DIVISION OF ECONOMIC AND RISK ANALYSIS1

Agency decision

Ask Donna

What actually matters in this document.

Text

MEMORANDUM

FILE

TO:

DIVISION OF ECONOMIC AND RISK ANALYSIS1

FROM:

SUBJECT: CORNERSTONE ANALYIS OF THE IMPACT OF THE

OPTION PENNY PILOT PROGRAM ON MARKET QUALITY

DATE:

DECEMBER 18, 2017

The Division of Economic and Risk Analysis contracted with Cornerstone Research

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

investigating the effects on market quality of the continuing pilot program that allows certain

exchange-traded options to be quoted in reduced price increments2 (“Option Penny Pilot” or

“Pilot”). Cornerstone’s analysis uses quoted and effective spreads as measures of market

quality.3 Effective and quoted spreads are frequently used as measures of trading costs and are

especially useful in assessing overall market quality for retail investors.4 The Division 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.

Main Finding:

The Cornerstone analysis finds that inclusion in the Pilot is associated with a statistically

significant decrease in effective and quoted spreads for the most liquid options - those with

trading volume in the top five percent of the universe of all listed option classes.5 However, for

less liquid options, the results do not suggest that inclusion in the Pilot has a significant effect on

This is a memorandum 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.

More specifically, the Pilot specifies that options trading at less than $3.00 have trading increments of one cent,

while those trading at $3.00 or more have trading increments of five cents. Options not in the Pilot have

corresponding minimum price increments of $0.05 for options trading below $3.00 and $0.10 for options above

$3.00. The Pilot makes an exception for options on the QQQQ, IMW and SPY which trade at or above one cent

increments, regardless of price level.

In Cornerstone’s analysis, spreads are volume-weighted across different trades in an option series, and then

equally weighted across option series, in order to obtain spreads for the option-class day. Those spreads are then

equally weighted across different days in the pre-event month and the post-event month, respectively, in order to

obtain option-class month observations.

“Decimal Pricing Has Contributed to Lower Trading Costs and a More Challenging Trading Environment,” GAO

report, May 2005, at 3 (available at http://www.gao.gov/new.items/d05535.pdf).

Note that Cornerstone’s analysis on option classes on common stock incorporates data on 3,833 option classes and

1,331,185 class-day observations.

market quality. Cornerstone reports that the option classes added to the Pilot were mostly

chosen from the most liquid and heavily traded option classes.6

Comparison to Analyses of Self-Regulatory Organizations:

Several Self-Regulatory Organizations (SROs) conducted analyses of the effects of the Options

Penny Pilot. The SROs made these analyses available on their websites.7 The reports of the

SROs focus on the activity on their own exchanges; none of the reports takes a holistic view of

the Pilot on all exchanges. Also, their analyses typically did not use control samples. Finally,

the most recent of the publicly available reports from an SRO dates from 2013, and thus the

Cornerstone analysis covers more recent time periods.8 However, the SRO analyses generally

agree with the conclusions drawn from Cornerstone’s analysis that the Pilot was associated with

a narrowing of spreads.

Analysis:

Cornerstone’s analysis uses quoted and effective spreads as a measure of market quality, and it

employs two methodologies.

The first methodology uses a univariate “difference-in-differences” approach, in which the

changes in quoted and effective spreads during the Pilot for options classes included in the Pilot

are compared to contemporaneous changes in spreads for a control sample of option classes that

were not included in the pilot. Options classes included in the control sample were matched on

option class volume and underlying market capitalization. This sample was referred to as the

“matched sample”.

The second methodology uses multivariate regression models to explain variation in the quoted

and effective spreads before and after option classes were added to the Pilot. These models

include independent variables to control for other factors that can impact spreads, including

option volume, underlying stock volume, and underlying stock volatility.

Using data on options on common stocks added to the Pilot between January 26, 2007 and

January 5, 2016,9 both methodologies have similar findings. The “difference-in-differences”

Although some of the original thirteen options were not among the most liquid, most heavily traded options, the

order approving the commencement of the penny pilot program describes the selection of the initial option classes

as follows; “The Commission believes that the thirteen options classes to be included in the penny pilot program

represent a diverse group of options classes with varied trading characteristics. This diversity should facilitate

analyses by the Commission, the options exchanges and others.” See Securities Exchange Act Release No. 55156

(January 23, 2007), 72 FR 4759, 4760 (February 1, 2007) (SR-NYSEArca-2006-73) (Order approving a penny

pilot program on NYSE Arca).

These data and analyses are available online at https://www.nyse.com/markets/amex-options/reports ,

http://www.ise.com/options-penny-pilot , https://www.cboe.org/generalinfo/pdfframed?content=/hybrid/cboepennypilotreport03042008.pdf&section=SEC_HYBRID_HOME&title=CB

OE%20Penny%20Pilot%20Report and http://boxoptions.com/regulatory/pilot-reports/

http://www.ise.com/assets/files/investors/Topaz_Penny_Pilot_Report_15.pdf

Results for 693 option classes on ETFs were generally similar to option classes on common equity.

tests find that, compared to the control sample, inclusion in the Pilot is associated with a

decrease in effective spreads of 3.5 cents and a decrease in quoted spreads of 5.1 cents.

Similarly, using the same matched sample, the regression model finds that the Pilot is associated

with a decrease in effective spreads of 3.6 cents and a decrease in quoted spreads of 5.2 cents.

These results are statistically significant.

Cornerstone reports results separately for sub-periods 2007-2009, 2010-2011, and 2012-2015.

The results are generally weaker in both statistical significance and in economic magnitude for

the period 2012-2015. One potential explanation advanced by Cornerstone for the smaller

decrease in spreads for later additions to the Pilot is that the benefits of the Pilot are most

pronounced for options on larger stocks or for option classes with the highest trading volume,

and to the extent that the later additions to the Pilot tend to be smaller stocks with lower-volume

options, there would be smaller declines in spreads for the later additions. Also, the 2012–2015

results are based on a relatively small sample of options.

One caveat to the results of the analysis that Cornerstone emphasizes is that the option

exchanges have a different fee structure for classes within and outside the Pilot. Cornerstone’s

analysis does not try to examine whether the improvements in market quality could be attributed

only to the Pilot, or also to the differing fee structures. However, Cornerstone also notes that the

average declines in execution costs for the most liquid option classes appear to be substantially

higher than the potential impact of any fee adjustments.

As option classes added to the Pilot were mostly the most liquid and heavily traded, Cornerstone

also includes analysis specifically aimed at testing whether the Pilot contributed to improved

market quality for less liquid option classes. The analysis uses more recent data, from February

2016, in order to allow the testing of the impact of the Pilot on option classes that may have been

more liquid when they were originally added to the Pilot, but that over time became less liquid.

Due to the low liquidity of these option classes, had the inclusion decision been made in

February 2016, they may have not been selected for inclusion in the Pilot.

However, Cornerstone’s analysis is neither consistent nor inconsistent with the proposition that

additional low volume option classes would benefit from pilot inclusion.10 Compared to

samples of non-Pilot options matched on option class volume, underlying market capitalization,

and underlying price, the results indicate evidence of statistically significant reductions in

quoted and effective spreads for the top two quartiles, by trading volume, of option classes

included in the Pilot. These option classes have trading volume in the top 5% of the universe of

all listed option classes. While Cornerstone also finds reductions in spreads for the bottom two

Also, there exist some option classes with liquidity equal to that of those in the Pilot, but they are not in the Pilot

because they did not meet the Pilot requirements at the time a replacement class was being added to the Pilot.

Some of these are included in the controls used for the Third and Fourth Quartile in Table 22 of Cornerstone’s

research, described as “Number of Clusters for Matching with Replacement”.

quartiles of option classes included in the pilot, these reductions are generally not statistically

significant.

