Swaps Execution: The Dogs Still Don’t Eat the Dog Food When They Have the Choice
The Bank of England has received a lot of attention for its just-released study on the liquidity impact of swap execution facilities (SEFs). It finds that:
as a result of SEF trading, activity increases and liquidity improves across the swap market, with the improvement being largest for USD mandated contracts which are most affected by the mandate. The associated reduction in execution costs is economically significant. For example, execution costs in USD mandated contracts, where SEF penetration is highest, drop, for market end-users alone, by $3 million–$4 million daily relative to EUR mandated contracts and in total by about $7 million–$13 million daily.
The basic methodology is to use a difference-in-difference approach to compare measures of liquidity pre-and-post SEF mandate, and which exploits the fact that non-US banks can avoid the mandate by avoiding trading with US banks: this avoidance issue will is important, and I will discuss it later.
The study is carefully done, but I am not persuaded. For one thing, the measures of liquidity employed are driven by data availability (transactions prices), and are not the ideal measures of liquidity. The primary liquidity measures employed are really measures of price dispersion, rather than preferred measures of liquidity such as bid-ask spreads, depth, or price impact (although greater (lower) price dispersion could be associated with lower (greater) price impact).
There is also the issue that transaction characteristics are endogenous. For instance, it may be the case that it is cheaper to do large deals off-SEF than on-SEF. These deals will tend to be done at prices that are further away from the mean price (or the end-of-day midpoint price), and in the volume-weighted measures employed in the paper, these deals get a bigger weight in the liquidity measures. Thus, price dispersion may be greater pre-mandate, and in Europe, where the mandate can be avoided not because SEFs improve liquidity, but because it is prohibitively costly to transact large deals on SEFs. That is, the results could be symptomatic of a loss of liquidity on some dimensions, rather than proof that the mandate improves liquidity.
The paper documents that there is less dealer intermediation where the mandate is binding. This could also reflect changes in transactions characteristics. Dealers are more likely to be needed to intermediate big deals, or deals that are exceptional on some other dimension.
The paper doesn’t break down transaction characteristics by mandate-impacted and non-mandate-impacted subsamples, and in particular doesn’t include a measure of the dispersion of transactions sizes. As a result, it’s not possible to determine whether the mandate has altered the mix of transaction characteristics.
This relates to another finding of the study: namely, that the mandate has led to the fragmentation of the interest rate swap markets along geographic/currency lines, with SEFs gaining far lower penetration in Euro-denominated swaps that are dominated by European banks who can avoid the mandate by trading with one another, and by trading with European end-users. There is confirmation of this result from Tabb Group, which finds that “European derivatives market continues to resist electronic trading.”
Well, this raises the dog food question: If the dog food is as great as the ads say, why don’t the dogs eat it? if SEFs are so much more liquid, why don’t traders flock to them?
When given a choice between a statistical finding, and revealed preference, I go with the latter. Those who actually internalize the cost of trading largely avoid SEFs. This suggests that they are actually costlier to use, at least for some users, such as those who want to trade in large size, or have other idiosyncratic needs. The choices of those who have the choice strongly suggests that the statistical evidence purportedly showing lower execution costs on SEFs is flawed and misleading.
With this in mind, it was gratifying indeed to see CFTC Chairman Massad stating that he favors allowing market participants to decide whether they transact swaps electronically or using traditional voice execution. There was never a compelling case-or even a weak one-for forcing diverse market users with diverse transactional needs to use a one-size-fits-all execution method. Massad’s free-to-choose approach is therefore a vast and welcome improvement over his predecessor Gary Gensler’s monomaniacal determination to bash everybody over the head with a CLOB.
Update. Here’s a more detailed description of the Tabb Group study I linked to above. One important takeaway: European end users really hate electronic execution, and really love voice execution:
Despite the cost benefits of e-trading, institutional investors still prefer to interact with their dealers via phone. Nearly 80 [!] per cent of the more than 200 European investors interviewed as part of the Greenwich Associates 2015 European Fixed-Income Study confirmed their trading protocol of choice was the phone.
“These trades often require white-glove treatment, and clients work with dealers that are best at limiting market impact and providing the support needed to get the trade done,” says Greenwich Associates Managing Director Andrew Awad (pictured). “As a result, clients still place a high value on the support provided by swaps salespeople in executing complex and large trades.”
This strongly suggests that there are likely to be considerable differences between deals done on SEFs and those done the old fashioned way. Not particularly the point about “limiting market impact.” Those who want to do trades that are “complex and large” go to dealers to trade bilaterally to avoid price impact. If they are not able to do that, because of an electronic execution mandate, they will almost certainly trade differently. Fewer big, complex trades. If that is correct, then the Bank of England study is comparing grapes to grapefruit. If so, the difference in price dispersion documented in the study does not demonstrate greater liquidity on SEFs: it demonstrates worse liquidity, at least for some kinds of trades.
Mind you: end users were the supposed beneficiaries of the SEF mandate. According to GiGi et al, they were being shamelessly exploited by dealers, and SEFs would set them free. Apparently they like their chains just fine, thank you very much.
Again: revealed preference rules. Believe it over a stat every time.
Do you know where IRS bid-ask and depth data can be obtained? In our (IMF) work on market liquidity (http://www.imf.org/External/Pubs/FT/GFSR/2015/02/pdf/c2_v3.pdf) we were able to buy Markit CDS data of this sort but never came across anything on IRS. And maybe this is a bit off-topic but I’m not totally convinced by ISDA’s claims of IRS market bifurcation as the LCH.Clearnet data they rely on doesn’t account for U.S. dealers masquerading as Europeans via corporate shell games (e.g., European affiliates of U.S. dealers). We asked LCH.Clearnet if it was possible to identify such masqueraders in their data but they said it would be too difficult.
Comment by Kiffmeister — January 27, 2016 @ 8:02 am
@kiffmeister-No, I’m not aware of any source of this data. I vaguely recall an ISDA study from about 4-5 years ago comparing liquidity in swaps vs. futures, and I think that they obtained some data from the dealers, but to my knowledge historical data is not available.
Regarding masquerading, that makes sense . . . and fits in perfectly with the dog food point. Those who can avoid trading on a SEF, do so.
@Kiffmeister-The stuff I remember. Here’s the ISDA paper I vaguely remembered. It appears that they hand collected bid and offer information from single dealer platforms from 7/9/11-8/12/11.