DIVISION OF ECONOMIC AND RISK ANALYSIS (“DERA”) 1

Agency decision

Ask Donna

What actually matters in this document.

Text

MEMORANDUM

TO:

FILE

FROM:

DIVISION OF ECONOMIC AND RISK ANALYSIS (“DERA”) 1

SUBJECT:

CORNERSTONE ANALYSIS OF LONG-LASTING STRADDLE STATES

DATE:

December 2017

The Division of Economic and Risk Analysis ("DERA") contracted with Cornerstone

Research (“Cornerstone”), an economic and financial consulting firm, to conduct analyses

related to investigating the occurrence of long-lasting "Straddle States" under the Limit Up-Limit

Down (“LULD”) National Market System (“NMS”) Plan (the “LULD Plan” or “Plan”). 2,3 Under

the LULD Plan, a Straddle State occurs when the National Best Bid (Offer) is below (above) the

lower (upper) Price Band and the NMS Stock is not in a Limit State. A Straddle State restricts

trading from occurring on the side of the quote that is outside the Price Bands. 4 The Cornerstone

analysis defines a long-lasting Straddle State as a Straddle State that lasts at least 5 minutes. 5

The Cornerstone analysis was commissioned by DERA as a result of the Notice of Filing

of the Twelfth Amendment to the LULD Plan, in which the Participants proposed eliminating the

definition of Straddle State from the Plan and the ability of a Primary Listing Exchange to

declare a Trading Pause while a security was in a Straddle State. 6,7 In the Notice of Filing of the

This is a memo by the Staff of the Division of Economic and Risk Analysis of the U.S. Securities and Exchange

Commission. The Commission has expressed no view regarding the analysis, findings or conclusions contained

herein.

See Order Approving, on a Pilot Basis, the National Market System Plan to Address Extraordinary Market

Volatility by BATS Exchange, Inc., BATS Y-Exchange, Inc., Chicago Board Options Exchange, Incorporated,

Chicago Stock Exchange, Inc., EDGA Exchange, Inc., EDGX Exchange, Inc., Financial Industry Regulatory

Authority, Inc., NASDAQ OMX BX, Inc., NASDAQ OMX PHLX LLC, The Nasdaq Stock Market LLC, National

Stock Exchange, Inc., New York Stock Exchange LLC, NYSE MKT LLC, and NYSE Arca, Inc. (altogether, the

"Participants") (“Approval Order”), Release No. 34-67091 (May 31, 2012), 77 FR 33498 (June 6, 2012) (File No. 4631). The original text of the Plan is attached to the Approval Order as Exhibit A (“The Plan”). Subsequent

amendments are available at https://www.sec.gov/rules/sro/nms.htm.

Capitalized terms used herein but not otherwise defined shall have the meaning ascribed to such terms in the Plan.

See supra note 2.

In some of the analysis, Cornerstone alternatively defines a long-lasting Straddle State as a Straddle State lasting at

least 30 minutes.

See Notice of Filing of the Twelfth Amendment to the National Market System Plan to Address Extraordinary

Market Volatility by Bats BZX Exchange, Inc., Bats BYX Exchange, Inc., Bats EDGA Exchange, Inc., Bats EDGX

Exchange, Inc., Chicago Stock Exchange, Inc., Financial Industry Regulatory Authority, Inc., Investors Exchange

LLC, NASDAQ BX, Inc., NASDAQ PHLX LLC, The Nasdaq Stock Market LLC, National Stock Exchange, Inc.,

Twelfth Amendment, the Commission noted that Commission staff had conducted an analysis of

Straddle States under the Plan over the period from May 12, 2014 to August 29, 2014. 8 The staff

analysis found that 2,073,497 Straddle States occurred in Tier 2 securities over this time period

and the vast majority of Straddle States lasted less than five minutes. However, more than 4,000

Straddle States lasted between five and 30 minutes, and more than 4,000 Straddle States lasted

longer than 30 minutes. 9 Unlike Limit States, which can last up to 15 seconds, Straddle States

can last indefinitely, unless the Primary Listing Exchange declares a Trading Pause in

accordance to its policies and procedures pursuant to Section VII(A)(2) of the Plan. 10 This

means that there may be a long period of time during which one or both quotes are nonexecutable (i.e. cannot be traded against).

The Cornerstone analysis is meant to provide further details on these long-lasting

Straddle States, which could help inform future changes to the Plan. In particular, the

Cornerstone analysis investigates the frequency of occurrence of long-lasting Straddle States,

factors that influence the occurrence of long-lasting Straddle States, market activity during and

after long-lasting Straddle States, and the impact of the Tenth Amendment of the LULD Plan on

the occurrence of long-lasting Straddle States and Trading Pauses. 11

New York Stock Exchange LLC, NYSE MKT LLC, and NYSE Arca, Inc. (“Twelfth Amendment Notice of

Filing”), ”), Release No. 34-79410 (November 28, 2016).

On January 17, 2017, the Participants submitted a letter to the Commission related to the Twelfth Amendment to

the LULD Plan, which requested that the Commission modify the Twelfth Amendment to retain provisions in the

Plan related to Straddle States. The modified version of Amendment 12 was approved on January 19, 2017. See

Securities Exchange Act Release No. 79845 (January 19, 2017).

See supra note 6.

The results and additional analysis concerning Straddle States are also presented in the DERA White Paper entitled

“Limit Up-Limit Down Pilot Plan and Associated Events” by Claudia Moise and Paca Flaherty (2017). Available at

https://www.sec.gov/files/dera-luld-white-paper.pdf.

However, in the Notice of Filing of the Twelfth Amendment it states that there have not been any Trading Pauses

declared following a Straddle State. See supra note 6.

Amendment 10 revised the methodology for determining the initial Reference Price when there is not trading

volume in the opening auction. The new methodology uses the closing price of the NMS Stock on the primary

listing exchange on the previous trading day, or if no such closing price exists, the last sale on the primary listing

exchange, as the initial Reference Price, instead of the midpoint of the opening auction. The changes in Amendment

10 were made based on analysis The Participants presented in a Transmittal Letter and in The Supplemental Joint

Assessment that showed using the midpoint of the opening bid and ask when no trading occurred in the opening

auction often resulted in what the Participants believe was a skewed initial Reference Price. The Participant’s

analysis also showed that most Trading Halts occurred in securities that did not have a trade in the opening auction.

Amendment 10 was implemented on July 18, 2016. See Plan to Address Extraordinary Market Volatility as

amended by SEC Approval Order: Rel. No. 34-77679 (Amendment 10), available at

https://www.sec.gov/rules/sro/nms/2016/34-77679.pdf. Also see Limit Up – Limit Down: National Market System

Plan Assessment To Address Extraordinary Market Volatility (the “Supplemental Joint Assessment” or

“Assessment”), available at https://www.sec.gov/comments/4-631/4631-39.pdf , and see Letter from Paul Roland,

Overall, the Cornerstone analysis finds that between August 1, 2014 and December 31,

2016, on average, over 140 long-lasting Straddle States occurred each day. It also finds that they

are more likely to occur in securities with lower trading volume, higher volatility, and smaller

market capitalizations and on days when no trade occurs during the opening. DERA believes

that the methodology employed by Cornerstone in the analysis is appropriate to address the

questions presented and agrees with the conclusions drawn by Cornerstone based on that

analysis. Below, we summarize the main findings presented in the Cornerstone analysis.

Overview of Analysis:

The Cornerstone analysis examines a sample of U.S. stocks and exchange traded products

during the period from August 1, 2014 to December 31, 2016 (609 trading days). 12 The analysis

focuses on long-lasting Straddle States, which the analysis defines as Straddle States lasting at

least 5 minutes. 13 Below, we summarize the main findings from the Cornerstone Analysis

regarding the frequency of occurrence of long-lasting Straddle States, factors that influence the

occurrence of long-lasting Straddle States, market activity during and after the occurrence of

long-lasting Straddle States, and the impact of Tenth Amendment of the LULD Plan on the

occurrence of long-lasting Straddle States and Trading Halts

1. Frequency of Occurrence of Long-Lasting Straddle States

The Cornerstone analysis examines how frequently long-lasting Straddle States occur. It

finds that, each day, an average of 84 symbols experience 141 long-lasting Straddle States,

which is five times more frequent than the rate at which Moise and Flaherty (2017) estimate

Trading Pauses occur. 14,15 However, the analysis finds that this represents a small percentage of

Straddle States, with over 98% of Straddle States lasting less than 5 minutes.

The analysis examines how frequently long-lasting Straddle States reoccur within the

same trading day and finds that the majority of symbols experience only one long-lasting

Straddle State during the same trading day. 16

Principal, U.S. Equities, Nasdaq, to Brent Fields, Secretary, Commission, dated February 18, 2016. (“Transmittal

Letter”).

Most of the analysis excludes August 24, 2015. Cornerstone does provide some analysis for August 24, 2015.

See supra note 5.

The Cornerstone analysis also examines the duration of long-lasting Straddles States and finds that 53.1% of longlasting straddle states last between 5 and 30 minutes, 11.8% last between 30 minutes and 1 hour, 15.1% last between

1 and 3 hours, 6.2% last between 3 and 5 hours, and 13.8% last between 5 and 6.5 hours.

Moise and Flaherty (2017) find that an average of 29.4 Trading Pauses occur each day (2295 Trading Pauses/78

days) during the period following the Phase-in of LULD for Tier II stocks. See supra note 9.

The analysis also finds that 29.5% of symbols that experience a long-lasting Straddle State experience at least two

long-lasting Straddles States in the same day and 12.8% experience at least three long-lasting Straddle States in the

same day.

The analysis also examines how frequently long-lasting Straddle States recur across days

and finds that long-lasting Straddle States do not frequently recur in most symbols that

experience at least one long-lasting Straddle State. 17 However, the analysis also finds that a

small group of 30 symbols frequently experience long-lasting Straddle States, with these

symbols experiencing long-lasting Straddle States on more than 200 days out of the 609-day

sample period. 18

2. Factors that Influence the Occurrence of Long-Lasting Straddle States

The Cornerstone analysis examines the factors that influence the occurrence of longlasting Straddle States. In our opinion, the results suggest that long-lasting Straddle States are

mainly associated with thin trading and gaps in liquidity (i.e. instances in which the depletion of

liquidity results in large price changes that revert once liquidity is restored). The analysis finds

that long-lasting Straddle States are more likely to occur in securities with lower trading volume,

higher volatility, and smaller market capitalizations. 19 The analysis also finds that long-lasting

Straddle States are more likely to occur on days when market volatility is high, i.e. days the

market experiences relatively large positive or negative returns or days when the VIX volatility

index is high. Additionally, the majority of long-lasting Straddle States occur in securities that

did not experience a trade during the opening, with the majority beginning at either 9:30am (the

start of trading) or 9:45am (when the LULD Price Bands narrow from double width).

