Updated to fix the claim that collateral protects repo borrowers.
There have been many reasons recently to attend to developments in the repo market. The Federal Reserve's balance-sheet expansion has since March been substantially funded by repo. What is more, this one-way, crisis-time intervention has moved the Fed closer to establishing a new two-way, non-crisis relationship with repo markets. In general, the US central bank's approach to backstopping the global dollar relies heavily on repo markets.
Repo markets are at the core of the financial system, but they still seem arcane to many. A conversation last week with a Soon Parted reader prompted me to try to explain repo and its importance from first principles.
Financial structure of a repo transaction
Repo is a secured, short-term money market transaction: it is a loan of money secured by a deposit of collateral. The name is short for "sale and repurchase," which refers to how the collateral moves: securities are first sold by the borrower to the lender, which causes money to move from lender to borrower; then bought back, which reverses the flows of both collateral and money. Usually the repurchase price is slightly higher, which amounts to an interest payment to the lender of money. But repo has, I think, grown out of its name.
The T accounts below reflect the way that I think about repo, in its most generic version: overnight repo against Treasury collateral. Above the dotted line are flows of money, below are flows of collateral. There are four entries in the transaction. Two are flows of money: (1) the lender provides reserve money to the borrower and (2) the borrower agrees to return it the next day. The other two are flows of collateral: (3) the borrower posts collateral with the lender and (4) the lender agrees to return the collateral the next day.
The T accounts show the balance-sheet relationships that are created at the beginning of the repo transaction. At maturity, the borrowed money and posted collateral are returned in fulfillment of the promises to do so. If the borrower fails to repay, then the lender is entitled to take ownership of the collateral.
Note that this representation describes only the economic endpoints of a repo transaction: real repo markets also involve custodians, who ensure that collateral is kept safe, or central clearing, which can facilitate netting of offsetting transactions. Accountants would also have something to say about my representation, because there are particular rules for the accounting treatment about all four parts of the transaction, according to circumstances. In particular, note that the Fed's balance sheet does not show flows of collateral.
Sometimes transactions can be driven by the lender's need to lend, rather than by the borrower's need to borrow. This can be seen in the Fed's overnight reverse repo (ON RRP) facility: the Fed is acting as borrower, meaning repo is on the liability side of its balance sheet. This does not represent a need for funds, as repo borrowing would in some cases. It represents, rather, a surplus of funds for its counterparties, money market funds, who need an asset to hold. For this reason I have sometimes described the Fed's ON RRP as a deposit facility.
Secured and unsecured lending
A repo loan is frequently described as secured or collateralized, referring to the flow of securities in the above transaction. Such explicit flow is absent from other kinds of lending. But in a way, the distinction between secured and unsecured may be less than it appears. Even in an unsecured loan, without explicit collateral, a lender of money still has some kind of recourse, which they can act on through the legal system, for example as part of bankruptcy proceedings. One could go so far as to record this claim on the balance sheet. These T accounts show a generic bank loan, where (1) deposits are created against (2) a promise to return them. At the same time, the bank records (3) a potential claim on the equity of the borrower, which claim will be (4) given up when the loan is paid off:
Such a loan would be called unsecured, but it would be more precise to say that there is no explicit security—only a general claim on the borrower.
Money for market-based credit
Repo protects lenders by giving them recourse to explicit collateral. But it also protects borrowers, because the explicit collateral is first to absorb the cost of default, before a general balance-sheet exposure. These features make repo a very cheap transaction, and so suitable for wholesale financial transactions. What is more, in repo the lending of money and the movement of collateral are closely connected. This makes repo the money-market instrument most suited for the short-term finance of securities positions, particularly for securities dealers. Repo, in other words, is the form of money used by the market-based credit system.
Why no mention of specials, which reverse the logic presented above?
《The demand for some specials can become so strong that the repo rate on that particular issue falls to zero or even goes negative in an otherwise positive interest rate environment. The repo market is the only financial market in which, historically, a negative rate of return has not been unusual.》
What if repo lenders are getting specials from the Fed, which they turn around and repo out for an additional profit?
Also, why no mention of rehypothecation?
《Rehypothecation is an alternative name for re-pledging. In the derivatives market, rehypothecation is sometimes called re-use. [...] Rehypothecation is regarded by prime brokers as essential to the economics of their business. In return for rights of rehypothecation, they can offer clients cheaper funding.》
Does rehypothecation allow both sides of a repo transaction to book a profit? Especially with specials and negative prices involved?
Is there a lot more going on in repo markets, which doesn't easily fit with zero-sum assumptions about money? If multiple sides can come away from a transaction booking a profit because they used different derivatives, does that pose a problem for standard economic scarcity assumptions?