Brian Begalle, Adam Copeland, Antoine Martin, Jamie McAndrews, and Susan McLaughlin
Repurchase agreement (repo) markets played an important role in the 2007-09 financial crisis in the United States, and much discussion since then has focused on the role of repo haircuts. A repo is essentially a loan collateralized by securities. Typically, the value of the securities exceeds the value of the loan and the amount of overcollateralization corresponds to the haircut. In a 2010 paper, Yale’s Gary Gorton and Andy Metrick identified a dramatic increase in haircuts in the bilateral segment of the repo market, which they interpreted as a run on repo. Separately, an industry task force aimed at reforming a different segment of the market, the tri-party repo market, indicated that haircuts may have been too low during the crisis, given the volatility of many of the underlying assets’ values. Maintaining higher haircuts throughout the business cycle could solve both problems: the excessively rapid increase in haircuts in the bilateral segment of the market and the low level of haircuts in the tri-party segment. But are permanent higher haircuts always better? In this post, we dig a little deeper and find that they can have paradoxical effects.
The Case for Higher Haircuts
There are a number of reasons to like high haircuts: 1) they limit the leverage and, thus, the fragility of borrowers, 2) they provide better protection for lenders, making them less likely to run, and 3) if already high in good times, they will likely need to increase less during times of stress, which could dampen the type of dramatic increase that Gorton and Metrick identified and help avoid the “scramble” for alternative sources of funding during a period of already tight credit conditions.
All of these effects reduce the risk of run on a repo borrower and the risk of what we term a pre-default fire sale in which a dealer is forced to sell large quantities of securities at distressed prices to generate liquidity. There is certainly a case to be made that haircuts were too low in many segments of the U.S. repo market before the crisis, and they may still be too low today. So, case closed?
The Danger of High Haircuts
If high haircuts can help mitigate the risk of pre-default fire sale, they can have a perverse effect and increase the risk of post-default fire sales. The latter type can occur because repos benefit from special treatment in the case of bankruptcy. Unlike many types of transactions that are subject to the automatic stay of bankruptcy, repos are exempt. (In the United States, a defaulting broker dealer is resolved under the Securities Investor Protection Act, which imposes a stay on liquidation of securities collateral. This stay is expected to be lifted in a short time period.) So, once a repo borrower has gone into default, the cash lenders can sell the repo securities to repay their loans.
Yet repo lenders may not be able to coordinate the sales of assets to limit their impact on market pricing and liquidity. In fact, some investors may have incentives to liquidate their securities as quickly as possible. Since the haircut protects the lender against a decrease in value of the securities that collateralize the repo, a larger haircut could actually enable cash lenders to conduct damaging fire sales of the repo securities they own post default.
Consider the simple example of a repo for $95, collateralized by a security worth $100 today. The haircut in this case is 5 percent. The lender doesn’t benefit directly from selling the security at a price higher than $95, because any remaining cash must be given back to the bankruptcy estate of the borrower. The investor may be hesitant to sell at a price below $95, because she would receive an unsecured claim on the estate of the borrower for the difference between the proceeds from the sale and the face value of the repo. So if the price drops below $95, and the investor believes that prices may recover from that level, she would be expected to be less inclined to sell quickly.
Now consider what happens if the same security backs a repo of $90, implying a haircut of 10 percent. In case of default, investors would be happy to sell the securities at any price exceeding $90. This example illustrates that the value of the security today, minus the haircut, serves as a “reservation” price. Above that threshold, the investor is willing to sell at any price. Below that threshold, the incentives to sell are much reduced. So with a higher haircut, the price could drop much more before investors have reduced incentives to sell. As we noted in our previously mentioned paper on fire sales, such a drop in price could spill over to other markets and institutions, potentially leading to systemic disruptions.
The Haircut Paradox
So where does that leave us? It is important to recognize that haircuts affect the risk of pre-default and post-default fire sales in contradictory ways. Higher haircuts are an effective microprudential tool that protects individual counterparties against credit loss, and reduces the leverage that borrowers can achieve for their holdings of securities. As such, they reduce the risk of runs and pre-default fire sales.
However, haircuts are a poor macroprudential tool. Once a counterparty has gone into default, high haircuts may permit some investors to sell at large discounts, exacerbating the risk of post-default fire sale. Haircuts: Use with caution!
Brian Begalle is an assistant vice president in the Federal Reserve Bank of New York’s Risk Group.
Adam Copeland is a senior economist in the Research and Statistics Group.
Antoine Martin is a vice president in the Research and Statistics Group.
Jamie McAndrews is the Bank’s director of research.
Susan McLaughlin is a senior vice president in the Markets Group.
In reply to Stefan: Thank you for your question. A commenter to our July 17 post, “Magnifying the Risk of Fire Sales in the Tri-Party Repo Market,” provides an answer to your question. He writes: “Post-default fire-sale risk may be ameliorated somewhat by the 21-day stay on the disposition of securities collateral (including securities sold under a repurchase agreement) typically imposed under section 78eee(b)(2)(C)(ii) of the Securities Investor Protection Act, which would apply to the insolvency of any major U.S. dealer. In other words, in a bankruptcy default situation, repo buyers would not be permitted to dispose of the purchased securities without obtaining the consent of the Securities Investor Protection Corporation, until the expiration of the 21-day stay typically ordered in a SIPA case.” So, an investor would have to wait twenty-one days to liquidate her repo securities, unless SIPC agrees to the liquidation. SIPC has issued letters in the past suggesting that it would act promptly on requests to liquidate collateral. See the original comment here: http://libertystreeteconomics.newyorkfed.org/2013/07/magnifying-the-risk-of-fire-sales-in-the-tri-party-repo-market.html#comments
A post-default fire sale is merely a recognition of the then-current market-clearing real value of the asset. Therefore, it should not be viewed as anything other than a means of forcing recognition of excessive financial speculation earlier, and at a less damaging stage in a bubble’s collapse. Collapsing bubbles earlier, from lower heights, and in atmospheres of lower leverage, should be the goal of every participant in the real economy. Yes, it will thus retain some risk of further loss to financial participants in these markets; and should, at the margin, encourage participants to actually underwrite risk, rather than just wager. And, yes, that will make financial speculation and manipulation more difficult; but that is an inherently good thing.
In regards to a a repo agreement when a dealer defaults – is tri-party repo exempt from the stay on collateral liquidation?