3. Activity During and After the Occurrence of Long-Lasting Straddle States

The Cornerstone analysis examines trading and quoting activity during long-lasting

Straddle States and finds little trading activity occurs during most long-lasting Straddle States

and finds that quotes tend to be wider during most long-lasting Straddle States than under normal

trading conditions. Specifically, the analysis finds that no trades occur in over 75% of all longlasting Straddle States and that 95% of long-lasting Straddle States have quoted spreads that are

wider than the spreads during the control period. 20

The analysis also finds that the quoted spreads are wider than the Price Bands in 39% of

long-lasting Straddle States and that both the bid and ask quotes are outside the Price Bands in

24% of long-lasting Straddle States. These results could indicate that the Price Bands are not

wide enough in these securities.

2,364 unique symbols in the sample experience at least one long-lasting Straddle State. Over half of these symbols

experience long-lasting Straddle States on 5 or fewer days out of the 609 day sample period.

The analysis also finds a group of 83 symbols experience long-lasting Straddle States for a total of 101 to 200

days out of the 609 day sample period.

These types of securities are also more likely to have wider bid-ask spreads.

The Cornerstone analysis compares the period during a long-lasting Straddles State to a control period, which is

constructed by taking the time window corresponding to the Straddle State on the five trading days prior to the

event.

The Cornerstone analysis examines how long-lasting Straddle States end. 21 The analysis

finds that while the majority of long-lasting Straddle States end with both quotes inside the price

bands, 11% of long-lasting Straddle States end by entering a Limit State and 22% immediately

reenter another Straddle State. 22,23 Moreover, although the majority of long-lasting Straddle

States end with both quotes inside the price bands, the Cornerstone analysis finds that 54% of

long-lasting Straddle States eventually reenter another Straddle State on the same day. 24

4. Impact of the Tenth Amendment of the LULD Plan on Long-Lasting Straddle States and

Trading Halts

The Cornerstone analysis examines the impact of the Tenth Amendment on the occurrence of

long-lasting Straddle States and Trading Pauses. 25 The results of the Cornerstone analysis

suggest that, after the implementation of Amendment 10, long-lasting Straddle States are

substituting for some situations that previously would have resulted in a Trading Pause.

Specifically, the analysis finds that Trading Pauses are less likely to occur and long-lasting

Straddle States are more likely to occur in securities with no opening trade after the

implementation of the Tenth Amendment. 26

According to the Plan, the primary listing exchange has the discretion to declare a Trading Pause in order to

resolve a Straddle State. However, as discussed in the Notice of Filing of the Twelfth Amendment to the LULD

Plan, a primary listing exchange has never declared a Trading Pause while a security was in a Straddle State. See

supra note 6.

Of the 11% of long-lasting Straddle States that enter a Limit State, 7% end in a Trading Pauses.

Of the 22% of long-lasting Straddle States that immediately reenter another Straddle State, 7% immediately

reenter another long-lasting Straddle State and 15% immediately reenter another Straddle-State lasting less than five

minutes.

Specifically, the analysis finds that 54.4% of long-lasting Straddle States are eventually followed by another

Straddle State on the same trading day. Among these follow-up Straddle States, 32.4% are long-lasting Straddle

States and 67.6% last less than five minutes.

See supra note Error! Bookmark not defined.

The Cornerstone analysis also finds that the majority of long-lasting Straddle States occur in securities that did not

experience a trade during the opening. See Section III of the Cornerstone analysis.

M E MOR A N D UM

DATE:

Friday, July 21, 2017

TO:

Amy Edwards, Office of Markets, Division of Economic and Risk Analysis

FROM:

Stewart Mayhew, Cornerstone Research

RE:

Long-Lasting Straddle States

This memo summarizes research we have conducted pursuant to your request

under Contract SECHQ1-16-C-0024, related to straddle states1 under the Limit

Up/Limit Down (LULD) rules.

Specifically, this memo provides a summary overview of economic analysis

investigating the frequency of “long lasting” straddle states (defined as LULD

straddle states lasting at least five minutes or at least 30 minutes), the

characteristics of the securities experiencing such long-lasting straddle states, and

other aspects of market conditions surrounding long-lasting straddle states. The

memo also summarizes research examining the impact of Amendment 10

(implemented on July 18, 2016) on the occurrence of these long-lasting straddle

states. In conjunction with this memo, we have provided a slide deck with tables

and charts.

This research was performed by the staff of Cornerstone Research under my

direction, with additional guidance and oversight from subject matter expert Dr.

Timothy McCormick.

The analysis described below and in the accompanying slide deck is based on 609

trading days from August 1, 2014 to December 31, 2016. Most of the analysis

described below excludes August 24, 2015, a date that experienced an

extraordinary number of LULD events. Some analysis is provided for August 24,

2015. The data for this project were provided to Cornerstone Research by the staff

of the Division of Economic and Risk Analysis (DERA). This includes data on

straddle states, limit states, and LULD trading halts (as identified by the selfregulatory organizations (“SROs”), data from the Center for Research in Security

Prices (CRSP), and consolidated intraday data from the Trade and Quote (TAQ)

database.

Under the LULD rules, a straddle state occurs when the National Best Bid (Offer) is below (above) the lower

(upper) price band and the NMS Stock is not in a limit state. During a straddle state, there are some prices within the

quoted spread at which trades can be executed, but there are prices within the quoted spread at which trades cannot

be executed because they fall outside of the price bands.

The results of this research can be summarized as follows. The structure of this

outline refers to corresponding sections of the accompanying slide deck.

Section I: Univariate Distributions of Long-Lasting Straddle States

August 24, 2015 was an outlier with almost 10 times more straddle events than

usual. There were 1,384 straddle events lasting five minutes or more on August

24, 2015, occurring on 768 different symbols.

Based on analysis of the remainder of the sample, excluding August 24, 2015:



98.6% of all straddle states lasted less than five minutes.2



On average, there were 141 straddle events per day lasting at least five

minutes, and 66 straddle events lasting at least 30 minutes (slide 6).



The majority of long-lasting straddle states begin at the start of trading or

at 9:45 AM, the moment when the LULD bounds become narrower (slides

7–8).



Graphs are provided summarizing the average number of long-lasting

straddle events over the course of the trading day (slides 9–10).

Section II: Comparing Five- and 30-Minute Long-Lasting Straddle States

This section contains graphs showing cumulative time in straddle state conditional

on the length of the longest straddle of the day. Separate graphs are provided for

symbol-days on which the longest straddle state was less than five minutes,

between five and 30 minutes, and greater than 30 minutes (slides 13–15).

Instances in which a security persistently flickered in and out of a straddle state—

wherein the security never experienced a straddle state lasting longer than five

minutes, but was in a straddle state for a total of more than 30 minutes of the

day—are analyzed further. There are 156 such symbol-days in the sample, which

corresponds to this event occurring roughly once every four days (slides 12–13).



This occurred on 51 unique symbols, with one symbol flickering in and

out of straddle states on 16 different dates.



Of these flickering symbol-days, 33 symbol-days were in a straddle state

for at least five hours of the trading day.

There were 6,215,889 straddle states in the sample and 6,130,081 lasted less than five minutes.

Section III: Factors That Influence the Occurrence of Straddle States

The following factors were evaluated with respect to their likelihood of

influencing the occurrence of straddle states:



Market index returns: The incidence of long-lasting straddle events

appears to be slightly higher on days when the S&P 500 or the Russell

2000 experiences a relatively large movement up or down (slides 18–24)

or when the VIX volatility index is high (slides 25–26).



Breakdown by primary exchange listing and type of security: NASDAQlisted securities were more likely to experience a long-lasting straddle

state than securities listed on other exchanges (slides 27–29).



Whether the security was in Tier 1 or Tier 2 (slide 30).



Price level of prior day’s close, by price band regime (slide 31).



Existence of an opening trade (slide 32).



Trading volume: Long-lasting straddle states are more prevalent for lowervolume securities (slides 33–37).



Market capitalization (for common stock only): Long-lasting straddle

states are more prevalent for smaller issuers (slides 38–42).

The following logistic panel regression was used to estimate the significance of

various factors contributing to the likelihood of a long-lasting straddle event:

The dependent variable

, , has a value of one if there

was a long-lasting straddle state on a given symbol-day and zero otherwise. The

independent variables, , , include the following:



Trading volume on that day (measured in logarithm).

Indicator variable with a value of one if the stock is in Tier 1.

Indicator variable with a value of one if the prior day close was $3 or below.

Indicator variable with a value of one if the prior day close was below $0.75.

Indicator variable with a value of one if there was no opening trade.

Abnormal volume measured as the quintile rank of the current day’s volume

among the prior 30 days.



Stock volatility measured by the standard deviation of returns over the prior

30 days.

The results indicate that stocks with higher return volatility, higher abnormal

trading volume, and stocks that lack an opening trade were associated with a

higher likelihood of experiencing a long-lasting straddle state. Tier 1 stocks,

stocks with a prior day’s closing price below $3 and below $0.75, respectively,

and stocks with higher trading volume were associated with a lower likelihood of

experiencing a long-lasting straddle state. These results hold for estimations with

and without time fixed effects (slide 44).

Section IV: Activity During and After Long-Lasting Straddle States

Appearing first in this section is analysis on the propensity of straddle states to

recur across days. Thirty symbols experienced at least one long-lasting straddle

state on more than 200 days in the sample of 609 trading days. On average, 84

symbols per day experienced a long-lasting straddle event, although this number

varies considerably across the sample period. Of the symbols experience a longstraddle event, typically 10–25 symbols per day had a long-lasting straddle state

in each of the five previous trading days and 10–25 had a long-lasting straddle

state in none of the five previous trading days (slides 47–49).

The majority of symbols that experience a long-lasting straddle state on a given

day experience only one long-lasting straddle event on that day. In addition, 22%

of long-lasting straddle states were resolved by immediately entering into another

straddle state, while 10.8% were resolved by immediately entering into a limit

state (slides 51–53).