Conclusion:

The evidence Cornerstone presents suggests that the existing Pilot program has resulted in

improvements in market quality (as measured by quoted and effective spreads) for the Pilot

option classes. Cornerstone’s research does not provide any clear evidence that expanding the

Pilot to include lower-volume options would result in similar benefits. While noting that the

analysis does not control for different fee structures that apply to option classes included and

excluded from the Pilot, Cornerstone asserts that the estimated reduction in execution costs for

option classes within the pilot appear to be substantially larger than the potential impact of fee

differentials.

MEMORANDUM

DATE:

July 3, 2017

TO:

Division of Economic and Risk Analysis

FROM:

Stewart Mayhew, Cornerstone Research

RE:

Analysis of the Impact of the Option Penny Pilot Program on Market Quality

Executive Summary

This memo summarizes research requested by the staff of the Commission’s Division of

Economic and Risk Analysis (DERA) under the terms of contract SECHQ1-16-C0024 with

Cornerstone Research.1 This research relates to an ongoing pilot program that, since 2007, has

allowed certain exchange-traded options to be quoted in reduced price increments (“Option

Penny Pilot” or “Pilot”). DERA staff requested an analysis of the impact of the Option Penny

Pilot on market quality, as part of an effort to evaluate the efficacy of the program. The current

memo summarizes our research investigating the extent to which inclusion in the Pilot impacts

measures of market quality. The results of this research should aid the staff of the Commission

as it seeks to evaluate the overall impact of the Pilot, and as it makes decisions regarding whether

The analysis summarized in this memo was undertaken by a team of researchers under the direction of Dr. Stewart

Mayhew at Cornerstone Research. The team also included Dr. Amber Anand of Syracuse University, Dr. D.

Timothy McCormick, and the staff of Cornerstone Research. The findings presented herein are results of analysis

performed by the research team in response to specific requests and questions from DERA staff. This memo does

not seek to offer any policy recommendations, and should not be construed as an endorsement by the authors or by

Cornerstone Research of any particular policy alternative. Any views and interpretations expressed in this memo are

solely those of Dr. Mayhew, Dr. Anand, and Dr. McCormick, who are responsible for the content, and do not

necessarily represent the views of Cornerstone Research.

the program should be made permanent and/or whether the program covers the appropriate

number of option classes.

The research summarized herein encompasses the addition of option classes to the Pilot

over an eight-year period between January 26, 2007 and January 5, 2016. It analyzes the impact

of addition to the Pilot on quoted and effective spreads, while attempting to control for other

factors that influence spreads and for any selection bias that may influence the addition of certain

option classes to the program.

Section I of this memo provides a brief overview of the background of the Pilot,

including the timing of when new option classes were added. Section II describes the data used

for the analysis.

Section III reports summary statistics for option class volume, underlying stock volume,

and underlying market capitalization for option classes already in the Pilot, new classes selected

for inclusion in the Pilot, and option classes not in the Pilot, as of each addition date (“event

date”). The option classes in the Pilot and those selected for inclusion tend to have significantly

higher trading volume and tend to be on underlying stocks with higher market capitalization

compared to classes not participating in the Pilot.

Section IV provides additional detail regarding the construction of control samples and

reports summary statistics for the control samples.

Section V presents results of univariate “difference-in-differences” tests, which compare

changes in quoted and effective spreads after option classes are added to the Pilot to

contemporaneous changes in spreads for a control sample of non-switching option classes

(matched on option class volume and underlying market capitalization). The difference-indifferences results (presented in Table 5) indicate that inclusion in the Pilot is associated with a

decrease in dollar effective spreads of 3.5 cents and a decrease in dollar quoted spreads of 5.1

cents. These decreases were found to be statistically significant at the 1% confidence level, and

similar results were found for relative spreads. Results are also reported separately for

subperiods 2007–2009, 2010–2011, and 2012–2015. The average decline in spreads was smaller

for the most recent additions to the Pilot in 2012–2015, although this group has relatively few

observations. The decline in spreads is found to be robust for option classes with different levels

of underlying market capitalization (Table 6) and option class trading volume (Table 7).

Section VI analyzes additions to the Pilot using a multivariate regression approach

designed to control for other factors that can impact spreads, including option volume,

underlying stock volume, and stock volatility. This analysis confirms that inclusion in the Pilot

is associated with a statistically significant decline in effective spreads and quoted spreads,

although the estimated magnitude of the effect varies somewhat, depending on how the model is

implemented. When the regression model is estimated on the entire universe of listed options,

the results indicate that inclusion in the Pilot is associated with a decrease in effective spreads of

8.8 cents (Table 13) and a decrease in quoted spreads of 16.1 cents (Table 14). When the model

is estimated only on the matched sample described in Section IV, the Pilot is associated with a

decrease in dollar effective spreads of 3.6 cents (Table 15) and a decrease in dollar quoted

spreads of 5.2 cents (Table 16). Similar results are found for percentage spreads. The decline in

effective spreads is found to be largest in the 2010–2011 subperiod, and smallest for the most

recent additions in 2012–2015. In some specifications, the effect of the Pilot program is found to

be close to zero and not statistically significant for the 2012–2015 subperiod. We reiterate the

caveat for the 2012–2015 results that they are based on a relatively small sample of switching

options.

Section VII explores further whether the Pilot is contributing to an improvement in

market quality for the least liquid option classes, a question that is significant if the Commission

staff wishes to consider whether penny quoting should be expanded to a larger set of option

classes or constricted to a smaller set. This question presents empirical challenges because the

option classes added to the Pilot are generally the most liquid, thus making it difficult to

extrapolate the benefits of the Pilot to less liquid option classes. We address this by examining

market quality for Penny Pilot option classes using a more recent period of February 2016. The

examination of a recent period allows us to study Penny Pilot option classes that may have been

more liquid when they were originally added to the Pilot but with the passage of time became

less liquid. The analysis compares spreads between Pilot options and a control sample of (nonPilot) options, matched on option volume, underlying market capitalization, and underlying

price. Results are reported by option volume quartiles to examine whether the benefits of

inclusion in the Pilot taper off at a particular threshold (Table 23). The results indicate evidence

of significant benefits of inclusion in the Pilot for the top two quartiles, corresponding to the

option classes with trading volume in the top 5% of the universe of all listed option classes. For

the bottom two quartiles of Pilot options, most of the results are not statistically significant,

suggesting that the benefits of the Pilot decline as option liquidity declines.

Background: Option Penny Pilot Program

On Friday, January 26, 2007, the options exchanges commenced the Pilot program to

quote and trade options on 13 stocks and ETFs in one-cent increments. The Pilot specified that

options trading at less than $3.00 would have trading increments of one cent, while those trading

at $3.00 or more would have trading increments of five cents, with the exception of options on

the QQQQ which would trade at one cent increments, regardless of price level. The change

represents a reduction relative to options not in the Pilot, which have corresponding minimum

price increments of five cents for options trading below $3.00 and ten cents for options trading

above $3.00.

As summarized in Table 1, the Pilot was subsequently expanded to include additional

option classes. Most notably, the Pilot was expanded by an additional 22 option classes on

September 28, 2007 and 28 option classes on March 28, 2008. The Pilot was subsequently

expanded to include the next 300 most actively traded, multiply-listed options classes that were

not yet included in the Pilot, excluding options with high premiums. These option classes were

added in groups of 75 on each of November 2, 2009, February 1, 2010, May 3, 2010, and August

2, 2010. Further, SPX and IWM began trading in one-cent increments regardless of the price

level on February 1, 2010. Since 2011, additions have occurred at the beginning of January and

July each year, when a handful of options (less than 10) have been added to the Pilot to replace

options that no longer exist.2 This analysis encompasses the addition to the Pilot of options on

common equity and ETFs, commencing on January 26, 2007 and concluding on January 5, 2016.

II.

Data and Sample

Options Price Reporting Authority (OPRA) data were obtained from CBOE LiveVol,

which contains options trades, along with contemporaneous quotes from each exchange, as well

as the National Best Bid and Offer (NBBO). For each of the event dates listed in Table 1

See https://www.ise.com/options/regulatory-and-fees/options-penny-pilot/.

beginning in January 2007 and ending with the addition in January 2016, OPRA data were

obtained for the full calendar month prior to the event date (“pre-event month”) and for the full

calendar month subsequent to the event date (“post-event month”). For example, for the event

date on January 26, 2007, data were obtained for the calendar months of December 2006 and

February 2007.

