‘Look, your worship,’ said Sancho, ‘what we see there are not giants but windmills, and what seem to be their arms are the sails that turned by the wind make the millstone go.’
—'Don Quixote' by Cervantes
I have to admit it’s exhausting debunking the regular stream of caustic mythology around ETFs. There’s a bull market in hyperbole headlines around ETFs, and it’s been going on since the 1990s.
The latest volley in this endless game comes from yesterday’s Wall Street Journal, in an article titled: “In a Down Market, ETFs Could Make Things Even Worse.” The article is a survey of an academic papers that’s been circulating (and being debunked) since 2017, called “ETF Short Interest and Failures-to-Deliver: Naked Short-Selling or Operational Shorting?” by a collection of authors from Villanova and Darden, which you can find here.
The core arguments of both are fairly simple: Market makers can sell shares of an ETF they don’t own, and then just not deliver them, known as a “fail.” Those authors go on to suggest this creates all sorts of perverse incentives and opportunities for shenanigans.
Their paper is 80 pages long, and I promised myself I wasn’t going to go line by line and rebut every angle of it, so let me just hit the high spots and hopefully put a few things to rest.
How ETF Settlement Actually Works
OK, so creation and redemption 101. The ETF “XYZ” is hot today. It trades up up up! Not enough shares exist to meet demand. The authorized participant (who is also a market maker) steps in and sells 100,000 shares. That selling pressure keeps the price of the ETF close to fair value. They do this because they can immediately arbitrage out the price difference between XYZ and whatever it holds, let’s say a basket of stocks.
Here’s the very first place the paper goes off the rails, saying that this shorting is often “naked”; as they put it, “operationally short.”
No reasonable person would suggest this.
The paper suggests that APs habitually leave half of this arbitrage open—selling XYZ while not doing anything on the other side. The paper proposes that APs do this because they act on the hope that natural order flow will even-out their positions.
‘Never Wear A Position’
The problem is that, in 25 years, I’ve never met an AP who’s acknowledged this happening. Modern risk management basics for an AP and market maker include the axiom “never wear a position.”
In other words, if they sell 100,000 shares of XYZ, that would create a significant short exposure to the basket of stocks.
While they may not buy precisely what would be required to do a creation today, they don’t do nothing. They hedge that risk immediately. They’re not in the business of making giant market calls, whether that hedge is in a related ETF, a sub-basket of stocks or a derivative. The price pressure between the ETF and what it holds is effectively neutralized.
So what actually happens? The AP holds on to both their short position in XYZ and their long hedge. They have up to five days to submit an actual creation order to fill that 100,000 share short position, or they can just neutralize that position naturally through trading between now and T+5.
Why not do the creation right away? Because doing a creation isn’t free! If they can avoid doing the creation for a few days, there’s a small economic incentive to do so.
What About Fails?
The paper makes a lot of noise about ETFs “failing to deliver.” Technically, this is correct, but only because of bad terminology. Reg SHO calls a “fail” any trade that has not been closed out by T+2 (two days after the trade). Market makers get till T+5; that is, five days after trade). So every time a market maker takes the time they are intentionally given to do their job, they trigger a failure under reg SHO.
This sounds scary, but literally nothing happens. The person who bought 100,000 shares of XYZ doesn’t have them ripped away from them, even if the market maker literally went bankrupt between T and T+5. The National Securities Clearing Corp. stands behind every single trade that goes through the overnight trade reconciliation process. On the evening of the original trade, the market maker is no longer the counterparty to those 100,000 shares anymore—the NSCC is. The person owed something by our market maker is the NSCC. Not the buyer. That’s how settlement works.
So does the NSCC go after the market maker? No, because the market maker is allowed to wait until T+5. At that point, if the securities aren’t delivered, the position is “bought in” for the delinquent market maker, fines are levied, hands are slapped, etc.
Real Costs In Failure To Deliver
This is not a regular occurrence, and there are real costs associated with it to the market maker. Firms with net debits to the NSCC post-settlement pay 2% on that debit, and also incur charges up to $10,000 each and every time it happens. (This is all public info; you can get it from the Depository Trust Company.)
Importantly, even if this were a rampant, huge problem, the counterparty risk is between the NSCC and each member trading firm. It is not between Wall St. banks or market makers.
So what actually isn’t a problem in the first place gets reported as a huge problem, because the “fail” tag kicks off at T+3, even when there are two days left to settle. It’s the regulatory equivalent of the police pulling you over for doing 45 mph in a 55 mph zone: “I’m gonna have to give you a warning, because at the end of this on-ramp, you might be speeding.”
Classic Statistical Mistake
Worse, the paper commits a classic statistics-class mistake of assigning causality to a relationship. One novel approach the authors of the paper take is to look at whether there's a correlation between different firms failing and different ETFs failing.
Unsurprisingly, they find the volume of “fails” tends to stay consistent across ETFs and firms. In other words, when this batch of ETFs has a lot of delays, so does the overall ETF market. When one firm has a lot of delays, so do other firms.
I would have been shocked if this weren’t the case, given that all APs are working in the same market conditions, trying to access the same underlying securities to hedge the same ETF order flows.
If, for example, one biotech ETF is on fire today, chances are all biotech ETFs are on fire. But the paper suggests this is a vector for “contagion,” as if there were some signaling between firms and ETFs that would create some kind of domino problem. As they put it, “operational shorting by one AP may spill over to other APs.” But they never actually get into the mechanics of either how this would happen, or why it represents an actual risk.