Next, the quoting and trading activity during the long-lasting straddle state is

analyzed. For several analyses in this section, a control period is constructed by

taking the time window corresponding to the straddle event on the five trading

days prior to the straddle event.3 The results show the following:



Quoting activity during the straddle event: Quoted relative spreads during

the straddle state were wider relative to the control period for 95% of

straddle events. For straddle events that occurred while the bands were

single wide, the time-weighted relative quoted spread was between 8%

and 16% more than half of the time (slides 55–59).

Straddle events were included in the sample for these analyses if there were five trading days to use as a control

period prior to the day of the straddle event and if there were no long-lasting straddle states during the control

period.



Trading volume during the straddle event: There were no shares traded

during 75.3% of long-lasting straddle states. Overall, volume during the

long-lasting straddle state was typically lower than the volume during the

same time interval on the five previous trading days (slides 60–61).



Breakdown of straddle state by straddle type: 41% of straddle states were

“limit up straddle states,” 32% were “limit down straddle states,” 24%

were instances in which the bid was below and the ask was above the price

bands, and 3% fell into more than one of these categories over the course

of the straddle state (slide 62).

Section V: Impact of Amendment 10 on Long-Lasting Straddle States

Amendment 10 was implemented on July 18, 2016 and was targeted at reducing

the number of trading halts on stocks that did not have an opening trade. The

amendment stated that stocks without any trading volume in the opening auction

would use the prior day’s closing price as the first reference price instead of the

midpoint of the opening quote.

The impact of Amendment 10 is analyzed by extending the regression

methodology outlined in Section III above. In particular, the specification adds

indicator variables for whether or not the stock had an opening trade and whether

the date was before or after Amendment 10 was implemented, as well as the

interaction of the two. The impact of Amendment 10 on the likelihood of

experiencing a trading halt is tested first.

∗

The independent variables include the following:



denotes an indicator variable with a value of one after

implementation.



, denotes an indicator variable with a value of one if

there was no opening trade.

∗

, denotes the interaction of the

two indicator variables described above.

, is the set of control variables described in Section III.

This analysis shows that the odds ratio of a stock experiencing a trading halt

conditional on that stock having no opening trade is approximately five to seven

times lower after the implementation of Amendment 10 (slide 65).

Next, the same model is used to test the impact of Amendment 10 on the

likelihood of experiencing a long-lasting straddle.

∗

This analysis shows that the odds ratio of a stock experiencing a long-lasting

straddle state conditional on that stock having no opening trade is approximately

1.2 times higher after the implementation of Amendment 10 (slide 67).

In summary, the results indicate that the implementation of Amendment 10

appears to have significantly reduced the likelihood of a trading halt, but

increased the likelihood of long-lasting straddle events.

Analysis of Long-Lasting Straddle States

July 21, 2017

Page 1

Analysis of Long-Lasting Straddle States

Table of Contents

• Section I: Univariate Distributions of Long-Lasting Straddle States

• Section II: Comparing Five– and 30–Minute Long-Lasting Straddle States

• Section III: Factors That Influence the Occurrence of Straddle States

• Section IV: Activity During and After Long-Lasting Straddle States

• Section V: Impact of Amendment 10 on Long-Lasting Straddle States

Page 2

Analysis of Long-Lasting Straddle States

Section I: Univariate Distributions of Long-Lasting Straddle States

Page 3

Analysis of Long-Lasting Straddle States

Univariate Distributions of Long-Lasting Straddle States

• Sample Period: August 2014 – December 2016

• 98.6% of straddle states lasted less than five minutes.

• Statistics using five–minute threshold for long-lasting straddle states:

– Average number of straddle events: 141 per day

– Average number of symbols that experienced at least one straddle event: 84 per day

• Statistics using 30–minute threshold for long-lasting straddle states:

– Average number of straddle events: 66 per day

– Average number of symbols that experienced at least one straddle event: 53 per day

• Majority of long-lasting straddle states begin at the start of trading or 9:45 AM.

• Majority of analyses will be done using five–minute threshold.

• August 24, 2015

– There were 1,384 straddle events lasting five minutes or more, or almost 10 times

more than usual. The 1,384 straddle events occurred on 768 symbols.

– Due to the unusual nature of this day, it was excluded from most of the analysis.

Page 4

Distribution of Straddle State Duration

8/1/14 – 12/31/16[1]

Number of Straddle

States per Day[2]

53.9

53.7

46.4

Number of Symbols in a Straddle State:

From Open to Close: 6.2 per day

While Bands Were Single-Wide: 4.3 per day

21.1

19.4

16.6

14.1

7.2

4.7

4.1

3 hrs - 4 hrs

4 hrs - 5 hrs

1 min - 2 min

2 min - 5 min

Source: SEC Straddle Data

5 min - 15 min 15 min - 30 min 30 min - 1 hr

1 hr - 2 hrs

2 hrs - 3 hrs

5 hrs - 6.5 hrs

Straddle Duration[3]

Note:

[1] There are 609 trading days in the sample, excluding August 24, 2015.

[2] This figure depicts the distribution of straddle state durations for straddle states lasting longer than one minute. The full sample also includes 6,069,029

straddle states that lasted less than one minute, or approximately 10,000 per day.

[3] Straddle duration represents the total amount of time a straddle state lasted.

Page 5

Daily Counts of Long-Lasting Straddle States

August 2014 – December 2016

Number of

Straddle States

300

250

200

150

100

Straddle States Lasting At Least 5 Minutes

Straddles States Lasting At Least 30 Minutes

Source: SEC Straddle Data

Note: The straddle states that occurred on August 24, 2015 have been excluded.

Page 6

'LVWULEXWLRQRI/RQJ/DVWLQJ6WUDGGOH6WDWHV3DUWLWLRQHGE\6WDUW7LPH

6WUDGGOH6WDWHV/DVWLQJ$W/HDVWFive0LQXWHV

$XJXVW±'HFHPEHU

$YHUDJH1XPEHURI

6WUDGGOHVSHU'D\

9:45 AM

6RXUFH6(&6WUDGGOH'DWD

1RWH7KHVWUDGGOHVWDWHVWKDWRFFXUUHGRQ$XJXVWKDYHEHHQH[FOXGHG

Page 7

'LVWULEXWLRQRI/RQJ/DVWLQJ6WUDGGOH6WDWHV3DUWLWLRQHGE\6WDUW7LPH

6WUDGGOH6WDWHV/DVWLQJ$W/HDVW0LQXWHV

$XJXVW±'HFHPEHU

$YHUDJH1XPEHURI

6WUDGGOHVSHU'D\

9:45 AM

6RXUFH6(&6WUDGGOH'DWD

1RWH7KHVWUDGGOHVWDWHVWKDWRFFXUUHGRQ$XJXVWKDYHEHHQH[FOXGHG

Page 8

$YHUDJH1XPEHURI6\PEROVLQD/RQJ/DVWLQJ6WUDGGOH6WDWHDW2QH-0LQXWH,QWHUYDOV7KURXJKRXWWKH'D\

6WUDGGOH6WDWHV/DVWLQJ$W/HDVWFive0LQXWHV

$XJXVW±'HFHPEHU

$YHUDJH1XPEHURI

6\PEROVSHU'D\

9:45 AM

3:35 PM

6RXUFH6(&6WUDGGOH'DWD

1RWH7KHVWUDGGOHVWDWHVWKDWRFFXUUHGRQ$XJXVWKDYHEHHQH[FOXGHG

Page 9

$YHUDJH1XPEHURI6\PEROVLQD/RQJ/DVWLQJ6WUDGGOH6WDWHDW2QH-0LQXWH,QWHUYDOV7KURXJKRXWWKH'D\

6WUDGGOH6WDWHV/DVWLQJ$W/HDVW0LQXWHV

$XJXVW±'HFHPEHU

$YHUDJH1XPEHURI

6\PEROVSHU'D\

9:45 AM

3:35 PM

6RXUFH6(&6WUDGGOH'DWD

1RWH7KHVWUDGGOHVWDWHVWKDWRFFXUUHGRQ$XJXVWKDYHEHHQH[FOXGHG

Page 10

Analysis of Long-Lasting Straddle States

Section II: Comparing Five– and Thirty– Minute Long-Lasting Straddle States

Page 11

Analysis of Long-Lasting Straddle States

Defining a Long-Lasting Straddle State

• 156 symbol-days (or one symbol every four days) persistently “flickered” in and out of a

straddle state, wherein they experienced no straddle states lasting longer than five minutes,

but were in a straddle state for more than 30 minutes of the day.

– These flickering symbol-days were spread across 51 unique symbols, with one symbol

experiencing flickering on 16 different dates.

– Flickering occurred on 85 unique dates, with two dates having a maximum of eight

different symbols experiencing flickering.

• Of these “flickering” symbol-days, 33 symbol-days (or one symbol every four weeks) were

in a straddle state for at least five hours of the day.

– These flickering symbol-days were spread across 11 unique symbols, with one symbol

experiencing flickering on 13 different dates.

– Flickering occurred on 30 unique dates, with three dates having a maximum of two

different symbols experiencing flickering.

Page 12

Average Number of

Symbols per Day[2]

Cumulative Time in Straddle State Conditional on the Longest

Straddle State Lasting Less Than Five Minutes

8/1/14 – 12/31/16[1]

16.0

11.0

2.7

0.2

0.0

0.1

1 hr - 2 hrs

2 hrs - 3 hrs

3 hrs - 4 hrs

4 hrs - 5 hrs

5 hrs - 6.5 hrs

1 min - 2 min

2 min - 5 min

5 min - 15 min 15 min - 30 min 30 min - 1 hr

Straddle Duration[3]

Source: SEC Straddle Data

Note:

[1] There are 609 trading days in the sample, excluding August 24, 2015.

[2] This figure does not include the 50,562 symbol-days (approximately 83 symbols per day) for which the cumulative time in straddle state was less than one

minute.

[3] Straddle duration represents the total amount of time a symbol spent in a straddle state on a given day. Symbol-days on which the longest straddle states

lasted less than five minutes are included.

Page 13

Average Number of

Symbols per Day

Cumulative Time in Straddle State Conditional on the Longest

Straddle State Lasting Between Five and Thirty Minutes

8/1/14 – 12/31/16[1]

16.7

10.5

2.4

0.0

1 min - 2 min

2 min - 5 min

0.4

0.1

0.0

0.1

0.2

1 hr - 2 hrs

2 hrs - 3 hrs

3 hrs - 4 hrs

4 hrs - 5 hrs

5 hrs - 6.5 hrs

5 min - 15 min 15 min - 30 min 30 min - 1 hr

Straddle Duration[2]

Source: SEC Straddle Data

Note:

[1] There are 609 trading days in the sample, excluding August 24, 2015.