The OPRA data were supplemented with data from the Center for Research in Security

Prices (CRSP) and OptionMetrics. CRSP data were used to obtain characteristics of the

underlying equity instrument, including market capitalization, average daily trading volume, and

prices. Daily OptionMetrics data were used to obtain additional characteristics of the traded

option series and to provide a mapping between the OPRA and CRSP data.

The analysis is focused on options on common stock and ETFs, with results reported

separately for common stock and ETFs.3 Option classes for which the underlying stock or ETF

traded below $1 or above $1,000 on any day during the pre-event month or the post-event month

of any addition date are excluded, as are option series with non-standard settlements. Outlier

spreads (effective or quoted) which are greater than $5 when the option price is less than $100,

or are greater than $10, are excluded from the analysis. The sample for option classes on

common stock incorporates data on 3,833 option classes and 1,331,185 class-day observations.

The sample for option classes on ETFs incorporates data on 693 option classes and 194,694

class-day observations.

The analyses reported in the body of this memo are based on a database of regular option trades (trades involving

only one leg). An analysis of complex trades shows similar results and is reported in the Appendix.

III.

Characteristics of Options Added to the Pilot

Table 2 contains summary statistics for option class volume, underlying volume, and

underlying market capitalization for option classes not in the Pilot, classes being added to the

Pilot, and classes already in the Pilot, computed during the pre-event month. Switching option

classes generally have a higher average daily option volume, higher average daily underlying

stock volume, and higher underlying market capitalization vis-à-vis option classes outside the

Pilot. This is not surprising and simply reflects the fact that more actively traded options on

more actively traded and larger stocks tend to be selected for inclusion in the Pilot. Note that the

volume and market cap metrics for switching options decreased substantially between 2007 and

2010, indicating that option classes with higher volume and market cap were added to the Pilot

prior to option classes with lower volume and market cap.

Since the large number of illiquid options may unduly influence the average for volume

and underlying market capitalization for options outside the Pilot, it is worth examining the

distribution of these characteristics across all option classes. Figure 1 shows a scatterplot of

option classes on common stock based on average daily option volume and underlying market

capitalization, while Figure 2 plots option classes on common stock based on average daily

underlying volume and underlying market capitalization. Figures 3 and 4 show analogous plots

for option classes on ETFs. All figures plot the data as of February 2016. The distributions

confirm that options in the Pilot have higher trading volume and larger underlying market

capitalizations vis-à-vis the overall population of traded options.

IV.

Control Sample

The primary goal of this research is to assess the degree to which including an option

class in the Pilot resulted in narrower quoted and effective spreads. Quoted and effective spreads

are frequently used as measures of trading costs and, more generally, to assess market quality.

Even though the analysis focuses on relatively short windows of one month before and one

month after each addition date, it is possible that other market-wide factors influencing option

spreads might have changed systematically between the pre-event and post-event months. The

difference-in-differences methodology recognizes this possibility and controls for any such

market-wide factors by comparing changes in market quality for the added option classes with

contemporaneous changes in a control sample of option classes that were not added.

Control samples are constructed by matching switching options with non-switching

options on the basis of option class volume and underlying market capitalization. Specifically,

each switching option was matched with a control option by rank ordering the entire universe of

traded options on option class volume and underlying market capitalization in the pre-event

month, and then minimizing the difference in ranks between the switching option and the control

option. Matching was completed without replacement, that is, each control option class is

matched with exactly one switching option class for each event date. The results are similar

across two different sets of matched controls: an “unconstrained” set, where the matched option

is any non-switching option regardless of whether it is in the Pilot, and a “constrained” set,

where the control option must be outside the Pilot. In later years, when most of the highest

option volume options are in the Pilot, the unconstrained sample is likely to provide a closer

match for the volume of the switching option. Table 3 contains summary statistics for the

control samples for option classes on common stock during the pre-event month. Table 4

contains the same summary statistics for option classes on ETFs.

Impact of the Pilot on Quoted and Effective Spreads

Figures 5–8 graphically illustrate the impact on spreads for the switching options after

being added to the Pilot vis-à-vis the control sample. Figure 5 depicts the average dollar

effective spreads during the pre-event month and the post-event month across the different

addition dates. As can be seen in Figure 5, dollar effective spreads for switching options are

substantially lower in the post-event month as compared to the pre-event month. No

corresponding decline in effective spreads occurs for the control sample. A similar pattern is

observed for relative effective spreads (Figure 6), dollar quoted spreads (Figure 7), and relative

quoted spreads (Figure 8). Figures 9–12 show a similar pattern for option classes on ETFs.

These figures suggest that inclusion in the Pilot is associated with substantially lower quoted and

effective spreads.4

Tables 5–12 present changes in effective and quoted spreads for options that are

switching into the Pilot,5 as compared to a control group of non-switching options. Spreads are

volume-weighted across different trades in an option series and then equally weighted across

option series, in order to obtain spreads for the option-class day. Those spreads are then equally

weighted across different days in the pre-event month and the post-event month, respectively, in

The graphs shown in Figures 5–8 are based on event-time data aggregated across all event dates. Similar graphs

could be made for any individual event date. For example, a graph showing the evolution of spreads surrounding

August 2, 2010 is provided in the Appendix.

The addition dates in January 2010 and July 2010 are the pre-event months for the additions in February 2010 and

August 2010, respectively and, hence, are excluded as addition dates from the analysis to ensure that event windows

do not overlap.

order to obtain option-class month observations. A similar methodology is adopted for

computing spreads for the control sample, and the results of a difference-in-differences

specification are reported. P-values are computed using a paired t-test and the Wilcoxon SignedRank test.

As shown in the first panel in Table 5, average dollar effective spreads declined from

10.5 cents in the pre-event month to 6.8 cents in the post-event month for the switching options

in the overall sample, a decrease of 3.7 cents, while effective spreads in the control sample

decreased by 0.2 cents. Relative to the control sample, effective spreads declined by 3.5 cents,

as reported in the column labeled “diff-in-diff.” This decline is statistically significant (at the 1%

level). Table 5 also reports results for the different subperiods. As indicated in the table,

inclusion in the Pilot is associated with a decrease in effective spreads of 3.3 cents for classes

added between 2007 and 2009, 3.8 cents for classes added between 2010 and 2011, and 1.4 cents

for classes added between 2012 and 2015. The second panel of Table 5 reports analogous results

based on percentage effective spreads.

Similarly, as shown in the third panel of Table 5, average dollar quoted spreads declined

from 13.9 cents in the pre-event month to 8.7 cents in the post-event month for the switching

options in the overall sample, while quoted spreads for the control sample decreased from 12.4 to

12.2 cents. Based on the difference-in-differences, inclusion in the Pilot is associated with a

decline in quoted spreads of 5.1 cents, which is statistically significant at the 1% level.

One potential explanation for the smaller decrease in spreads for later additions to the

Pilot is that the benefits of the Pilot are most pronounced for options on larger stocks or for

option classes with the highest trading volume, and to the extent that recent additions to the Pilot

tend to be smaller stocks with lower-volume options, there would be smaller declines in spreads

in recent years.

To examine this, Table 6 decomposes the sample into categories based on the underlying

market capitalization, using breakpoints calculated from NYSE stocks. Because most options

being added are in the largest market capitalization quartile, they are decomposed into more

granular categories. The difference-in-differences spreads are negative and statistically

significant for all categories, with the Wilcoxon Signed-Rank test significant at the 1% level.

Average dollar effective spreads declined by 3.5 cents for the largest quartile, by 3.6 cents for the

third quartile, and by 2.9 cents for options below the 50th percentile, vis-à-vis the changes in the

control group. Thus, the spread declines for Pilot stocks are consistent across all market cap

categories. In addition, all four spread measures show similar results.

