A June 25, 2012 blog post at the Fed’s Liberty Street Economics by Adam Copeland, Isaac Davis, Eric LeSueur, and Antoine Martin, caught our attention. Entitled “Mapping and Sizing the U.S. Repo Market” it’s not hard to see why. It is a good primer on the repo markets. But what really got us interested was the sizing of the repo market: $3.054 trillion in repos, $2.446 in reverse repos.
Earlier numbers benchmarked the US repo market at $10 trillion. These numbers came from a well known paper by Gorton and Metrick “Securitized Banking and the Run on Repo”, Jan. 2009, revised Nov. 2010, which, in addition to research the authors did, looked at numbers published by Singh and Aitken (2010) and Bank for International Settlements economists Hördahl and King (2008). The Liberty Street Economics blog notes that the Gorton and Metrick approach suffers from double counting.
We are not sure where the double counting comes into play, and the post doesn’t really explain. Gorton and Metrick say
“…For every “repo,” the other side of the transaction is a “reverse repo.” This raises the issue of double counting of both repo and reverse repo. The issue concerns whether the relevant number is gross repo or net repo for financial firms. For example, a dealer bank that lent money via repo to a hedge fund (which provides the dealer bank with a bond as collateral) and then borrowed against that bond from another dealer bank would have a net of zero (ignoring the haircuts). But, in a run the depositors want their cash bank and the dealer bank cannot instantaneously get the cash back from the hedge fund. (This would be possible in a clearinghouse arrangement.) So, it would seem that gross repo borrowing is the relevant number…”
We wonder if there is some confusion about the term “gross”. Certainly including both repos and reverses for the same trade is wrong. For every repo there is a reverse, but it is simply the flip side of the same transaction. You wouldn’t count a sale by a car dealer as one sale and a purchase of that car by a consumer as another. But when a broker/dealer is intermediating between two counterparties – say a different broker/ dealer on one hand and a hedge fund on another and acts as principal in the middle — then there are two independent trades from a credit point of view and the numbers should be additive. We think this is what Gorton and Metrick are saying, but we can’t vouch for how others are counting. It is an exercise in triangulation.
Several recent speeches by regulators have bemoaned the lack of transparency in the repo market (in particular the bilateral market) and have advocated a trade repository. In the US there are a couple places to look for the information. The blog post wrote about the tri-party data that the Fed collects as well as information that comes from DTCC. Bilateral trade data is, indeed, hard to come by. There is bilateral specials trading that goes through FICC that wasn’t mentioned in the blog post. Equilend and their fixed income arm Bondlend, who put together counterparties but don’t clear the trades, also might be a place to look for bilateral trade data. DTCC already runs a swaps data repository and must be pretty far up the learning curve on how to get it done.
The post also stated that equities are not a significant component of repo. As a follow-up, we would ask if equities financing in the Prime Brokers are included? That is a pretty big business. Even though the underlying agreements might not be the same used in repo, if it walks like a duck, talks like a duck…. Also, don’t forget some PB businesses (and repo & derivatives groups in broker/dealers) use total return swaps (TRS) to synthetically finance their clients. Booked under ISDAs, these are financing deals in function and ought to be included too.
So what is holding back the process of capturing all the repo/financing flows in a central data depository? The first place to look is common client identifiers. Without a common language, it will look more like the tower of babel. Those trades, when they are reported, should also include fields (in addition to the basic collateral, start date, rate, face, etc.) which specifies that it is a financing transaction, perhaps with a way to further sort trades by client type, underlying agreement, maturity, GC or special, haircut, etc. The IT guys won’t be happy about this.
And when it is all said and done, don’t get your hopes up. It will be hard to extract meaningful data from a repo trade repository. To use it to, for example, track haircuts over time – a good way to sense if the market is adjusting to stress –when each dealer can treat client credit differently or may have trades which hedge their risk but aren’t repo trades — can result in a mass jumble of meaningless information. At best we can hope for better information on the size of the financing market (as long as the net is cast wide enough), a broad brush of exposure by broker/dealer or client, and some macro trends.
A link to the blog post is here.