[2] Straddle duration represents the total amount of time a symbol spent in a straddle state on a given day. Symbol-days on which the longest straddle states

lasted greater than or equal to five minutes but less than 30 minutes are included.

Page 14

Cumulative Time in Straddle State Conditional on the Longest

Straddle State Lasting More Than 30 Minutes

Average Number of

Symbols per Day

8/1/14 – 12/31/16[1]

24.8

8.8

7.2

5.2

3.9

3.5

0.0

1 min - 2 min

2 min - 5 min 5 min - 15 min 15 min - 30 min 30 min - 1 hr

Source: SEC Straddle Data

1 hr - 2 hrs

2 hrs - 3 hrs

3 hrs - 4 hrs

4 hrs - 5 hrs

5 hrs - 6.5 hrs

Straddle Duration[2]

Note:

[1] There are 609 trading days in the sample, excluding August 24, 2015.

[2] Straddle duration represents the total amount of time a symbol spent in a straddle state on a given day. Symbol-days on which the longest straddle

state lasted at least 30 minutes are included.

Page 15

Analysis of Long-Lasting Straddle States

Section III: Factors That Influence the Occurrence of Straddle States

Page 16

Analysis of Long-Lasting Straddle States

Factors That Influence the Occurrence of Straddle States

• Frequency of long-lasting straddle states shown by the following day characteristics:

– S&P 500 Return

– Russel 2000 Return

– VIX Level

• Frequency of long-lasting straddle states shown by the following stock characteristics[1]:

– Exchange

– Type of security

– Tier 1 or Tier 2 stock

– Price level of prior day’s close

– Whether or not there was an opening trade

– Volume

– Market capitalization (includes only common stock)

[1] Most of these analyses only include stocks that are included in the CRSP database. CRSP does not have data on BATS listed securities, ETNs, preferred stock, and units.

Page 17

Average Number of Symbols Experiencing at Least One Long-Lasting

Average Number of

Straddle State Partitioned by S&P 500 Return[1]

Symbols per Day

120

8/1/14 – 12/31/16[2]

100

(-5.2% to -1.0%) (-1.0% to -0.5%) (-0.5% to -0.2%) (-0.2% to -0.1%) (-0.1% to 0.0%) (0.0% to 0.2%) (0.2% to 0.4%) (0.4% to 0.6%) (0.6% to 1.2%) (1.2% to 3.9%)

S&P 500 Daily Return Deciles (Bounds)

Source: SEC Straddle Data; Bloomberg S&P 500 Total Return Index Data

Note:

[1] This graph shows the average number of unique symbols with a long-lasting straddle (a straddle state lasting at least five minutes) occurring on days partitioned by

the daily dividend-adjusted return of the S&P 500 Index.

[2] There are 609 trading days in the sample, excluding August 24, 2015.

Page 18

Average Number of Long-Lasting Straddle States

Partitioned by S&P 500 Return[1]

8/1/14 – 12/31/16[2]

Average Number of

Straddles per Day

180

169

160

140

137

129

132

135

139

140

127

120

100

(-5.2% to -1.0%) (-1.0% to -0.5%) (-0.5% to -0.2%) (-0.2% to -0.1%) (-0.1% to 0.0%) (0.0% to 0.2%) (0.2% to 0.4%) (0.4% to 0.6%) (0.6% to 1.2%) (1.2% to 3.9%)

S&P 500 Daily Return Deciles (Bounds)

Source: SEC Straddle Data; Bloomberg S&P 500 Total Return Index Data

Note:

[1] This graph shows the average frequency of long-lasting straddle states (straddle states lasting at least five minutes) occurring on days partitioned by the daily

dividend-adjusted return of the S&P 500 Index.

[2] There are 609 trading days in the sample, excluding August 24, 2015.

Page 19

Average Number of Symbols Experiencing at Least One Long-Lasting

Straddle State Partitioned by S&P 500 Return

Average Number of

[1]

Symbols per Day

Highest

10%

Daily

Returns

140

8/1/14 – 12/31/16[2]

120

115

114

107

100

(1.2% to 1.2%) (1.2% to 1.3%) (1.3% to 1.3%) (1.3% to 1.4%) (1.4% to 1.5%) (1.5% to 1.6%) (1.6% to 1.7%) (1.7% to 1.9%) (1.9% to 2.2%) (2.2% to 3.9%)

S&P 500 Daily Return Percentiles (Bounds)

Source: SEC Straddle Data; Bloomberg S&P 500 Total Return Index Data

Note:

[1] This graph shows the average number of unique symbols with a long-lasting straddle (a straddle state lasting at least five minutes) occurring on days partitioned by

the daily dividend-adjusted return of the S&P 500 Index.

[2] There are 609 trading days in the full sample, excluding August 24, 2015. The 61 trading days included here are those with daily index returns in the highest 10% of

the full sample.

Page 20

Average Number of Symbols Experiencing at Least One Long-Lasting

Straddle State Partitioned by S&P 500 Return

Average Number

of Symbols per

[1]

Lowest

10%

Daily

Returns

140

8/1/14 – 12/31/16[2]

120

119

103

100

101

(-5.2% to -2.5%) (-2.5% to -2.0%) (-2.0% to -1.7%) (-1.7% to -1.6%) (-1.6% to -1.5%) (-1.5% to -1.4%) (-1.4% to -1.3%) (-1.3% to -1.2%) (-1.2% to -1.1%) (-1.1% to -1.0%)

S&P 500 Daily Return Percentiles (Bounds)

Source: SEC Straddle Data; Bloomberg S&P 500 Total Return Index Data

Note:

[1] This graph shows the average number of unique symbols with a long-lasting straddle (a straddle state lasting at least five minutes) occurring on days partitioned by

the daily dividend-adjusted return of the S&P 500 Index.

[2] There are 609 trading days in the full sample, excluding August 24, 2015. The 60 trading days included here are those with daily index returns in the lowest 10% of

the full sample.

Page 21

Average Number of Symbols Experiencing at Least One Long-Lasting

Straddle State Partitioned by Russell 2000 Return[1]

Average Number of

Symbols per Day

8/1/14 – 12/31/16[2]

120

100

(-4.6% to -1.4%) (-1.4% to -0.8%) (-0.8% to -0.4%) (-0.4% to -0.2%) (-0.2% to 0.1%) (0.1% to 0.3%) (0.3% to 0.6%) (0.6% to 0.9%) (0.9% to 1.4%) (1.4% to 3.2%)

Russell 2000 Daily Return Deciles (Bounds)

Source: SEC Straddle Data; Bloomberg Russell 2000 Total Return Index Data

Note:

[1] This graph shows the average number of unique symbols with a long-lasting straddle (a straddle state lasting at least five minutes) occurring on days partitioned by

the daily dividend-adjusted return of the Russell 2000 Index.

[2] There are 609 trading days in the sample, excluding August 24, 2015.

Page 22

Average Number of Symbols Experiencing at Least One Long-Lasting

Straddle State Partitioned by Russell 2000 Return

Average Number of

Symbols per Day

Highest 10% of Daily Returns[1]

120

8/1/14 – 12/31/16[2]

109

107

103

100

(1.4% to 1.4%) (1.4% to 1.5%) (1.5% to 1.6%) (1.6% to 1.7%) (1.7% to 1.8%) (1.8% to 1.9%) (1.9% to 2.2%) (2.2% to 2.3%) (2.3% to 2.6%) (2.6% to 3.2%)

Russell 2000 Daily Return Percentiles (Bounds)

Source: SEC Straddle Data; Bloomberg Russell 2000 Total Return Index Data

Note:

[1] This graph shows the average number of unique symbols with a long-lasting straddle (a straddle state lasting at least five minutes) occurring on days partitioned by

the daily dividend-adjusted return of the Russell 2000 Index.

[2] There are 609 trading days in the full sample, excluding August 24, 2015. The 61 trading days included here are those with daily index returns in the highest 10% of

the full sample.

Page 23

Average Number of Symbols Experiencing at Least One Long-Lasting

Straddle State Partitioned by Russell 2000 Return

Average Number of

Symbols per Day

Lowest 10% of Daily Returns[1]

120

8/1/14 – 12/31/16[2]

111

107

100

(-2.8% to -4.6%) (-2.5% to -2.8%) (-2.2% to -2.5%) (-1.9% to -2.2%) (-1.7% to -1.9%) (-1.6% to -1.7%) (-1.5% to -1.6%) (-1.5% to -1.5%) (-1.4% to -1.5%) (-1.4% to -1.4%)

Russell 2000 Daily Return Percentiles (Bounds)

Source: SEC Straddle Data; Bloomberg Russell 2000 Total Return Index Data

Note:

[1] This graph shows the average number of unique symbols with a long-lasting straddle (a straddle state lasting at least five minutes) occurring on days partitioned by

the daily dividend-adjusted return of the Russell 2000 Index.

[2] There are 609 trading days in the full sample, excluding August 24, 2015. The 60 trading days included here are those with daily index returns in the lowest 10% of

the full sample.

Page 24

Average Number of Symbols Experiencing at Least One Long-Lasting

Average Number of

Straddle State Partitioned by VIX Value[1]

Symbols per Day

8/1/14 – 12/31/16[2]

120

113

100

(14.7 to 15.5)

(15.5 to 16.7)

(11.3 to 12.3)

(12.3 to 12.9)

(12.9 to 13.5)

(13.5 to 14.0)

(14.0 to 14.7)

(16.7 to 18.8)

(18.8 to 22.0)

(22.0 to 36.0)

Daily VIX Value Deciles (Bounds)

Source: SEC Straddle Data; Bloomberg VIX Data

Note:

[1] This graph shows the average number of unique symbols with a long-lasting straddle (a straddle state lasting at least five minutes) occurring on days partitioned by

the VIX Value.

[2] There are 609 trading days in the sample, excluding August 24, 2015.