Table 7 decomposes the sample based on the percentile of the option volume in the preevent month. Because almost the entire population of the switching options is in the top 10% by

volume, they are decomposed into more granular categories. The difference-in-differences

spreads are negative and statistically significant for all categories, with the Wilcoxon SignedRank test significant at the 1% level. Relative effective spreads declined by 4.4% for the top two

percentiles, by 6% for the 95th to 97th percentile, by 5.5% for the 90th to 94th percentile, and by

7.2% for options below the 90th percentile, vis-à-vis changes in the control groups. The spread

declines for Pilot option classes are consistent across all volume categories for all four spread

measures.

Table 8 decomposes the sample based on the quoted price for an option. The differencein-differences spreads are negative and statistically significant for all categories, with the

Wilcoxon Signed-Rank test significant at the 1% level. Average dollar effective spreads

declined by 2.7 cents for options priced below $1.00, by 2.9 cents for options priced between

$1.00 and $3.00, by 4.2 cents for options priced between $3.00 and $5.00, and by 4.6 cents for

options priced over $5.00, vis-à-vis the changes in the control group. The spread declines for

Pilot option classes are consistent across all price categories for all four spread measures.

Table 9 decomposes the sample based on the time to expiration for an option. The

difference-in-differences spreads are negative and statistically significant for all categories, with

the Wilcoxon Signed-Rank test significant at the 1% level. Average dollar effective spreads

declined by 3.4 cents for options expiring within a week, by 3 cents for options expiring between

one and four weeks, and by 3.6 cents for options expiring in more than four weeks, vis-à-vis the

changes in the control group. The spread declines for Pilot option classes are consistent across

all expiration categories for all four spread measures.

Tables 10–12 show results for option classes on ETFs decomposed by addition year,

underlying market capitalization (as measured by price multiplied by shares outstanding), and

option volume. Overall, average dollar effective spreads declined by 5 cents and average dollar

quoted spreads declined by 6.8 cents. When decomposed by the market capitalization of the

underlying ETF or by option trading volume, all difference-in-differences spreads are negative

and statistically significant at the 1% level for the Wilcoxon Signed-Rank test.

The difference-in-differences analysis described above has focused on examining the

impact of the Pilot on effective spreads and quoted spreads. The same technique can also be

applied to examine other market characteristics. For example, analysis showing the impact on

average daily trading volume, average trade size, and quoted depth at the NBBO is presented in

the Appendix.

VI.

Regression Analysis

The difference-in-differences analysis reported in the previous section used a control-

sample methodology to control for any market-wide difference in factors that might have

affected spreads between the pre-event and post-event months. This section presents further

results based on a multivariate regression methodology that further controls for changes in other

factors that might affect spreads, including option volume, stock volume, and equity volatility.

Baseline Regression Specification (Contains All Options)

The baseline multivariate regression approach is estimated at the class-day level for

periods covering a calendar month prior to, and a calendar month subsequent to, each date of

addition to the Pilot and contains all traded option classes:

where

particular day;

is the binary variable of interest, taking a value of 1 in the post-switch

is the log of option class volume;

is the log of underlying stock volume on the

is the underlying stock volatility

calculated as the difference between the high and low prices of the day divided by the closing

price;

is a measure of quoted or effective spreads for a particular option class on a

period for options switching to the Pilot;

day;

is an option class fixed effect; and

The variable of interest

is a date fixed effect.

is the interaction term between indicator variables for the

switching options and the post-switch period. The specification does not include individual

(main effect) indicator variables for switching options and the post-switch period because these

are subsumed by the option class and date fixed effects. The model is estimated on observations

at the class/day level, and the dependent variable is constructed by aggregating spread measures

across the option series within the same class. The spreads are first volume-weighted across

trades within each option series-day and then equally weighted across option series to obtain

option class-day spreads. Hence, this specification incorporates all option series within the

option classes with positive trading volume. The model is estimated on a sample including all

option class/day measures for the calendar month prior to, and the calendar month after, each

date of addition to the Pilot for all options. Standard errors are clustered on option class and

date.

Matched Sample Regression Specification (Contains Only Matched Sample)

We also estimate an alternative specification, which has three key differences with the

baseline specification. First, the alternate specification contains only the matched sample for

each event, where an event constitutes a pre-event and post-event month for an addition.

Second, the variable of interest is a switching indicator (which takes on a value of 1) only in the

post-event month for the switching options. This indicator is zero for the control options,

regardless of whether they are in the Pilot or not. Third, the option class fixed effect is replaced

with an option class event fixed effect, since the same control option may be outside the Pilot for

a particular event and may be in the Pilot for another event.

Findings

Table 13 reports results for effective spreads using the baseline specification estimated

over the pre-event month and post-event month for the full sample, as well as for the different

time periods. The indicator variable of interest is negative and statistically significant (at the 1%

level) for effective spreads (for the full sample and for all time periods, except for the most

recent additions in 2012–2015), showing a decline in trading costs for options being added to the

Pilot. Dollar effective spreads declined by an average of 8.8 cents after addition to the Pilot for

the full sample. Comparing the different subperiods separately, the maximum decline in spreads

is for the additions in the 2010–2011 period (8 cents). The indicator is not statistically

significant for the most recent additions from 2012–2015, with a coefficient close to zero.

Similarly, Table 14 reports results for quoted spreads using the baseline multivariate

regression framework. Dollar quoted spreads declined by 16.1 cents for the entire sample, which

is statistically significant. Comparing the subperiods individually, the decline is strongest in the

2010–2011 subperiod (12.2 cents), and is not statistically significant for additions made in the

2012–2015 subperiod.

Table 15 reports regression results for effective spreads using the matched sample

approach, for the complete time period as well as for different subperiods. Dollar effective

spreads declined by 3.6 cents for the full sample, which is statistically significant at the 1% level.

The coefficient of interest (i.e., switching indicator) is statistically significant and negative across

all specifications, including the 2012–2015 subperiod. Similar results are reported in Table 16

for quoted spreads, with dollar quoted spreads declining by 5.2 cents for the full sample, which is

statistically significant at the 1% level.

Table 17 reports regression results for effective spreads using the matched sample

approach, decomposed into different market capitalization categories. The coefficient of interest

(i.e., switching indicator) is statistically significant and negative across all categories. Similar

results are reported in Table 18 for quoted spreads.

Table 19 reports regression results for effective spreads using the matched sample

approach, decomposed into different categories based on option volume. The coefficient of

interest (i.e., switching indicator) is statistically significant and negative across all categories.

Similar results are reported in Table 20 for quoted spreads.

Table 21 reports regression results for option classes on ETFs using the full sample and

matched sample approach. Similar to the common stock sample, the results show that switching

to the Penny Pilot has a significant (at the 5% level) negative effect on quoted spreads. For

effective spreads, we find similar negative effects (significant at the 1% level), with the

exception of the effective dollar specification for the full ETF sample.

VII.

Impact of Pilot on the Least Liquid Option Classes

Based on the results reported above, we observe a decline in spreads for option classes in

the Pilot vis-à-vis the control options, which is robust across different spread measures, as well

as different partitions based on market capitalization, option volume, etc. This does not

necessarily imply that all option classes would benefit from inclusion in the Pilot because, as

noted above, the Pilot option classes are not representative of the overall population of option

classes, but tend to be the most liquid. Many option classes currently not in the Pilot have lower

trading volume than those currently in the Pilot. It is not clear the extent to which similar

benefits might be realized should penny quoting be expanded to include such option classes.

With the passage of time, some Pilot option classes (which may have had high trading

volume at the time of inclusion) may have become less active, to the point that they would no

longer be selected for the Pilot if the decision were made today. Among the Pilot option classes,

those with the lowest current volume and market capitalization are the most similar to the

options outside the Pilot. Analysis of these options could provide insight into the benefits that

might be experienced by less liquid options upon inclusion.

