What Is Market-On-Close?
It's worth understanding what this "market-on-close" thing is, anyway. Throughout the day, the exchanges accumulate order flow marked for "market-on-close," or MOC, pricing. As the name suggests, a market-on-close order means that your order will be filled at the closing price.
That closing price is actually a somewhat complex calculation based on an automated order-imbalance auction run by the New York Stock Exchange (or other exchanges). In principle though, it's the price where the maximum number of market-on-close and limit-on-close shares can trade hands, and is generally very close to the last traded price at 3:59 p.m. Eastern time.
The primary users of market-on-close orders are institutions and index fund managers, who are mainly concerned with tracking error. By definition, putting your trades in the MOC auction means you'll get extremely good tracking to an index like the S&P 500, which itself is calculated on those same closing prices.
But you'll notice that the volume in the final 10 minutes far exceeds the volume actually marked at and after the close, and this is fairly typical. Investment managers often trade the close rather than relying on the auction.
Regardless, the worst-case scenario is that these seven leveraged or inverse funds represented only 5-6 percent of value traded in the final 10 minutes of that hugely volatile day.
A valid criticism of this analysis would be that it undercounts the impact of leveraged and inverse funds as a whole, as it focuses only on S&P 500 funds.
To address that critique, we ran the same study against all levered and inverse ETFs investing in U.S. equities with static leverage factors.
When all those 102 funds are considered, the net end-of-day trading required to rebalance based on the performance of the funds themselves would have been roughly $3.8 billion. While this sounds like a large number, consider the comparable value-traded data for the Russell 3000 securities, representing effectively the entire U.S. equity market:
|Full Day's Trading:||218,919,513,824|
|Final 10 minutes+MOC||25,118,084,644|
|Final 30 minutes+MOC||45,348,146,711|
That $3.8 billion to sell represents just 15 percent of the final 10-minute volume, and well under 10 percent of the final 30 minutes of trading.
The Momentum Argument
While we could argue about whether 5 percent or 10 percent or 15 percent of closing volume is enough to matter, the market can settle the argument for us. If the pro-cyclical pressure of these funds was causing the market to move further than it might otherwise, this should be evident in the data. However, it's not.
Looking back to the beginning of this recent downturn—April 1, to pick a fairly stable date—we looked at the momentum of the market from 3:00, 3:15, 3:30 and 3:45 to close. Because of the pro-cyclical nature of the leveraged/inverse daily rebalance, if that activity were to have a serious effect, we'd expect to see an acceleration of the day's trend. That is, on a day with the S&P down 1 percent at 3 p.m., we'd expect it to be down more than 1 percent by the end of the day.
We chose four "start times" for this study, as in our conversations with issuers, it's clear they all begin their swap coverage negotiations at different times following 3:00 pm.
Here are the results for the S&P 500:
Far from a trend, the data suggests it's truly a coin flip at any time in the last hour of trading about whether the market accelerates or reverses course. It's even more obvious when looking at the momentum of the Russell 3000:
Not only is there no clear trend in the first three time periods, the data for the last 15 minutes of trading would suggest that on the vast majority of days, the market actually reverses heading right into and through the closing auction. In other words, the leverage/inverse effect was not just neutralized but reversed by other factors.
Another version of this analysis would look at the change in price from close to the next morning's open, testing the idea perhaps that closing prices were not reflective of market sentiment, and thus would revert the next morning.
Excellent work on just this phenomenon was done by William Trainor of East Tennessee State University last year, so I have not duplicated his efforts. He found no predictive value in previous days' closing price movements.