Page 25

Average Number of Symbols Experiencing at Least One Long-Lasting

Straddle State Partitioned by VIX Value

Average Number of

Symbols per Day

Highest 10% of Daily Values[1]

140

8/1/14 – 12/31/16[2]

128

121

120

116

109

104

105

106

107

(22.5 to 23.4)

(23.4 to 24.1)

(24.1 to 24.5)

(24.5 to 25.4)

119

(26.7 to 27.8)

(27.8 to 36.0)

100

(22.0 to 22.4)

(22.4 to 22.5)

(25.4 to 26.1)

(26.1 to 26.7)

Daily VIX Value Percentiles (Bounds)

Source: SEC Straddle Data; Bloomberg VIX Data

Note:

[1] This graph shows the average frequency of long-lasting straddle states (straddle states lasting at least five minutes) occurring on days partitioned by the daily VIX

Value.

[2] There are 609 trading days in the sample, excluding August 24, 2015. The 61 trading days included here are those with daily VIX Values in the highest 10% of the

full sample.

Page 26

Long-Lasting Straddle State Count Aggregated by CRSP

Share Code and Primary Listing Exchange[1]

8/1/14 – 12/31/16[2]

Long-Lasting Straddle State Counts

CRSP Share Code Definitions

Across All

Exchanges

Percentage of Symbol-Days That Experienced a Long-Lasting

[3]

Straddle State

Across All

Exchanges

NASDAQ

ARCA

NYSE

MKT

BATS

NASDAQ

ARCA

NYSE

MKT

BATS

Domestic Common Stock[4]

36,267

–

141

155

36,563

1.54%

0.00%

0.01%

0.11%

0.00%

0.91%

International Common Stock[5]

7,747

–

7,763

3.11%

0.00%

0.01%

0.02%

0.00%

1.46%

American Depositary Receipts[6]

4,009

–

104

–

4,115

3.88%

0.00%

0.06%

0.04%

0.00%

1.26%

Exchange Traded Funds[7]

1,114

2,353

–

3,540

0.65%

0.24%

0.00%

0.21%

0.29%

376

463

–

890

0.52%

1.59%

0.00%

0.04%

0.00%

0.10%

49,513

2,816

278

191

52,871

1.66%

0.28%

0.01%

0.08%

0.21%

0.72%

Other[8]

Total Across All CRSP Share

Codes

Source: SEC Straddle Data; CRSP Data; CRSP Data Guide; TAQ Data

Note:

[1] Long-lasting straddle states are defined as a straddle state lasting at least five minutes. The sample is limited to the universe of stocks included in CRSP and the universe of stocks traded on BATS.

[2] There are 609 trading days in the sample, excluding August 24, 2015.

[3] Each cell in this panel of the table shows the percent of days on which symbols trading on the given exchange with the given share code experienced at least one long-lasting straddle state. For example,

the cell that corresponds to share code 11 and NASDAQ represents the number of days on which symbols with share code 11 trading on NASDAQ experienced a straddle state, divided by the total number of

symbol-days in CRSP with share code 11 that trade on NASDAQ.

[4] Corresponds to CRSP share code 11.

[5] Corresponds to CRSP share code 12.

[6] Corresponds to CRSP share code 31.

[7] Corresponds to CRSP share code 73.

[8] Corresponds to all other CRSP share codes and includes securities such as closed-end funds and REITs.

Page 27

Long-Lasting Straddle State Count Aggregated by CRSP

[1]

Share Code and Primary Listing Exchange

8/24/15

Long-Lasting Straddle State Counts

CRSP Share Code Definitions

NASDAQ

ARCA

NYSE

Domestic Common Stock[3]

227

–

260

4.63%

0.00%

2.27%

0.49%

0.00%

3.57%

International Common Stock[4]

–

10.68%

0.00%

1.81%

0.00%

5.73%

American Depositary Receipts[5]

–

7.34%

0.00%

2.20%

0.00%

3.82%

Exchange Traded Funds[6]

760

–

852

26.59%

30.77%

0.00%

17.14%

29.99%

Other[7]

–

8.33%

9.38%

2.33%

0.00%

2.82%

380

765

1,215

6.68%

30.27%

2.25%

0.28%

17.14%

9.28%

Total Across All CRSP Share

Codes

MKT

BATS

Across All

Exchanges

Percentage of Symbol-Days That Experienced a Long-Lasting

[2]

Straddle State

Across All

Exchanges

NASDAQ

ARCA

NYSE

MKT

BATS

Source: SEC Straddle Data; CRSP Data; CRSP Data Guide; TAQ Data

Note:

[2] Each cell in this panel of the table shows the percent of days on which symbols trading on the given exchange with the given share code experienced at least one long-lasting straddle state. For example,

the cell that corresponds to share code 11 and NASDAQ represents the number of days on which symbols with share code 11 trading on NASDAQ experienced a straddle state, divided by the total number of

symbol-days in CRSP with share code 11 that trade on NASDAQ.

[3] Corresponds to CRSP share code 11.

[4] Corresponds to CRSP share code 12.

[5] Corresponds to CRSP share code 31.

[6] Corresponds to CRSP share code 73.

[7] Corresponds to all other CRSP share codes and includes securities such as closed-end funds and REITs.

Page 28

Long-Lasting Straddle State Count Aggregated by CRSP

Share Code and Primary Listing Exchange[1]

8/1/14 – 12/31/16[2]

CRSP Share Code Definitions

Long-Lasting Straddle State Counts for Symbols That Experienced a

Long-Lasting Straddle on More Than 100 Days

Across All

Exchanges

NASDAQ

ARCA

NYSE

MKT

BATS

Long-Lasting Straddle State Counts for Symbols That Experienced a

Long-Lasting Straddle on More Than 200 Days

Across All

Exchanges

NASDAQ

ARCA

NYSE

MKT

BATS

Domestic Common Stock[3]

15,351

–

15,351

7,392

–

7,392

International Common Stock[4]

3,997

–

3,997

1,951

–

1,951

2,170

–

2,170

1,213

–

1,213

–

21,518

–

21,518

10,556

–

10,556

American Depositary Receipts

Exchange Traded Funds

[5]

[6]

Other[7]

Total Across All CRSP Share

Codes

Source: SEC Straddle Data; CRSP Data; CRSP Data Guide; TAQ Data

Note:

[2] There are 609 trading days in the sample, excluding August 24, 2015.

[3] Corresponds to CRSP share code 11.

[4] Corresponds to CRSP share code 12.

[5] Corresponds to CRSP share code 31.

[6] Corresponds to CRSP share code 73.

[7] Corresponds to all other CRSP share codes and includes securities such as closed-end funds and REITs.

Page 29

Breakdown of Symbol-Days with at Least One Long-Lasting

Straddle State by Tier[1]

8/1/14 – 12/31/16[2]

0.3%

Tier 1

Tier 2

99.7%

Source: CRSP; SEC Price Bands Data

Note:

[1] Long-lasting straddle states are defined as a straddle state lasting at least five minutes. The sample is limited to the universe of symbols in CRSP.

[2] The straddle states that occurred on August 24, 2015 have been excluded.

Page 30

Breakdown of Symbol-Days with at Least One Long-Lasting

Straddle State by Closing Price on the Prior Day[1]

8/1/14 – 12/31/16[2]

1.5%

6.4%

Less than $0.75

Between $0.75 and

$3.00 (inclusive)

Greater than $3.00

92.1%

Source: CRSP

Note:

[1] Long-lasting straddle states are defined as a straddle state lasting at least five minutes. The sample is limited to the universe of symbols in CRSP.

[2] The straddle states that occurred on August 24, 2015 have been excluded.

Page 31

Breakdown of Symbol-Days with at Least One Long-Lasting

Straddle State by Whether or Not There Was an Opening Trade[1][2]

8/1/14 – 12/31/16[3]

36.9%

Yes

63.1%

Source: CRSP; SEC Price Bands Data; TAQ Data

Note:

[1] Long-lasting straddle states are defined as a straddle state lasting at least five minutes. The sample is limited to the universe of symbols in CRSP.

[2] For straddle states that began at 9:30 AM, 32% of the symbols had an opening trade. For straddle states that began at 9:45 AM, 43% of the symbols had

an opening trade.

[3] The straddle states that occurred on August 24, 2015 have been excluded.

Page 32

Average Number of Symbols Experiencing at Least One Long-Lasting

Straddle State Partitioned by Monthly Trading Volume[1]

Average Number of

8/1/14 – 12/31/16[2]

Symbols per Day

29.1

Volume Decile Breakpoints[3]:

Minimum: 0

First Decile: 4,703

Second Decile: 17,644

Third Decile: 40,445

Fourth Decile: 79,475

Fifth Decile: 143,767

Sixth Decile: 253,120

Seventh Decile: 455,792

Eighth Decile: 873,377

Ninth Decile: 2,039,576

Maximum: 122,069,228

Number of Symbols that experience at least

one long-lasting straddle state and have zero

trading volume: 8.5 per day

14.5

5.6

1.9

0.7

0.2

0.1

0.0

Trading Volume Decile[4]

Source: SEC Straddle Data; CRSP Data

Note:

[1] Monthly trading volume is calculated by summing the daily trading volume of a given symbol within the month.

[2] There are 609 trading days in the sample, excluding August 24, 2015.

[3] Symbol-days were assigned to trading volume deciles based on each individual month of trading. Therefore, a given symbol's percentile can vary from month to

month.

[4] The volume decile breakpoints are reported in shares per day and are averaged across the months of the sample period.

Page 33

Average Number of Symbols Experiencing at Least One Long-Lasting

Straddle State Partitioned by Monthly Trading Volume[1]

Lowest 20 Percentiles

Average Number of

Symbols per Day

8/1/14 – 12/31/16[2]

5.0

3.2

3.1

2.8

2.7

2.5

2.0

1.8

1.7

1.5

1.3

1.1

1.2

0.9

Trading Volume Percentile[3]

Source: SEC Straddle Data; CRSP Data

Note:

[1] Monthly trading volume is calculated by summing the daily trading volume of a given symbol within the month.

[2] There are 609 trading days in the sample, excluding August 24, 2015.

[3] Symbol-days were assigned to trading volume deciles based on each individual month of trading. Therefore, a given symbol's percentile can vary from month to

month.

Page 34

Number of Symbols Experiencing at Least One Long-Lasting

Straddle State Partitioned by Monthly Trading Volume[1]

8/24/15

Number of Symbols

300

250

240

200

155

150

100

Trading Volume Decile[2]

Source: SEC Straddle Data; CRSP Data

Note:

[1] Monthly trading volume is calculated by summing the daily trading volume of a given symbol within the month.

