Careful ETF traders may have noticed what can best be described as “weirdness” in ETF trading on March 31. First, during the early morning, the New York Stock Exchange had a glitch that caused some 160 ETFs to be untradeable for much of the hour between 10:00 and 11:00 a.m. It wasn't just ETFs that were hit, but it was anything with a ticker from UTG to ZSML alphabetically. That means all the Vanguard ETFs that start with V and all those SSgA tickers that start with X. Because of these accidental trading halts, the Market on Close system for these securities – a whole five hours or so later – was also affected. The end result? Quite a few ETFs, like the Vanguard Mid Cap ETF, traded wildly off of fair value – right at the close. So how is this possible?
The Market-on-Close Process
Many casual investors think that the closing price is whatever the last trade happened to be before someone rings a bell. In the modern markets, that bell is largely symbolic, and there’s far too much activity at the end of the day to leave to chance, so instead, most modern markets run what’s called a “closing auction.”
Throughout the day, NYSE collects orders flagged as either Market or Limit-on-Close. Starting at 3:00 p.m., NYSE starts sending out information about how imbalanced the close is going to be: are there a lot more buyers than sellers or vice-versa. It continues to collect orders and disseminate this imbalance information right up until the close; however, after 3:45, once you’ve got your MOC order in, you can’t cancel it. Market makers can enter orders to offset the imbalance, whatever it might be at 4:00 p.m., but that’s it.
Once 4:00 p.m. hits, all orders are done, and the NYSE runs an electronic Dutch auction – a system that ensures the maximum number of shares can settle at a closing price.
The Information Problem
While the Dutch auction process is relatively simple and used in all sorts of markets, the most critical part of the whole end of day process is the dissemination of information. The goal of the entire process is for the closing price to be as logically derived from all the trading leading up to 4:00 p.m. as possible, minimizing volatility and maximizing crossing. As the close approaches, all the disseminations of imbalances NYSE does are designed to prompt market makers and traders into recognizing where things are headed. If you know, for instance, that there’s a HUGE amount of stock being sold at the close, well, you might want to sell now, and then put in an order to re-buy on the close – selling high, buying low. The whole process should settle down trading and make for an orderly, predictable close.
But when there’s a hiccup in that information flow, that process breaks down. And that’s apparently what happened here. Traders used to seeing indications of the closing imbalance every five seconds right up to the close were in the dark, with no real indication of whether there were more buyers than sellers. In this case, the market makers, who might otherwise have placed closing-offset orders designed to automatically take the other side of the imbalance and lock in a small gain on the spread difference, held off. So all that was left were the sellers with market on close orders.
The Accidental Paint
The irony here is that this was also the final trading day of the first quarter. Such gains, of course, are millisecond-short lived. VO, and every other ETF affected by this glitch, opened up right on fair value, just like they generally do every day, and all those gains disappeared as quickly as they appeared.