Table 22 provides characteristics of Pilot option classes as well as matched control

options as of February 2016. Control samples were constructed by matching Pilot options with

non-Pilot options on the basis of option class volume, underlying market capitalization, and

underlying price.6 To ensure robustness, control samples were designed so that the matching

was completed with replacement (i.e., an out-of-Pilot option class can match with any number of

in-Pilot option classes), as well as without replacement (i.e., an out-of-Pilot option class can

match with only one in-Pilot option class). Even with the passage of time, Pilot options are more

liquid than the overall option universe, with the top quartile of Pilot options by option volume in

the top three percentiles of the overall universe. Moreover, even the smallest Pilot option classes

have option volumes greater than the volume of the 40th percentile in the overall option class

Each Pilot option class was matched with a control option outside the Pilot by rank ordering the entire universe of

traded options on option class volume, underlying market capitalization, and underlying price as of February 2016,

and then minimizing the difference in ranks between the Pilot option and the control option.

universe; hence, the identification of the impact of inclusion of option classes below the 40th

percentile is not possible using this sample.

Table 23 analyzes the difference in spreads between the Pilot option classes and the

control options as of February 2016.7 Results are reported by option volume quartiles to analyze

if there is a discontinuity at which point the benefits of being in the Pilot taper off. Analyzing

the results across the different quartiles, the difference in spreads is negative and mostly

statistically significant for the top two quartiles. Hence, consistent with the difference-indifferences analysis, the Pilot appears to be beneficial for the top two quartiles of Pilot options.

The top two quartiles are in the 95th percentile of the overall option universe. Thus, these results

show that the penny tick size is associated with the benefit of narrower spreads for option classes

in the top 5% of option volume.8 Analyzing the difference in spreads for the bottom two

quartiles, the test statistics are negative, but most are not statistically significant, suggesting that

the benefits of the Pilot decline as option liquidity declines. However, the absence of evidence

that market quality has deteriorated suggests that inclusion of lower-volume option classes may

not be not harmful.

VIII. Conclusion

This research has examined data covering the entire history of the Option Penny Pilot

from its inception and has analyzed how inclusion in the Pilot has impacted effective spreads and

Note that this is a difference in spreads specification, rather than the difference-in-differences specification used

previously.

We note two limitations of our analysis–one, the control sample of options have significantly lower volume than

the Penny Pilot options since most of the high volume options are in the Pilot, and two, that our analysis is limited to

liquidity metrics that can be calculated using available data. For example, we are unable to comment on the impact

of the Pilot on overall trading costs faced by institutions with large orders.

quoted spreads for option classes added to the Pilot through January 2016. The results indicate

that inclusion in the Pilot is associated with a statistically significant decline in effective spreads

and quoted spreads. This result is found to hold across all subperiods, but the magnitude of the

effect appears to be smaller for additions after 2012. Results are presented for options with

varying levels of trading volume and underlying market capitalization and appear to be robust

across categories. The results presented above focus on options on common stock, although

similar results are found for ETF options as discussed above. Similar results are found for

complex trades and are presented in the Appendix. The impact of inclusion in the Pilot on other

dimensions of market quality, such as quoted depth, are also presented in the Appendix.

This research further seeks to provide evidence to assist the staff in evaluating whether

the program should be expanded (or contracted) by examining the spreads of Pilot option classes

vis-à-vis matched controls as of February 2016. The results show a benefit of narrower spreads

for option classes in the top 5% of the universe. However, there is less less evidence of benefit

for lower-volume option classes. We find no evidence of a harmful increase in spreads for any

category of option classes.

In summary, the evidence suggests that the existing Pilot program has resulted in

improvements in market quality (as measured by quoted and effective spreads) for the Pilot

option classes. This research does not provide any clear evidence that expanding the Pilot to

include lower-volume stocks would result in similar benefits.

It is worth noting that this research does not account for the different fee structure option

exchanges have for classes within and outside the Pilot. It is common among exchanges to have

separate fee structures for Penny Pilot options. The research summarized herein does not

attempt to disentangle whether the observed improvements in market quality can be attributed to

the Pilot alone, or if they may also be influenced by the differential fee structure. The average

declines in execution costs for the most liquid option classes appear to be substantially higher

than the potential impact of any fee adjustments.

Figure 1

Average Daily Option Class Volume vs. Market Capitalization

February 2016

Log Average

Option Class Volume

Not in Pilot

In Pilot

-1

-2

7.5

Source: LiveVol; OptionMetrics; CRSP.

8.5

9.5

10.5

11.5

Log10 Market Capitalization

Note: Includes all option classes on common stock for which LiveVol has data in February 2016. Share codes 10 and 11 are used to identify common stock in

the CRSP data.

Figure 2

Average Daily Underlying Volume vs. Market Capitalization

February 2016

Log Average

Underlying Volume

8.5

7.5

6.5

5.5

Not in Pilot

In Pilot

4.5

3.5

7.5

Source: LiveVol; OptionMetrics; CRSP.

8.5

9.5

10.5

11.5

Log10 Market Capitalization

Note: Includes all option classes on common stock for which LiveVol has data in February 2016. Share codes 10 and 11 are used to identify common stock in

the CRSP data.

Figure 3

Average Daily Option Class Volume vs. Market Capitalization

Option Classes on ETFs

Log Average

Option Class Volume

February 2016

Not in Pilot

In Pilot

Source: LiveVol; OptionMetrics; CRSP.

Log10 Market Capitalization

Note: Includes all option classes on ETFs for which LiveVol has data in February 2016. Share code 73 is used to identify ETFs in the CRSP data.

Figure 4

Average Daily Underlying Volume vs. Market Capitalization

Option Classes on ETFs

Log Average

Underlying Volume

February 2016

Not in Pilot

In Pilot

Source: LiveVol; OptionMetrics; CRSP.

Log10 Market Capitalization

Note: Includes all option classes on ETFs for which LiveVol has data in February 2016. Share code 73 is used to identify ETFs in the CRSP data.

Figure 5

Impact of Addition to the Penny Pilot on Effective Spreads ($)

Effective Spread

$0.14

Penny Pilot

Addition

$0.12

$0.10

$0.08

$0.06

Switching

Control

$0.04

$0.02

$0.00

Pre-Addition Month

Post-Addition Month

Source: LiveVol; OptionMetrics, CRSP.

Note: Includes 235 option classes on common stock that were added to the Penny Pilot between January 26, 2007 and July 6, 2015 for which

LiveVol has data. 12 option classes were removed because the switching option class or the control option class was missing data on an option

class-day in the pre-addition or post-addition month. Each switching option class is matched with a non-switching option class outside the Pilot

based on underlying market capitalization and option volume. Effective spreads for the switching and control samples are averaged across each

trading day.

Figure 6

Impact of Addition to the Penny Pilot on Effective Spreads (%)

Effective Spread

20%

Penny Pilot

Addition

16%

12%

Switching

Control

Pre-Addition Month

Post-Addition Month

Source: LiveVol; OptionMetrics, CRSP.

Note: Includes 235 option classes on common stock that were added to the Penny Pilot between January 26, 2007 and July 6, 2015 for which

LiveVol has data. 12 option classes were removed because the switching option class or the control option class was missing data on an option

class-day in the pre-addition or post-addition month. Each switching option class is matched with a non-switching option class outside the Pilot

based on underlying market capitalization and option volume. Effective spreads for the switching and control samples are averaged across each

trading day.

Figure 7

Impact of Addition to the Penny Pilot on Quoted Spreads ($)

Quoted Spread

$0.18

Penny Pilot

Addition

$0.15

$0.12

$0.09

Switching

$0.06

Control

$0.03

$0.00

Pre-Addition Month

Post-Addition Month

Source: LiveVol; OptionMetrics, CRSP.

Note: Includes 235 option classes on common stock that were added to the Penny Pilot between January 26, 2007 and July 6, 2015 for which

LiveVol has data. 12 option classes were removed because the switching option class or the control option class was missing data on an option

class-day in the pre-addition or post-addition month. Each switching option class is matched with a non-switching option class outside the Pilot

based on underlying market capitalization and option volume. Quoted spreads for the switching and control samples are averaged across each

trading day.