[2] Symbol-days were assigned to market capitalization deciles based on trading in August 2015.

Page 35

Average Number of Symbols Experiencing at Least One Long-Lasting

Straddle State Partitioned by Monthly Trading Volume[1]

Average Number of

8/1/14 – 12/31/16[2]

Symbols per Day

29.1

All Symbols

Symbols That Experienced Long-Lasting Straddle States on 101 – 200 Days

Symbols That Experienced Long-Lasting Straddle States More Than 200 Days

14.5

6.8

6.1

5.6

2.7

1.9

0.7

0.6

0.1

0.1 0.0

0.7

0.0 0.0

0.2 0.0 0.0

0.1 0.0 0.0

0.0 0.0 0.0

Trading Volume Decile[3]

Source: SEC Straddle Data; CRSP Data

Note:

[1] Monthly trading volume is calculated by summing the daily trading volume of a given symbol within the month.

[2] There are 609 trading days in the sample, excluding August 24, 2015.

[3] Symbol-days were assigned to trading volume deciles based on each individual month of trading. Therefore, a given symbol's percentile can vary from month to

month.

Page 36

Percent of Long-Lasting Straddle States per Trading Day by CRSP Volume Deciles

August 2014 – December 2016

% of Straddles

Per Day

100

Second Decile

First Decile

All Other Deciles

Source: SEC Straddle Data; CRSP Data

Note: The straddle states that occurred on August 24, 2015 have been excluded.

Page 37

Average Number of Symbols Experiencing at Least One Long-Lasting

Straddle State Partitioned by Average Daily Market Capitalization[1]

Common Stock[2]

Average Number of

Symbols per Day

8/1/14 – 12/31/16[3]

13.8

12.8

8.1

5.1

2.1

0.5

0.1

0.0

Market Capitalization Decile[4]

Source: SEC Straddle Data; CRSP Data

Note:

[1] Average daily market capitalization is calculated on a monthly basis by multiplying the daily closing price by the daily number of shares outstanding, summing this

value across the month, and then dividing this sum by the number of trading days in the month.

[2] Common stocks are symbols with CRSP share code 11 or 12. Symbols with other share codes are excluded.

[3] There are 609 trading days in the sample, excluding August 24, 2015.

[4] Symbol-days were assigned to market capitalization deciles based on each individual month of trading. Therefore, a given symbol's percentile can vary from

month to month.

Page 38

Average Number of Symbols Experiencing at Least One Long-Lasting

Straddle State Partitioned by Average Daily Market Capitalization[1]

Lowest 20 Percentiles of Common Stock[2]

Average Number of

Symbols per Day

8/1/14 – 12/31/16[3]

1.9

1.5

1.4

1.3

1.5

1.4

1.3

1.4

1.3

1.2

1.3

1.2

1.3

1.2

1.1

1.0

Market Capitalization Percentile[4]

Source: SEC Straddle Data; CRSP Data

Note:

[1] Average daily market capitalization is calculated on a monthly basis by multiplying the daily closing price by the daily number of shares outstanding, summing this

value across the month, and then dividing this sum by the number of trading days in the month.

[2] Common stocks are symbols with CRSP share code 11 or 12. Symbols with other share codes are excluded.

[3] There are 609 trading days in the sample, excluding August 24, 2015.

[4] Symbol-days were assigned to market capitalization deciles based on each individual month of trading. Therefore, a given symbol's percentile can vary from month

to month.

Page 39

Number of Symbols Experiencing at Least One Long-Lasting

Straddle State Partitioned by Average Daily Market Capitalization[1]

Common Stock[2]

Number of Symbols

8/24/15

Market Capitalization Decile[3]

Source: SEC Straddle Data; CRSP Data

Note:

[1] Average daily market capitalization is calculated on a monthly basis by multiplying the daily closing price by the daily number of shares outstanding, summing this

value across the month, and then dividing this sum by the number of trading days in the month.

[2] Common stocks are symbols with CRSP share code 11 or 12. Symbols with other share codes are excluded.

[3] Symbol-days were assigned to market capitalization deciles based on trading in August 2015.

Page 40

Average Number of Symbols Experiencing at Least One Long-Lasting

Straddle State Partitioned by Average Daily Market Capitalization[1]

Common Stock[2]

Average Number of

Symbols per Day

8/1/14 – 12/31/16[3]

13.8

12.8

All Symbols

Symbols That Experienced Long-Lasting Straddle States on 101 – 200 Days

Symbols That Experienced Long-Lasting Straddle States More Than 200 Days

8.1

5.1

3.2

3.1

2.3

2.1

1.2

1.3

1.6

1.3

0.7

0.4

0.1

0.5

0.2 0.0

0.1 0.0 0.0

0.0 0.0 0.0

Market Capitalization Decile[4]

Source: SEC Straddle Data; CRSP Data

Note:

[1] Average daily market capitalization is calculated on a monthly basis by multiplying the daily closing price by the daily number of shares outstanding, summing this

value across the month, and then dividing this sum by the number of trading days in the month.

[2] Common stocks are symbols with CRSP share code 11 or 12. Symbols with other share codes are excluded.

[3] There are 609 trading days in the sample, excluding August 24, 2015.

[4] Symbol-days were assigned to market capitalization deciles based on each individual month of trading. Therefore, a given symbol's percentile can vary from

month to month.

Page 41

Percent of Long-Lasting Straddle States per Trading Day by CRSP Market Cap Deciles

Common Stocks Only

August 2014 – December 2016

% of Straddles

Per Day

100

Second Decile

Third Decile

First Decile

All Other Deciles

Source: SEC Straddle Data; CRSP Data

Note: The straddle states that occurred on August 24, 2015 have been excluded.

Page 42

Analysis of Long-Lasting Straddle States

Factors That Influence the Occurrence of Straddle States

• Implement a Logistic Regression Model to test which factors significantly increase the likelihood

of experiencing a long-lasting straddle state.

• Run at the symbol-day level.

– Dependent Variable: Indicator variable with a value of one if the stock experienced at least

one long-lasting straddle state on that particular day.

– Independent Variables:

• Log of trading volume on that day.

• Indicator variable with a value of one if the stock is in Tier 1.

• Indicator variable with a value of one if the prior day close was $3 or below.

• Indicator variable with a value of one if the prior day close was below $0.75.

• Indicator variable with a value of one if there was no opening trade.

• Abnormal volume measured as quintile rank of the current day’s volume among the prior 30 days.

• Stock volatility measured by the standard deviation of returns over the prior 30 days.

– Sample is limited to the universe of stocks in CRSP.[1]

[1] CRSP does not have data on BATS listed securities, ETNs, preferred stock, and units. In addition, symbols-days for which the standard deviation of the

daily returns of the prior 30 days exceeded 20% and instances in which the existence of an opening trade could not be identified are excluded. One share is

added to the volume on all symbol-days in order to preserve symbol-days with zero volume.

Page 43

Logistic Regression Results[1]

[2]

Likelihood of Experiencing a Long-Lasting Straddle

[3]

8/1/14 – 12/31/16

Including Time Fixed Effects[4]

Parameter

Log Volume

Tier 1

Prior Day's Close Below $3

[5]

Prior Day's Close Below $0.75

Abnormal Volume

No Opening Trade

Equity Volatility

Number of Symbol/Days

All Symbols

Common Stock

ETFs

All Symbols

Common Stock

ETFs

(1)

(2)

(3)

(4)

(5)

(6)

-0.277**

-0.362**

-0.232**

-0.275**

-0.361**

-0.243**

(0.000)

-2.714**

-3.605**

-0.825**

-2.724**

-3.632**

-0.721**

(0.000)

(0.004)

(0.000)

-1.385**

-1.464**

-1.406*

-1.390**

-1.485**

-1.488**

(0.000)

(0.047)

(0.000)

(0.002)

-1.420**

-1.136**

–

-1.460**

-1.197**

–

(0.000)

0.363**

0.449**

0.472**

0.354**

0.443**

0.471**

(0.000)

0.698**

0.570**

1.090**

0.727**

0.607**

1.179**

(0.000)

37.373**

31.067**

36.465**

37.361**

30.805**

41.804**

(0.000)

4,261,730

2,247,031

915,771

4,261,730

2,247,031

839,834

Sources: CRSP; SEC Straddle Data; SEC Price Bands Data; TAQ Data

[1] This table reports results from a logit regression estimating the effects of various factors on the likelihood of experiencing a long-lasting straddle state. The sample includes data for all stocks included in the

CRSP database. The dependent variable is an indicator variable that equals one if a stock experienced a long-lasting straddle on that particular day. Standard errors are estimated by clustering on stocks for the

specifications run without fixed effects. The "Permno" in CRSP is used to identify stocks over time. P-values are reported in parentheses below the coefficient estimates. Coefficient estimates that are statistically

different from zero at the 5% significance level are marked with one asterisk and those that are statistically different from zero at the 1% significance level are marked with two asterisks.

[2] Long-lasting straddle states are defined as a straddle state lasting at least five minutes.

[3] The straddle states that occurred on August 24, 2015 have been excluded.

[4] Time fixed effects were implemented at the day level.

[5] There were only 11 observations where the prior day's closing price was below $0.75 for ETFs. This variable is excluded for the specifications that only include ETFs.

Page 44

Analysis of Long-Lasting Straddle States

Section IV: Activity During and After Long-Lasting Straddle States

Page 45

Analysis of Long-Lasting Straddle States

Propensity of Straddle States to Recur Across Days

• Of the 2,364 symbols that experienced a long-lasting straddle state, 26% experienced

long-lasting straddle states on just one day.

• 30 symbols experienced at least one long-lasting straddle state on more than 200 days out

of a sample of 609 days.

– These 30 symbols are not concentrated within a particular volume decile, market

capitalization decile, or price level.

• Typically 10 to 25 symbols per day had a long-lasting straddle state in each of the five

previous trading days.

• Similarly, 10 to 25 symbols per day had no long-lasting straddle states in the previous five

trading days, although there was more day-to-day variation in this metric.

Page 46

Distribution of Number of Trading Days on Which a Stock

Experienced a Long-Lasting Straddle State

Number of

Symbols[1]

700

610

600

500

400

308

300

313

266

200

183

169

184

135

100

6 – 10

11 – 25

Number of Trading Days[2]

26 – 50

51 – 100

101 – 200

More than

200

Source: SEC Straddle Data

Note:

[1] There are 2,364 unique symbols in the full sample.

