3 December 2017 - Matt Levine
No one is around to make sure you can buy or sell at a moment's notice.
Here's the central intellectual problem in bond market liquidity. Practitioners -- bond traders and investors -- all go around saying that liquidity in corporate bond trading is bad. If you call up your dealer and ask her to buy a big slug of bonds from you, she will put you on hold and call other customers to see if they're interested in buying those bonds. If they are, she'll buy from you and sell to them; if they're not, there's no trade. In the good old days she'd just buy them herself and deal with the risk of finding buyers later, but now she is too nervous and capital-constrained and Volcker-limited.
Academics and regulators, meanwhile, look at the data, and they almost invariably say: Meh, corporate bond liquidity looks fine. Volumes are up; bid/ask spreads are tight; price impact is modest. Dealer inventories are down -- dealers really are buying fewer bonds for their own accounts -- but it doesn't seem to be affecting the actual liquidity that anyone experiences in the market.
The practitioners reply: Look, you can't learn anything from the data. The data just shows trades that happened. Of course bid/ask spreads are tight. In a world where dealers buy and sell bonds at their own risk, they will charge a lot for doing it: They'll buy at 99.5 hoping to sell at 100, because there's a risk they might have to sell at 98. In a world where dealers only match customer trades, they don't need to charge as much: They can buy from one customer at 99.5 and sell to another at 99.75 and clip a totally risk-free 0.25. Bid/ask spreads are tighter, but if you actually need to trade a bond, liquidity is worse. An important element of liquidity -- immediacy, the assurance that you can sell a bond whenever you want to -- is gone; you can only sell a bond if the dealer can line up a buyer on the other side. If it can't, then it just won't buy your bond.
Here is an extremely delightful Federal Reserve Board discussion paper from Jaewon Choi and Yesol Huh called "Customer Liquidity Provision: Implications for Corporate Bond Transaction Costs." They find a neat way to shed light on this question using TRACE data of corporate bond trades, which identifies the dealer involved in each trade. They identify when a dealer buys and sells each bond, and classify trades as "DC-DC" (dealer-customer/dealer-customer, where a dealer buys a bond from one customer and sells it to another customer within 15 minutes, or vice versa), "DC-ID" (where a dealer buys a bond from a customer and sells it in the interdealer market within 15 minutes), or "invt>15min" (where a dealer buys a bond from a customer and holds it for longer than 15 minutes). Their assumption is that DC-DC trades -- where a dealer offloads its risk to another customer immediately -- are more or less what I described in the first paragraph above: The dealer is not taking risk on her own book to buy the bonds, but buys them only because she already has an ultimate buyer lined up. Choi and Huh call this "customer liquidity provision," and they find that the customers who provide liquidity -- the ones who are willing to be lined up as buyers to facilitate the dealers' trades -- get paid for it:
We provide empirical evidence that is consistent with predictions implied by customer liquidity provision. First, we show that customer trades that are matched with other customer trades (i.e., DC-DC trades) have lower average spreads than customer trades that are not matched. We find that DC-DC trades have 20–40% lower bid-ask spreads compared with trades where customers are demanding liquidity, indicating that average bid-ask spreads underestimate trading costs for customers demanding liquidity. We find that 40% of DC-DC trades exhibit negative spreads, higher than that of DC-ID or invt>15min trades. Furthermore, DC-DC trades have lower bid-ask spreads and are more likely to have negative spreads particularly for customer buy trades, which is consistent with our prediction that customers have more capacity to provide liquidity on their long positions. For investment grade bonds, for example, DC-DC trades for customer buys (sells) are on average 42.6 (23.7) bps and 17.17 (6.6) bps lower than DC-ID and invt>15min customer buy (sell) trades, respectively. Also, we find that the bid-ask spreads for DC-DC trades are lower particularly for bonds that trade infrequently, consistent with our customer liquidity provision hypothesis in that liquidity-providing customers are compensated more for bonds with high dealer inventory risk.
This seems like pretty good evidence for the practitioners' argument that dealers are not providing as much liquidity. "Our results indicate that liquidity decreased in corporate bond markets and can help explain why, despite the decrease in dealers’ risk capacity, average bid-ask spread estimates remain low," they write.
Still, you could ask: So what? You could imagine a market where dealers don't matter much, where bond investors (mutual funds, pensions, insurance companies, etc.) buy and sell from each other, with dealers playing a minor role. (Lots of people imagine this, which is why there are so many all-to-all electronic trading platforms designed to let bond investors buy and sell from each other.) "Since the periods before the 2008 financial crisis, substantial amounts of liquidity provision have moved from the dealer sector to the non-dealer sector," write Choi and Huh. The system works -- the customers provide liquidity to each other -- so why do you need the dealers?
One answer is that investors like immediacy, and customer-to-customer trading is not great at providing it. "Among trades where customers are demanding liquidity" -- where dealers aren't just matching two customers with each other but actually providing immediacy themselves -- "we find that these customers pay 35 to 50 percent higher spreads than before the crisis." Bid/ask spreads have gone up for the dealers' core work of providing liquidity at their own risk, of buying when there is no obvious natural buyer. There is a social role of a bond dealer in providing liquidity, and dealers were optimized to do it, and now they are not. Customers can do it themselves, but they are not optimized to do it -- they are primarily in the business of investing in bonds, not providing liquidity -- and so they are not as good at it.