Figure 8

Impact of Addition to the Penny Pilot on Quoted Spreads (%)

Quoted Spread

21%

Penny Pilot

Addition

18%

15%

12%

Switching

Control

Pre-Addition Month

Post-Addition Month

Source: LiveVol; OptionMetrics, CRSP.

Note: Includes 235 option classes on common stock that were added to the Penny Pilot between January 26, 2007 and July 6, 2015 for which

LiveVol has data. 12 option classes were removed because the switching option class or the control option class was missing data on an option

class-day in the pre-addition or post-addition month. Each switching option class is matched with a non-switching option class outside the Pilot

based on underlying market capitalization and option volume. Quoted spreads for the switching and control samples are averaged across each

trading day.

Figure 9

Impact of Addition to the Penny Pilot on Effective Spreads ($)

Option Classes on ETFs

Effective Spread

$0.16

Penny Pilot

Addition

$0.14

$0.12

$0.10

$0.08

$0.06

Switching

Control

$0.04

$0.02

$0.00

Pre-Addition Month

Post-Addition Month

Source: LiveVol; OptionMetrics, CRSP.

Note: Includes 53 option classes on ETFs that were added to the Penny Pilot between January 26, 2007 and January 5, 2016 for which LiveVol

has data. Two option classes were removed because the switching option class or the control option class was missing data on an option class-day

in the pre-addition or post-addition month. Each switching option class is matched with a non-switching option class outside the Pilot based on

underlying market capitalization and option volume. Effective spreads for the switching and control samples are averaged across each trading day.

Figure 10

Impact of Addition to the Penny Pilot on Effective Spreads (%)

Option Classes on ETFs

Effective Spread

20%

Penny Pilot

Addition

16%

12%

Switching

Control

Pre-Addition Month

Post-Addition Month

Source: LiveVol; OptionMetrics, CRSP.

Note: Includes 53 option classes on ETFs that were added to the Penny Pilot between January 26, 2007 and January 5, 2016 for which LiveVol

has data. Two option classes were removed because the switching option class or the control option class was missing data on an option class-day

in the pre-addition or post-addition month. Each switching option class is matched with a non-switching option class outside the Pilot based on

underlying market capitalization and option volume. Effective spreads for the switching and control samples are averaged across each trading day.

Figure 11

Impact of Addition to the Penny Pilot on Quoted Spreads ($)

Option Classes on ETFs

Quoted Spread

$0.24

Penny Pilot

Addition

$0.21

$0.18

$0.15

$0.12

$0.09

Switching

Control

$0.06

$0.03

$0.00

Pre-Addition Month

Post-Addition Month

Source: LiveVol; OptionMetrics, CRSP.

Note: Includes 53 option classes on ETFs that were added to the Penny Pilot between January 26, 2007 and January 5, 2016 for which LiveVol

has data. Two option classes were removed because the switching option class or the control option class was missing data on an option class-day

in the pre-addition or post-addition month. Each switching option class is matched with a non-switching option class outside the Pilot based on

underlying market capitalization and option volume. Quoted spreads for the switching and control samples are averaged across each trading day.

Figure 12

Impact of Addition to the Penny Pilot on Quoted Spreads (%)

Option Classes on ETFs

Quoted Spread

24%

Penny Pilot

Addition

21%

18%

15%

12%

Switching

Control

Pre-Addition Month

Post-Addition Month

Source: LiveVol; OptionMetrics, CRSP.

Note: Includes 53 option classes on ETFs that were added to the Penny Pilot between January 26, 2007 and January 5, 2016 for which LiveVol

has data. Two option classes were removed because the switching option class or the control option class was missing data on an option class-day

in the pre-addition or post-addition month. Each switching option class is matched with a non-switching option class outside the Pilot based on

underlying market capitalization and option volume. Quoted spreads for the switching and control samples are averaged across each trading day.

Table 1

Option Classes Added to Penny Pilot

Since Inception

Date

New Option Classes Added

(%)

1/26/2007

9/28/2007

3/28/2008

11/2/2009

1/5/2010

2/1/2010

5/3/2010

7/6/2010

8/2/2010

1/4/2011

7/6/2011

1/4/2012

7/3/2012

1/3/2013

7/3/2013

1/3/2014

7/3/2014

1/5/2015

7/6/2015

1/5/2016

7/6/2016

Total

415

3.19%

5.39%

5.64%

18.38%

0.98%

18.38%

0.74%

18.38%

0.49%

0.98%

0.74%

0.49%

2.21%

0.49%

1.72%

0.98%

0.49%

1.47%

0.49%

1.69%

100.00%

Source: ISE Penny Pilot Addition Dates.

Table 2

Liquidity Metrics for Penny Pilot Addition Dates

Option Classes on Common Stock

Daily Volume of Option Class

(Contracts)

Number of Option Classes

Month

Not in Pilot

Switching

Dec-06

2,030

Aug-07

2,011

Feb-08

In Pilot

Not in Pilot

Switching

1,636

32,572

1,843

2,001

Oct-09

1,890

Dec-09

1,950

Jan-10

Daily Volume of Underlying Stock

(Thousands)

In Pilot

Not in Pilot

Switching

–

1,390

26,137

51,330

29,632

1,983

1,481

30,794

48,095

1,177

22,411

755

1,823

Apr-10

1,839

Jul-10

1,767

Dec-10

In Pilot

Market Cap of Underlying Stock

($Millions)

Not in Pilot

Switching

In Pilot

–

6,920

102,942

–

22,195

35,759

6,341

94,746

106,900

1,684

21,527

30,027

4,951

88,186

80,364

65,407

1,596

17,257

39,943

3,144

43,505

68,857

10,000

33,977

1,205

11,925

23,211

3,365

23,018

60,045

866

9,080

44,974

1,260

8,250

30,600

2,719

22,664

57,192

140

652

7,528

35,862

1,281

6,237

23,736

2,881

10,254

49,019

180

457

3,807

20,062

1,042

5,379

15,530

2,361

13,425

36,905

1,790

224

477

4,869

18,171

838

8,813

10,840

2,763

209

36,900

Jun-11

1,824

225

502

15,136

15,679

960

7,886

10,004

2,902

20,199

38,065

Dec-11

1,888

233

381

11,989

14,393

878

4,695

9,122

2,637

5,077

35,131

Jun-12

1,899

233

502

15,814

15,015

976

18,695

9,575

2,781

9,861

38,136

Dec-12

1,885

225

378

40,885

16,414

764

16,823

8,422

2,892

12,502

40,393

Jun-13

1,989

233

486

64,427

15,264

865

9,097

9,000

3,276

13,037

43,755

Dec-13

2,047

226

567

16,306

16,847

773

5,198

7,072

3,809

6,630

50,908

Jun-14

2,144

227

505

123,725

15,879

803

31,263

6,311

3,951

24,170

52,975

Dec-14

2,187

229

437

8,256

16,487

821

5,921

7,014

4,003

26,587

55,218

Jun-15

2,206

227

391

18,684

13,805

830

7,078

6,817

4,035

6,374

55,257

Source: CRSP; OptionMetrics; LiveVol.

Note: Liquidity metrics are shown for the month prior to each addition date. The liquidity metrics includes all option classes on common stock for which LiveVol has data. Share codes 10 and 11 are used to

identify common stock in the CRSP data. The sample excludes option classes for which the underlying stock traded below $1 or above $1,000 on any day on which an average spread is calculated during the

December 2006 through February 2016 period. Data for the July 2010 and January 2016 additions is not shown above because none of the option classes that switched on those dates are in the sample of option

classes on common stock.