[2] There are 609 trading days in the full sample. August 24, 2015 has been excluded from the sample.

Page 47

Daily Counts of Symbols Experiencing at Least One Long-Lasting Straddle State

Compared to Symbols with a Long-Lasting Straddle State in Each of the Previous Five Trading Days

August 2014 – December 2016

Number of

Symbols

180

160

140

120

100

Symbols with a Straddle

Symbols with a Straddle and a Straddle in Each of the Previous Five Trading Days

Source: SEC Straddle Data

Note: The symbols that experienced straddle states on August 24, 2015 have been excluded.

Page 48

Daily Counts of Symbols Experiencing at Least One Long-Lasting Straddle State

Compared to Symbols without a Long-Lasting Straddle in the Last Five Trading Days

August 2014 – December 2016

Number of

Symbols

180

160

140

120

100

Symbols with a Straddle

Symbols with a Straddle That Did Not Have a Straddle in the Previous Five Trading Days

Source: SEC Straddle Data

Note: The symbols that experienced straddle states on August 24, 2015 have been excluded.

Page 49

Analysis of Long-Lasting Straddle States

Propensity of Straddle States to Recur Within Days

• The majority of symbols that experience a long-lasting straddle state on a given day

experience only one long-lasting straddle state on that day.

– Of the symbols that experience a long-lasting straddle state on a given day, 29.5%

experience at least two long-lasting straddles, and 12.8% experience at least three

long-lasting straddles.

• 54.4% of long-lasting straddle states are eventually followed by another straddle state, short

or long, on the same day.

– 67.6% of the next straddle states are short and 32.4% are long.

– 22% of long-lasting straddle states are immediately followed by another straddle state,

and another 13.5% are followed by another straddle state within one minute.[1]

• The majority of long-lasting straddle states did not end by immediately entering into a limit

or straddle state.

– 10.8% of long-lasting straddle states resulted in a limit state.

– 7.0% of long-lasting straddle states ultimately resulted in a trading halt.

• Symbols in the first decile of trading volume were more likely to immediately enter into

another straddle state and about equally likely to enter into a limit state when compared to

symbols with higher trading volume.

[1] “Immediately” means that the symbol entered into another straddle state within the same millisecond.

Page 50

Daily Counts of Symbols Experiencing at Least One Long-Lasting Straddle State

August 2014 – December 2016

Number of

Symbols

180

160

140

120

100

At Least 1 Straddle

At Least 2 Straddles

At Least 3 Straddles

At Least 4 Straddles

Source: SEC Straddle Data

Note: The symbols that experienced straddle states on August 24, 2015 have been excluded.

Page 51

Long-Lasting Straddle States Partitioned by Time to Recurrence and

Whether the Ensuing Straddles State Was Short or Long[1]

8/1/14 – 12/31/16[2]

As a Percentage of All Long-Lasting Straddle States[3]

Number of Straddles States

Time to Next Straddle State

Entered into a

Short Straddle State

Entered into a

Long Straddle State

Total

Entered into a

Short Straddle State

Entered into a

Long Straddle State

Total

Immediately Re-entered[4]

12,837

5,998

18,835

15.0%

7.0%

22.0%

> 0 seconds – 1 minute

8,465

3,156

11,621

9.9%

3.7%

13.5%

1 minute – 10 minutes

5,253

3,243

8,496

6.1%

3.8%

9.9%

10 minutes – 1 hour

2,878

1,572

4,450

3.4%

1.8%

5.2%

1 hour – 6.5 hours

2,141

1,145

3,286

2.5%

1.3%

3.8%

Total

31,574

15,114

46,688

36.8%

17.6%

54.4%

Source: SEC Straddle Data

Note:

[1] Long-lasting straddle states are defined as lasting five minutes or more and short-lasting straddle states include all other straddle states.

[2] The straddle states that occurred on August 24, 2015 have been excluded.

[3] Of the 85,808 long-lasting straddle states in the sample, 39,120 (45.6% of the full sample) of the long-lasting straddle states were the last straddle state of the day, short or long. Of those 39,120

straddle states, 8,569 (10% of the full sample) lasted until the end of the trading day.

[4] “Immediately” means that the symbol entered into another straddle state within the same millisecond.

Page 52

Long-Lasting Straddle Count by Manner in which the

Straddle State Was Resolved[1]

8/1/14 – 12/31/16[2]

10.0%

Did Not Immediately Enter

Limit or Straddle State

3.8%

Immediately Entered into a

Short-Lasting Straddle State

7.0%

Immediately Entered into a

Long-Lasting Straddle State

7.0%

57.2%

Immediately Entered Limit

State, Then Halted

Immediately Entered Limit

State, But Did Not Halt

15.0%

Reached the End of the

Trading Day

Source: SEC Straddle Data; SEC Halts Data; SEC Limit Data

Note:

[1] Long-lasting straddle states are defined as a straddle state lasting at least five minutes. “Immediately” means that the symbol entered into another

straddle state within the same millisecond.

[2] The straddle states that occurred on August 24, 2015 have been excluded.

Page 53

Analysis of Long-Lasting Straddle States

Trading and Quoting Activity During Long-Lasting Straddle States

• For a sample of straddle states with a clean control period[1][2], spreads during the straddle state

were wider relative to the spread during the same time interval on the five previous trading days

for 95% of straddle events.

– For straddle events that occurred while the bands were single wide, relative quoted spreads

were typically between 8% and 16%, though with more variation for symbols in the lowest

quintile of volume.[3]

• In 14% of long-lasting straddle events the relative quoted spread was locked, crossed, or

exceeded 100% for at least 80% of the duration of the straddle state.[4]

• There were no shares traded during 75.3% of long-lasting straddle states.[5]

– For a sample of straddle states with a clean control period[1], volume during the long-lasting

straddle state was typically lower than the volume during the same time interval on the five

previous trading days.

• Limit up long-lasting straddle states were somewhat more common than limit down long-lasting

straddle states.

– For roughly one-quarter of straddle states, both the bid and the offer were outside the price

bands.

[1] The control period for these analyses is the time window corresponding to the straddle event on the five trading days prior to the straddle event. A control period is “clean” if there

were five trading days to use as a control period prior to the day of the straddle event and there were no long-lasting straddle states during the control period.

[2] Intervals during which the quoted relative spread was greater than 100% (or greater than 150% if the bid-ask midpoint was below $0.75) and intervals during which the quoted

spread was locked or crossed have been excluded. If the total amount of time excluded from the time-weighted spread calculation exceeded 50% of the straddle duration, the

straddle state was excluded from the analysis.

[3] This sample includes symbols trading between $5 and $100 and symbols that appear in the CRSP data. Approximately 10% of straddle events are included in this sample.

[4] If the bid-ask midpoint was below $0.75, a threshold of 150% was used instead of 100%.

[5] Traded volume includes regular trades, intermarket sweep orders, and odd lots.

Page 54

Quoted Spreads during Straddle Events Relative to Typical Quoted Spreads[1][2]

8/1/14 – 12/31/16[3]

Percent of Events

25%

23.3%

20%

18.6%

15%

12.8%

11.0%

10%

7.9%

6.9%

5.8%

5.4%

4.4%

3.9%

0–1

1–2

2–3

3–4

4–5

5 – 10

10 – 25

25 – 50

50 – 100

100+

Ratio

Source: SEC Straddle Data; SEC Halts Data; TAQ Data

Note:

[1] Calculated as the ratio of the time-weighted relative quoted spread during the straddle event to the time-weighted relative quoted spread during the control

period. The control period spread is the average quoted spread during the same window of time as the straddle event on the five preceding trading days. If a symbol

was in a long-lasting straddle state during any of the control windows of time, or if there were not five prior trading days (e.g., the stock listed less than five days ago),

the straddle state was excluded from the analysis.

[2] Intervals during which the quoted relative spread was greater than 100% (or greater than 150% if the bid-ask midpoint was below $0.75) and intervals during

which the quoted spread was locked or crossed have been excluded. If the total amount of time excluded from the time-weighted spread calculation exceeded 50%

of the straddle duration, the straddle state was excluded from the analysis.

[3] The straddle states that occurred on August 24, 2015 have been excluded.

Page 55

Distribution of Time-Weighted Quoted Relative Spreads during the Control Period[1][2]

Partitioned by CRSP Volume Quintile[3]

Percent of Events

8/1/14 – 12/31/16[4]

50%

41.6%

38.6% 39.4%

40%

30%

Lowest Volume Quintile

(0% – 20%)

Second Lowest Volume Quintile

(20% – 40%)

28.0%

24.2% 24.4%

24.2%

Middle Volume Quintile

(40% – 60%)

21.4%

20%

15.1%

11.7%

11.3%

9.9%

10%

5.8%

2.9%

1.5%

< 2%

2% – 4%

4% – 6%

6% – 8%

> 8%

Quoted Relative Spread

Source: SEC Straddle Data; SEC Halts Data; TAQ Data; CRSP

Note:

[1] The sample includes straddle states that occurred during the time period that the bands were single wide and on symbols with a closing price between $5 and

$100. The control period spread is the average quoted spread during the same window of time as the straddle event on the five preceding trading days. If a symbol

was in a long-lasting straddle state during any of the control windows of time, or if there were not five prior trading days (e.g., the stock listed less than five days

ago), the straddle state was excluded from the analysis.

[2] Intervals during which the quoted relative spread was greater than 100% (or greater than 150% if the bid-ask midpoint was below $0.75) and intervals during

which the quoted spread was locked or crossed have been excluded. If the total amount of time excluded from the time-weighted spread calculation exceeded 50%

of the straddle duration, the straddle state was excluded from the analysis.

[3] Symbols are partitioned into quintiles on a monthly basis using trading volume in CRSP.

[4] The straddle states that occurred on August 24, 2015 have been excluded.

Page 56

Distribution of Time-Weighted Quoted Relative Spreads during the Straddle Event[1][2]

Partitioned by CRSP Volume Quintile[3]

Percent of Events

8/1/14 – 12/31/16[4]

50%

Lowest Volume Quintile

(0% – 20%)

40%

36.0%

38.7%

37.2%

Second Lowest Volume Quintile

(20% – 40%)

34.8%

Middle Volume Quintile

(40% – 60%)

30%

26.0%

25.0%

21.9%

20%

16.8%

11.2%

12.4%

10.3% 10.2%

10%

7.8% 7.3%

4.4%

< 8%

8% – 12%

12% – 16%

16% – 20%

> 20%

Quoted Relative Spread

Source: SEC Straddle Data; SEC Halts Data; TAQ Data; CRSP

Note:

[1] The sample includes straddle states that occurred during the time period that the bands were single wide and on symbols with a closing price between $5 and

$100. In order to make the sample parallel to the control period sample, if a symbol was in a long-lasting straddle state during any of the control windows of time, or if

there were not five prior trading days (e.g., the stock listed less than five days ago), the straddle state was excluded from the analysis.

