Volatility Series: Trading Halts

September 29, 2015 Philip Pearson


  • Single-stock circuit breakers occurred almost 1300 times on 8/24/15
  • ETFs represent 79% of halts
  • Halts are imprecise but a net benefit

Equity market volatility has come roaring back over the past month. As the VIX flirts with levels unseen since the financial crisis and 2% index moves represent a regular occurrence, the typically sleepy months of August and September have been anything but for equity traders. The past month has put global market structure through an unexpected stress test, and we have been reviewing the data for observations. In this note we will focus on trading halts (single-stock trading halts). In the coming weeks, we will look at additional important issues, including some of the effects of volatility on volume curves, spreads, and liquidity.

The media has covered the market movement on Monday August 24th quite thoroughly. In addition to the intensity of the single day market dislocation, one of the more interesting aspects of the session was the historic number of limit up/down halts triggered in individual stocks and ETFs. A regulatory response to the May 2010 “Flash Crash”, these pauses are generally triggered whenever a stock moves down/up 5%, followed by a 5-minute period before trading resumes . To give some context, there were almost 1300 such occurrences on August 24th; this figure represents more than 30 times the daily average of 40 halts over the past year. The following day, August 25th, actually saw the second largest number of stocks being halted at 130.

Daily Trading Halts

Sequential Halts

One peculiar outcome is that 257 symbols were halted multiple times, often in opposite directions. Data on NYSE: CVS, a large cap stock that trades almost 6mm shares per day, can be seen below. The red tinted area is a halt due to limit down and the yellow is a halt due to limit up. This doesn’t seem right!

CVS Trading on August 24, 2015

The below represents the frequency of halts for symbols with multiple halts on August 24th. 85 stocks were halted 5 or more times.

Halts per symbol


As has been widely reported, the biggest circuit breaker issues were those affecting ETFs. Some have dubbed the day, the “ETF Flash Crash”. Almost 78% of the total halts were on ETFs, a major departure from a typical limit up /down patterns where ETFs represented 32% of total halts. Many market observers have blamed this on market makers struggling to price underlying assets, often linking this to slow openings and the NYSE invoking Rule 48 limiting the pricing information available. This seems to be plausible, and the effect was a real shock to established consensus liquidity expectations for ETFs in a volatile market.

Trading halts on August 24, 2015


Market structure rules and durable price boundaries are necessarily challenged when fat-tail trading occurs. Assume that limit up/down halts follow a normal distribution of outcomes, then August 24th is a 46-sigma event. Also, 5% intraday volatility may seem like a lot, but some stocks, like LTRP-B shares, are halted multiple times nearly every day and seventeen different times on August 24th. We share a desire to carefully define behavior as “extreme”, and perhaps we “know it when we see it” (Jacobellis v. Ohio), but setting concrete policy guardrails is a tricky enterprise. This is especially true for ETFs where you could halt the underlying securities but not the fund or the inverse—is this wise? Hard to say.

There are improvements that could be made to the process, but most of them serve to burden the market with more complexity. Below are a few precision (and complexity) increasing suggestions:

Trading halts table 1

So should we just leave it alone? The current rule is a heuristic, and one should not judge a heuristic on how often it is wrong; rather, what is the impact when it is wrong. August 24th exposed how arbitrary and imprecise the single stock circuit breaker rule can be, but the impact was reasonable for most investors. Stocks eventually traded in a range that was most likely representative of investor expectations, and that’s all we can ask of market structure.

One improvement that might make some sense with very low (negative?) complexity cost would be to extend the halt period from 5 minutes to 30 and only allow one halt per day. This would allow investors to really focus on what’s happening in a halted stock and get the resumption auction right, a process that feels broken currently. Furthermore, this is in line with the original motivation for this process: the idea that a high speed market can get a price level wrong and feed on itself, so humans should intervene when this risk exists. If limit up/down auctions aren’t producing quality price discovery, then the auction parameters should be revisited.

15% for stocks in the Russell 1000 and most ETF’s, 10% for other stocks. Both numbers are doubled in the first 15 and last 25 minutes of the trading day. 2

  • Philip Pearson

    CFA, Director, ITG Algorithms
  • Ben Polidore

    Managing Director, Head of Algorithmic Trading