Table 3

Liquidity Metrics for Penny Pilot Addition Dates

Matched Pairs Sample

Daily Volume of Option Class (Contracts)

Daily Volume of Underlying Stock (Thousands)

Market Cap of Underlying Stock ($Millions)

Month

Number of Option

Classes

Switching

Unconstrained

Constrained

Switching

Unconstrained

Constrained

Switching

Unconstrained

Constrained

Dec-06

32,572

27,896

26,137

18,426

102,942

76,960

Aug-07

51,330

36,314

36,353

22,195

27,659

18,002

94,746

101,936

85,218

Feb-08

30,794

44,490

22,451

21,527

31,756

16,848

88,186

85,362

65,385

Oct-09

22,411

23,793

11,851

17,257

17,117

9,666

43,505

37,261

25,440

Jan-10

9,080

9,246

6,915

8,250

7,668

5,598

22,664

18,788

14,558

Apr-10

7,990

9,654

5,705

6,271

7,387

7,463

10,010

10,163

9,934

Jul-10

3,807

3,853

3,241

5,379

5,516

3,930

13,425

13,119

11,582

Dec-10

4,869

1,793

8,813

1,993

209

259

Jun-11

15,136

14,755

8,719

7,886

12,246

2,248

20,199

20,167

11,981

Dec-11

11,989

15,117

6,524

4,695

11,795

3,703

5,077

5,099

4,783

Jun-12

15,814

11,767

18,695

6,119

9,861

9,777

Dec-12

40,885

37,142

8,880

16,823

17,370

4,000

12,502

13,173

10,139

Jun-13

64,427

37,025

9,730

9,097

3,079

2,784

13,037

12,438

9,251

Dec-13

16,306

15,206

9,946

5,198

4,131

4,420

6,630

6,930

7,677

Jun-14

123,725

59,340

26,590

31,263

2,402

9,795

24,170

26,476

33,033

Dec-14

8,256

8,265

7,583

5,921

8,410

8,833

26,587

29,393

28,724

Jun-15

18,684

25,054

8,856

7,078

13,617

23,484

6,374

6,202

6,616

Source: CRSP; OptionMetrics; LiveVol.

Note: Liquidity metrics are shown for the month prior to each addition date. The matched pairs sample is constructed from the universe of option classes on common stock for which LiveVol has data. Each

switching option class is matched with a non-switching option class based on underlying market capitalization and option volume. For “constrained” matching, the non-switching class must be outside of the Pilot.

Data for the January 2010 and July 2010 additions are not shown above because the post-month for these two addition dates overlapped with the February 1, 2010 and August 2, 2010 additions respectively. Data

for the January 2016 additions is not shown above because none of the option classes that switched on that date are in the sample of option classes on common stock.

Table 4

Liquidity Metrics for Penny Pilot Addition Dates

Matched Pairs Sample: Option Classes on ETFs

Daily Volume of Option Class (Contracts)

Daily Volume of Underlying Stock (Thousands)

Market Cap of Underlying Stock ($Millions)

Month

Number of Option

Classes

Switching

Unconstrained

Constrained

Switching

Unconstrained

Constrained

Switching

Unconstrained

Constrained

Dec-06

140,034

77,111

63,380

50,208

22,732

21,345

10,224

23,872

24,586

Aug-07

230,200

208,085

101,194

49,351

22,471

9,615

Feb-08

43,739

584,226

22,226

151,137

25,362

16,420

Oct-09

27,803

90,613

6,741

24,030

32,762

8,190

6,404

8,236

2,567

Jan-10

10,953

9,705

8,186

17,443

9,969

5,050

1,811

2,324

2,473

Apr-10

5,283

6,152

2,920

7,414

7,624

3,770

2,368

2,462

2,083

Jul-10

1,727

2,390

1,696

3,552

3,273

2,817

1,190

1,123

1,047

Dec-10

–

Jun-11

–

Dec-11

–

Jun-12

–

Dec-12

15,267

6,892

6,685

7,671

3,712

1,399

726

763

767

Jun-13

–

Dec-13

16,713

10,133

7,237

7,116

3,259

2,045

6,446

6,666

7,653

Jun-14

–

Dec-14

–

Jun-15

10,538

12,301

11,827

7,536

5,672

3,882

5,606

5,249

7,690

Dec-15

8,011

12,763

4,426

1,600

834

3,386

400

340

464

Source: CRSP; OptionMetrics; LiveVol.

Note: Liquidity metrics are shown for the month prior to each addition date. The matched pairs sample is constructed from the universe of ETFs for which LiveVol has data. Each switching option class is matched

with a non-switching option class based on underlying market capitalization and option volume. For “constrained” matching, the non-switching class must be outside of the Pilot. Data for the January 2010 and July

2010 additions are not shown above because the post-month for these two addition dates overlapped with the February 1, 2010 and August 2, 2010 additions respectively.

Table 5

Univariate Statistics Reported by Penny Pilot Addition Date Periods

Unconstrained Matching

Switching

Addition Dates

Average Daily Average Daily

Trade Count

Volume

Pre

Post

Change

Non-Switching

P-Value

Pre

Post

Change

P-Value

Difference-in-Differences

Wilcoxon

Diff-in-Diff T-Test Signed-Rank

Effective Spread ($)

All

247

912

16,940

0.105

0.068

-0.037

0.000

0.096

0.094

-0.002

0.195

-0.035

0.000

2007 – 2009

2010 – 2011

146

1,363

443

30,563

7,181

0.109

0.098

0.066

0.062

-0.042

-0.036

0.000

0.101

0.090

0.093

-0.009

0.002

0.000

0.213

-0.033

-0.038

0.000

2012 – 2015

2,795

33,618

0.144

0.129

-0.015

0.011

0.122

0.120

-0.001

0.779

-0.014

0.019

0.009

Effective Spread (%)

All

247

912

16,940

0.138

0.090

-0.048

0.000

0.113

0.122

0.009

0.004

-0.058

0.000

2007 – 2009

2010 – 2011

2012 – 2015

146

1,363

443

2,795

30,563

7,181

33,618

0.140

0.132

0.179

0.085

0.087

0.138

-0.055

-0.046

-0.041

0.000

0.113

0.111

0.131

0.116

0.126

0.122

0.004

0.015

-0.009

0.252

0.004

0.223

-0.058

-0.060

-0.032

0.000

0.003

0.000

0.006

0.124

0.122

0.118

0.180

0.122

0.113

0.122

0.176

-0.001

-0.009

0.004

-0.004

0.531

0.000

0.203

0.613

-0.051

-0.047

-0.055

-0.040

0.000

0.003

0.000

0.005

0.138

0.130

0.138

0.175

0.149

0.136

0.156

0.165

0.012

0.006

0.018

-0.010

0.002

0.098

0.003

0.275

-0.079

-0.074

-0.083

-0.063

0.000

0.003

Quoted Spread ($)

All

2007 – 2009

2010 – 2011

2012 – 2015

247

146

912

1,363

443

2,795

16,940

30,563

7,181

33,618

0.139

0.136

0.131

0.230

0.087

0.080

0.186

-0.053

-0.056

-0.052

-0.043

0.000

Quoted Spread (%)

All

2007 – 2009

2010 – 2011

2012 – 2015

247

146

912

1,363

443

2,795

16,940

30,563

7,181

33,618

0.173

0.165

0.169

0.256

0.106

0.096

0.104

0.183

-0.067

-0.069

-0.065

-0.073

0.000

Source: LiveVol; OptionMetrics; CRSP.

Note: Each switching option class is matched with a non-switching option class based on underlying market capitalization and option volume. Pilot addition dates 1/26/07, 9/28/07, 3/28/08, and 11/2/09 are

included in the 2007 – 2009 addition date period; Pilot addition dates 2/1/10, 5/3/10, 8/2/10, 1/4/11, and 7/6/11 are included in the 2010 – 2011 period; and Pilot addition dates 1/4/12, 7/3/12, 1/3/13, 7/3/13,

1/3/14, 7/3/14, 1/5/15, and 7/6/15 are included in the 2012 – 2015 period. The average daily volume and average daily trade count are calculated on a per symbol basis and are calculated using the month

prior to the month the option class was added to the Pilot (the "pre" month).