[2] Intervals during which the quoted relative spread was greater than 100% (or greater than 150% if the bid-ask midpoint was below $0.75) and intervals during

which the quoted spread was locked or crossed have been excluded. If the total amount of time excluded from the time-weighted spread calculation exceeded 50%

of the straddle duration, the straddle state was excluded from the analysis.

[3] Symbols are partitioned into quintiles on a monthly basis using trading volume in CRSP.

[4] The straddle states that occurred on August 24, 2015 have been excluded.

Page 57

Percentage of Time during Straddle Event That the Quoted Spread Was

Locked, Crossed, or the Relative Spread Exceeded 100%[1]

Percent of Events

[2]

8/1/14

–

12/31/16

100%

90%

85.0%

80%

70%

60%

50%

40%

30%

20%

10.4%

10%

3.5%

0.5%

0.2%

1% – 20%

20% – 40%

40% – 60%

60% – 80%

< 1%

80% – 99%

> 99%

Percentage of Time during Straddle Event

Source: SEC Straddle Data; SEC Halts Data; TAQ Data

Note:

[1] Calculated as the percentage of time during the straddle event that the quoted relative spread was locked, crossed, or greater than 100% (or greater than 150% if

the bid-ask midpoint was below $0.75).

[2] The straddle states that occurred on August 24, 2015 have been excluded.

Page 58

Quoted Spreads during Straddle Events Relative to the Price Bands[1][2]

8/1/14 – 12/31/16[3]

Percent of Events

50%

44.6%

40%

30%

19.2%

20%

16.0%

10%

7.8%

5.0%

2.2%

1.6%

1.2%

1.3%

3.5 – 4.0

4.0 – 4.5

4.5 – 5.0

0.04%

0 – 0.5

0.5 – 1.0

1.0 – 1.5

1.5 – 2.0

2.0 – 2.5

2.5 – 3.0

3.0 – 3.5

5.0+

Ratio

Source: SEC Straddle Data; SEC Halts Data; SEC Price Bands Data; TAQ Data

Note:

[1] Calculated as the ratio of the time-weighted relative quoted spread during the straddle event to the width of the price bands.

[2] Intervals during which the quoted relative spread was greater than 100% (or greater than 150% if the bid-ask midpoint was below $0.75) and intervals during

which the quoted spread was locked or crossed have been excluded. If the total amount of time excluded from the time-weighted spread calculation exceeded 50%

of the straddle duration, the straddle state was excluded from the analysis.

[3] The straddle states that occurred on August 24, 2015 have been excluded.

Page 59

Distribution of Traded Volume during Long-Lasting Straddle States[1]

Percentage of

8/1/14 – 12/31/16[2]

Sraddle States

80%

75.3%

70%

60%

50%

40%

30%

20%

11.2%

10%

7.4%

2.6%

1.9%

0.9%

0.6%

501 – 1,000

1,001 – 2,000

2,001 – 4,000

4,000+

1 – 100

101 – 500

Volume during Straddle State

Source: SEC Straddle Data; TAQ Data

Note:

[1] Traded volume includes regular trades, intermarket sweep orders, and odd lots.

[2] The straddle states that occurred on August 24, 2015 have been excluded.

Page 60

Distribution of Traded Volume during Long-Lasting Straddle States and the Control Period[1]

Sample of Symbols with No Long-Lasting Straddle States in the Control Period[2]

8/1/14 – 12/31/16[3]

Percentage of

Sraddle States

80%

70%

66.7%

Volume during Straddle Event

Average Window Volume during Control Period

60%

50%

40%

37.0%

30%

20.7%

17.7%

20%

13.1%

11.2%

8.3%

10%

3.9%

7.0%

2.9%

5.2%

1.3%

4.1%

0.9%

1 – 100

101 – 500

501 – 1,000

1,001 – 2,000

2,001 – 4,000

4,000+

Volume

Source: SEC Straddle Data; SEC Halts Data; TAQ Data

Note:

[1] Traded volume includes regular trades, intermarket sweep orders, and odd lots.

[2] The control period spread is the average daily volume during the same window of time as the straddle event on the five preceding trading days. If a symbol

was in a long-lasting straddle state during any of the control windows of time, or if there were not five prior trading days (e.g., the stock listed less than five days

ago), the straddle state was excluded from the analysis.

[3] The straddle states that occurred on August 24, 2015 have been excluded.

Page 61

Distribution of Long-Lasting Straddle States by Straddle Type

8/1/14 – 12/31/16[1]

24%

41%

Limit Up Straddle State

Limit Down Straddle State

Both Quotes Outside of Bands

Multiple Straddle Types

32%

Source: SEC Straddle Data; SEC Price Bands Data; TAQ Data

Note:

[1] The straddle states that occurred on August 24, 2015 have been excluded.

Page 62

Analysis of Long-Lasting Straddle States

Section V: Impact of Amendment 10 on Long-Lasting Straddle States

Page 63

Analysis of Long-Lasting Straddle States

Impact of Amendment 10 on Trading Halts

• Amendment 10 stated that stocks with no volume in the opening auction would use the price of

the prior day’s close as the first reference price instead of the midpoint of the opening quote.

– Implemented on July 18, 2016

• Extend methodology of logistic regression model from Section III to test the impact of rule

change on trading halts (the target of the rule change).

∗

–

is an indicator variable with a value of one after implementation.

–

is an indicator variable with a value of one if there was no opening trade.

–

∗

is the interaction of the two indicators.

are the control variables from the prior regression.

• In this specification, the coefficient on the variable of interest, the interaction term, is significant

and corresponds to the odds ratio of a stock experiencing a trading halt conditional on that stock

having no opening trade being roughly five to seven times lower after the implementation of

Amendment 10.

Page 64

Logistic Regression Results[1]

[2]

Likelihood of Experiencing a Trading Halt

[3]

8/1/14 – 12/31/16

Full Sample

Balanced Panel

(1)

(2)

-0.181**

-0.233**

(0.000)

-0.484**

-0.212

(0.000)

(0.162)

-1.217**

-1.236**

(0.000)

-0.680**

-0.455**

(0.000)

(0.009)

2.248**

2.143**

(0.000)

-0.160*

-0.022

(0.044)

(0.820)

No Opening Trade∗Post

Amendment 10

-1.743**

-1.909**

(0.000)

Abnormal Volume

0.400**

0.498**

(0.000)

30.200**

33.884**

(0.000)

4,261,730

3,242,584

Parameter

Log Volume

Tier 1

Prior Day's Close Below $3

Prior Day's Close Below $0.75

No Opening Trade

Post Amendment 10

Equity Volatility

Number of Symbol/Days

Sources: CRSP; SEC Halts Data; SEC Price Bands Data; TAQ Data

[1] This table reports results from a logit regression estimating the effects of various factors on the likelihood of experiencing a trading halt. The sample

includes data for all stocks included in the CRSP database. The dependent variable is an indicator variable that equals one if a stock experienced a halt on

that particular day. Standard errors are estimated by clustering on stocks. The "Permno" in CRSP is used to identify stocks over time. P-values are

reported in parentheses below the coefficient estimates. Coefficient estimates that are statistically different from zero at the 5% significance level are

marked with one asterisk and those that are statistically different from zero at the 1% significance level are marked with two asterisks.

[2] Trading halts that were unrelated to the Limit Up-Limit Down rules were removed.

[3] The straddle states that occurred on August 24, 2015 have been excluded.

Page 65

Analysis of Long-Lasting Straddle States

Impact of Amendment 10 on Long-Lasting Straddle States

• We use the same methodology to test whether Amendment 10 had any impact on the

likelihood of a stock experiencing a long-lasting straddle state.

∗

–

is an indicator variable with a value of one after implementation.

–

is an indicator variable with a value of one if there was no opening trade.

–

∗

is the interaction of the two indicators.

are the control variables from the prior regression.

• The coefficient on the variable of interest, the interaction term, is significant and corresponds to

the odds ratio of a stock experiencing a long-lasting straddle state conditional on that stock

having no opening trade being roughly 1.2 times greater after the implementation of Amendment

10.

Page 66

Logistic Regression Results[1]

[2]

Likelihood of Experiencing a Long-Lasting Straddle

[3]

8/1/14 – 12/31/16

Full Sample

Balanced Panel

(1)

(2)

-0.276**

-0.352**

(0.000)

-2.713**

-2.685**

(0.000)

-1.384**

-1.437**

(0.000)

-1.418**

-1.060**

(0.000)

0.674**

0.512**

(0.000)

-0.026

-0.036

(0.683)

(0.640)

No Opening Trade∗Post

Amendment 10

0.153*

0.226**

(0.032)

(0.009)

Abnormal Volume

0.362**

0.468**

(0.000)

37.397**

39.391**

(0.000)

4,261,730

3,242,584

Parameter

Log Volume

Tier 1

Prior Day's Close Below $3

Prior Day's Close Below $0.75

No Opening Trade

Post Amendment 10

Equity Volatility

Number of Symbol/Days

Sources: CRSP; SEC Straddle Data; SEC Price Bands Data; TAQ Data

[1] This table reports results from a logit regression estimating the effects of various factors on the likelihood of experiencing a long-lasting straddle state.

The sample includes data for all stocks included in the CRSP database. The dependent variable is an indicator variable that equals one if a stock

experienced a long-lasting straddle on that particular day. Standard errors are estimated by clustering on stocks. The "Permno" in CRSP is used to identify

stocks over time. P-values are reported in parentheses below the coefficient estimates. Coefficient estimates that are statistically different from zero at the

5% significance level are marked with one asterisk and those that are statistically different from zero at the 1% significance level are marked with two

asterisks.

[2] Long-lasting straddle states are defined as a straddle state lasting at least five minutes.

[3] The straddle states that occurred on August 24, 2015 have been excluded.

Page 67

This is a copy of a public record, reproduced as it was published. It is not legal advice, and it may not be the version a court would rely on. Check the official source before you cite it.