Volatility Series 3

February 29, 2016 Philip Pearson
The Winter of Our Discontent vs. the Long, Hot Summer


  • Dark volumes were significantly lower during the volatile period of August–September 2015 compared with the first two months of 2016.
  • Block volumes were also much lower during the summer.
  • NYSE opening auction delays were sharply reduced in 2016.

With the calendar turning to 2016, US equity markets were hit with a fresh bout of volatility, which looked at first glance to be similar to market conditions in August– September 2015. These periods give us an opportunity to take a fresh look at how volatility affects trading behaviors. Previously in the volatility series, we have looked into the August trading halts1 as well as intraday volume patterns2 . In this latest note, we compare recent events to those of the summer. Were there material differences in the types of volatility we saw? Did market structure issues amplify the effect of the volatility?

To start, we focused on numerous equity trading factors such as volume curves, spreads, opening and closing auctions, and transaction costs. We measured ITG Algorithms and Smart Router usage and performance. An initial look at the results revealed similarities between the two periods. In many ways, a volatile day in August is very similar to a volatile day in October is very similar to a volatile day in January. However, there are a few notable differences between these periods. Most intriguing was that the August and September volatility occurred concurrently with a notable decrease in dark and block trading. What was the cause? Additionally, given the increased scrutiny of NYSE’s use of Rule 48 after August 24th, we found a marked reduction in the NYSE opening auction delays.

To compare the periods, we looked at all trading from July 1, 2015– February 17, 2016, breaking it into 4 buckets:

  • August – volatile days between 8/20/15 – 9/30/15
  • January – volatile days between 1/1/16 – 2/17/16
  • Other Volatile – all other volatile days from 7/1/15 – 2/17/16 (excluding “August” and “January”)
  • Not Volatile – All non-volatile days between 7/1/15 – 2/17/16

Volatile days are defined as days where intraday price change for the S&P 500 surpassed 1.5% at any time during the day versus previous day’s close.

Dark Trading Overall VolumeDARK TRADING

On volatile days in August, dark trading accounted for 28.1% vs. 31.5% on average for January, a difference of 11%. Here dark trading is defined as ADF prints for S&P 500 stocks compared with overall market volume in the same securities.3 Interestingly, per the Dark Trading as Percent of Overall Volume table, other days of heightened volatility in 2015 and thus far in 2016 have seen less significant differences in overall dark volumes.

As seen in the Percent Dark by ADV chart, this effect is consistent across ADV buckets. It is most dramatic in lower ADV stocks, specifically <5 million shares. This seems intuitive as dark trading is traditionally more effective for less liquid stocks. Non-volatile days saw the most dark trading (proportionally) across the board; this can probably be attributed to lower overall volumes.

Percent Dark ADV
Block Trades of Overall VolumeFurthermore, as seen in the Block Trades as a Percent of Overall Volume table, block trading4 was also substantially lower in August than any other period, at 7.56% of all trading. This figure is 17% lower than was seen on non-volatile days and 23% lower than January volatile days.

Interestingly, January volatile-day block trading (9.86% of all trading volume) actually increased as proportion of overall trading, representing 8.2% more of trading volume as compared with non-volatile days.

The August combination of lower dark trading and lower block trading is curious. VIX levels during the different periods were similar, but volume was heavier in the January/February period.5 The one obvious difference is that after the first few days of the August period, the price action was relatively neutral with a similar amount of volatile up and volatile down days. Comparatively, the beginning of 2016 saw more of a consistent market decline with only a few up-days mixed in. When considering this fact and the higher volume levels, it leads to the notion that perhaps the recent sell-off has seen more institutional volume compared with more retail speculation in August and September. This would also explain the higher block and dark trading levels in January.


One of the biggest areas of market structure criticism regarding August 24th concerned the delayed openings on the NYSE. Unlike NASDAQ, the NYSE opening can be a manual process that causes many stocks to open significantly after 9:30:00. While ordinarily this is not problematic, August 24th witnessed extreme circumstances. More than 60% of S&P 500 NYSE-listed symbols opened after 9:35; 14% opened after 9:45.6 These delays caused halts in ETFs, which could not get reliable quotes or liquidity, adding to the day’s challenges.

Rule 48 allows the NYSE DMMs during times of high volatility to avoid the requirement of publishing imbalances in an effort to speed up the opening process. Often, it does not actually help the process. Because of media and industry scrutiny, the NYSE – albeit informally – has greatly reduced the usage of Rule 48 in recent months. In the past few weeks, the exchange has indicated an intention to do away with the rule.

To measure this effect over time, we examined auction delays since August 24th. We looked at how long after 9:30:00 the opening auctions were for all NYSE-listed stocks.

NYSE Open Auction Time by ADV
Clearly there has been improvement since August. The median open delay for the three days from August 24th to August 26th alone was over 5 minutes. The worst day since, January 15, saw a median open delay of less than 3 minutes. It seems that the emphasis the NYSE has put on improving the opening process has begun to bear fruit. There might be more impetus for further improvement given the changing ownership in the DMM ranks. Will a more automated auction process follow?

NYSE Open Delays


The beginning of 2016 has been marked by two months of higher levels of volatility, accompanying one of the worst Januaries in stock market history performance-wise. The current macro backdrop has been oft-cited as driver of both the weakness and the volatility. In comparison, while the August swoon continued for 5 weeks, it was largely influenced by the events of one day, August 24th. However, setting aside the causes and persistence of instability, in many ways, January and August were very similar. Both periods saw the VIX rise to greater than double its 250-day average. Average spreads for both periods increased about 17% compared with other volatile days and 25% compared with non-volatile days. Despite this, not all volatility is created equal.

New questions about market quality emerged following the August period. Some of these issues, such as trading halts and delayed openings, have objectively improved. In light of this, we expected to see more disparities between the two market events. As we can tell, there were two major changes in January worth noting; first, an improved NYSE opening auction process, and second, a larger institutional trading presence.

Much attention has been paid to the market microstructure issues, but we find the changing liquidity profile to be more intriguing. Block liquidity is essential to institutional traders. The ability to trade size with reduced impact, acts to smooth large price fluctuations that hurt the perception of market quality. Many traders struggled in both August and January, but the price movements were likely exacerbated in August by the lack of dark liquidity.

The summer traditionally sees lower volumes, institutionally and otherwise, but what explains the lower levels of dark trading? The beginning of 2016 feels different. As this macro instability continues to impact markets, it is worth watching the quality of liquidity available. Is this a new type of institutional selloff? The increased institutional activity certainly warrants further study.

3We are aware that this is an imprecise proxy for “dark” volume for several reasons, primarily because it excludes exchange hidden orders.
4For the purpose of this analysis, we consider all trades of >= 5000 shares for S&P 500 stocks, excluding the Opening and Closing Auctions.
5Excluding August 24th which saw the highest volatility and volume of the period.
6More information about that event can be found in the SEC’s very helpful 88-page report on August 24th.