and Goldman, and tries to understand how government intervention could be “indispensable” for the financial system, but not indispensable for Goldman.

I think the answer to this conundrum lies in the system of collateral posting for OTC derivatives (that was put in place with heavy lobbying from Goldman as well other TBTF financial institutions).   As Lehman made clear, what happens when a financial firm is about to fail is that all of its counterparties swoop in and take ever increasing amounts of collateral — leaving shareholders and unsecured creditors with almost nothing.  Thus, in the event of a financial collapse, there will be one or two firms left with almost all the assets of the financial system (that existed prior to collapse).  I think it’s probably a pretty safe bet that the last firms standing will in fact be solvent.

The way I read Goldman’s statements:  We had every intention of managing our collateral demands in the event of a financial collapse so that Goldman would be one of the last firms standing.