The flash crash of August 24th, 2015 was not a one-off event, and is quite likely to happen again. At least according to a December 2nd report from Goldman Sachs Equity Research, which argues that another flash crash is almost inevitable unless steps are taken to mitigate the underlying causes.
GS analysts Robert D. Boroujerdi and Katharine Fogertey argue that the August flash crash was caused by structural issues in the financial markets and is very likely to happen again. They note: “On August 24th the equity market went haywire. While many pointed the finger at ETFs as the driver of volatility, that was not the case. Specifically a series of new rules and procedures born out of the 2010 Flash Crash coupled with the lack of harmonization across markets paved the way…”
Boroujerdi and Fogertey go on to describe the structural problems that led to the August 24th flash crash, and offer a number of suggestions on how to improve the situation and minimize the chance of future flash crashes.
Fixing flash crashes (or steps to prevent a flash crash)
The GS analysts suggest that SEC Rule 48 played a major role in the August flash crash. Rule 48 is designed to allow market makers to be able to open stocks more quickly and efficiently than the usual pairing off orders and providing a clearing price.
[drizzle]However, Boroujerdi and Fogertey say that “rules that delay the open might have not stemmed volatility, but rather liquidity likely rushed to open markets.” They point out that when exchanges delay opening trading for equities, transparency is severely limited. That said, even with Rule 48, the NYSE did not open fully until after 9:40 am, whereas NASDAQ and BATS opened fully at 9:30 am.
Another way to mitigate the problems that led to the flash crash is to improve transparency in the market concerning pre-market order imbalances and prices. Although the NYSE Order Imbalances data feed provides real-time auction imbalances, it only operates until 9:35 am, so the market has to digest a major information gap for any stocks that open later than this.
Another part of the solution to flash crashes is revising the limit up – limit down (LULD) rules. LULD rules were implemented following the 2010 Flash Crash to develop more robust circuit breakers for individual stocks when major news hits.
Boroujerdi and Fogertey note that LULD rules have been helpful in helping level out volatility out before August 24th, the large majority of the 1,200 halts that day were on stocks that had traded sharply lower and had started to recover. This dried up liquidity and delayed a recovery. Keep in mind, again, that LULD was enacted for stock moves down as well as up.
Harmonizing the rules of the various exchanges would also go a long way in mitigating flash crashes. Each financial market and asset class has its own unique openings, halts, cancellations and regulators. The problem is when products that are not “harmonized” are tied to each other (for example, ETFs that use futures) or are more liquid (for example, futures vs derivatives) price and liquidity imbalances often result.
For example, erroneous trade rules do not match LULD bands for some securities. The bands for erroneous trades can be out of alignment with the LULD bands, so that a trade that happens within the LULD bands could be cancelled due to erroneous trade guidelines. This also impacts the ability to pursue arbitrage on products across the market. Also of note, LULD bans are wider at the open and close than during regular trading hours.
Another problem is that timing of trading halts also varies across exchanges. For example, a stock is halted for 5 minutes under the LULD plan, but CME futures are typically halted for just two minutes.
Another issue of concern is how to handle trades coming out of a trading halt as the auction procedures following a trading halt again vary by the exchange. Boroujerdi and Fogertey note that since over 85% of the US Exchange Traded Products are listed on NYSE, rules on this exchange have a greater impact on ETFs. After a halt, the stock has specific collars before it can again trade freely in the market.
On August 24th, the auction collars for NYSE Arca were 5% for stocks priced from $0.01 to $25.00, 2% for stocks from $25.01 to $50.00 and 1% for stocks greater than $50.00. The collars are designed to prevent securities from opening much higher or lower than the last trade. But when there is a serious imbalance of buy and sell orders (as in a flash crash), these tight collars actually boost the chances that the auction process will lead to a LULD halt. This is why a number of ETFs were halted multiple times on the morning of August 24th.