Table 6

Univariate Statistics Reported by Underlying Market Capitalization

Unconstrained Matching

Switching

Market Cap

Percentile

Average Daily Average Daily

Trade Count

Volume

Pre

Post

Change

Non-Switching

P-Value

Pre

Post

Change

P-Value

Difference-in-Differences

Wilcoxon

Diff-in-Diff T-Test Signed-Rank

Effective Spread ($)

95 +

75th – 94th

50th – 74th

Below 50th

123

1,279

811

30,296

12,444

0.102

0.107

0.061

0.071

-0.041

-0.036

0.000

0.085

0.106

0.079

0.106

-0.006

0.000

0.012

0.954

-0.035

0.000

780

449

11,117

8,279

0.106

0.095

0.069

0.066

-0.037

-0.029

0.000

0.001

0.085

0.093

0.084

0.093

-0.001

0.000

0.874

0.941

-0.036

-0.029

0.000

0.003

0.000

0.004

0.098

0.109

0.132

0.160

0.101

0.111

0.158

0.214

0.003

0.002

0.026

0.054

0.281

0.504

0.040

0.083

-0.060

-0.047

-0.077

-0.088

0.000

0.011

0.000

0.001

0.104

0.137

0.116

0.124

0.099

0.139

0.113

0.122

-0.005

0.002

-0.003

0.059

0.479

0.655

0.586

-0.050

-0.053

-0.041

0.000

0.001

0.000

0.002

0.114

0.133

0.167

0.200

0.119

0.136

0.197

0.263

0.005

0.003

0.029

0.062

0.103

0.284

0.048

0.085

-0.078

-0.067

-0.104

-0.112

0.000

0.009

0.000

0.001

Effective Spread (%)

95 +

75th – 94th

50th – 74th

Below 50th

123

1,279

811

780

449

30,296

12,444

11,117

8,279

0.132

0.128

0.163

0.176

0.075

0.084

0.112

0.142

-0.057

-0.045

-0.051

-0.035

0.000

0.020

Quoted Spread ($)

95 +

75th – 94th

50th – 74th

Below 50th

123

1,279

811

780

449

30,296

12,444

11,117

8,279

0.130

0.143

0.151

0.130

0.074

0.092

0.094

0.086

-0.056

-0.051

-0.057

-0.043

0.000

Quoted Spread (%)

95 +

75 – 94

50 – 74th

Below 50th

123

1,279

811

780

449

30,296

12,444

11,117

8,279

0.158

0.162

0.211

0.226

0.085

0.099

0.136

0.176

-0.073

-0.063

-0.074

-0.050

0.000

0.011

Source: LiveVol; OptionMetrics; CRSP.

Note: Each switching option class is matched with a non-switching option class based on underlying market capitalization and option volume. Matched pairs of switching and non-switching option classes

are categorized into market capitalization percentiles using the underlying market capitalization of the switching option class at the end of the month prior to the addition date. Market capitalization

percentiles are defined using the stocks listed on the NYSE. The average daily volume and average daily trade count are calculated on a per symbol basis and are calculated using the month prior to the

month the option class was added to the Pilot (the "pre" month).

Table 7

Univariate Statistics Reported by Option Class Volume

Unconstrained Matching

Switching

Option Volume

Percentile

Average Daily Average Daily

Trade Count

Volume

Pre

Post

Change

Non-Switching

P-Value

Pre

Post

Change

P-Value

Difference-in-Differences

Wilcoxon

Diff-in-Diff T-Test Signed-Rank

Effective Spread ($)

98 +

95th – 97th

90th – 94th

Below 90th

2,461

869

48,820

15,309

0.129

0.105

0.089

0.064

-0.041

0.000

0.102

0.082

0.093

0.077

-0.009

-0.005

0.002

0.011

-0.032

-0.037

0.000

411

162

6,691

2,600

0.098

0.088

0.064

0.058

-0.034

-0.030

0.000

0.101

0.106

0.105

0.108

0.004

0.002

0.178

0.620

-0.037

-0.032

0.000

0.108

0.111

0.100

0.142

0.108

0.122

0.107

0.162

0.000

0.011

0.008

0.020

0.989

0.138

0.021

0.067

-0.044

-0.060

-0.055

-0.072

0.000

0.124

0.105

0.134

0.138

0.115

0.100

0.139

-0.009

-0.005

0.006

0.002

0.007

0.018

0.160

0.748

-0.047

-0.055

-0.044

0.000

0.124

0.135

0.126

0.174

0.125

0.149

0.135

0.197

0.001

0.014

0.009

0.024

0.774

0.096

0.013

0.056

-0.059

-0.083

-0.075

-0.096

0.000

Effective Spread (%)

98 +

95th – 97th

90th – 94th

Below 90th

2,461

869

411

162

48,820

15,309

6,691

2,600

0.136

0.135

0.138

0.144

0.092

0.086

0.090

0.092

-0.044

-0.049

-0.048

-0.052

0.000

Quoted Spread ($)

98 +

95th – 97th

90th – 94th

Below 90th

2,461

869

411

162

48,820

15,309

6,691

2,600

0.165

0.142

0.134

0.117

0.109

0.082

0.085

0.074

-0.057

-0.060

-0.049

-0.042

0.000

Quoted Spread (%)

98 +

95 – 97

90 – 94th

Below 90th

2,461

869

411

162

48,820

15,309

6,691

2,600

0.163

0.171

0.176

0.182

0.105

0.101

0.109

0.110

-0.058

-0.069

-0.067

-0.072

0.000

Source: LiveVol; OptionMetrics; CRSP.

Note: Each switching option class is matched with a non-switching option class based on underlying market capitalization and option volume. Matched pairs of switching and non-switching option classes

are categorized into option class volume percentiles using the option class volume of the switching option class during the month prior to the addition date. Option class percentiles are defined using the

universe of option classes on common stock in the LiveVol data. The average daily volume and average daily trade count are calculated on a per symbol basis and are calculated using the month prior to

the month the option class was added to the Pilot (the "pre" month).

Table 8

Univariate Statistics Reported by Quoted Price

Unconstrained Matching

Switching

Quoted Price

< $1

$1 – $3

$3 – $5

> $5

247

245

234

Average Daily Average Daily

Trade Count

Volume

Pre

Post

Change

0.054

0.075

0.122

0.195

0.027

0.046

0.084

0.145

-0.027

-0.029

-0.039

-0.050

Non-Switching

P-Value

Difference-in-Differences

Wilcoxon

Diff-in-Diff T-Test Signed-Rank

Pre

Post

Change

P-Value

0.044

0.066

0.112

0.189

0.044

0.066

0.115

0.185

0.000

0.004

-0.004

0.761

0.809

0.099

0.291

-0.027

-0.029

-0.042

-0.046

0.000

0.259

0.038

0.030

0.021

0.265

0.038

0.031

0.021

0.006

0.000

0.001

-0.001

0.252

0.899

0.105

0.261

-0.135

-0.018

-0.011

-0.006

0.000

0.055

0.082

0.147

0.247

0.055

0.083

0.151

0.249

0.000

0.001

0.004

0.002

0.789

0.454

0.034

0.742

-0.039

-0.043

-0.060

-0.070

0.000

0.312

0.048

0.039

0.028

0.318

0.049

0.040

0.028

0.006

0.001

0.000

0.272

0.206

0.041

0.846

-0.179

-0.026

-0.016

-0.009

0.000

Effective Spread ($)

322

307

102

148

7,690

5,496

1,447

1,720

0.000

Effective Spread (%)

< $1

$1 – $3

$3 – $5

> $5

247

245

234

322

307

102

148

7,690

5,496

1,447

1,720

0.324

0.044

0.033

0.022

0.195

0.027

0.022

0.016

-0.129

-0.018

-0.010

-0.006

< $1

$1 – $3

$3 – $5

> $5

247

245

234

322

307

102

148

7,690

5,496

1,447

1,720

0.071

0.098

0.166

0.264

0.032

0.056

0.111

0.195

-0.039

-0.042

-0.055

-0.069

< $1

$1 – $3

$3 – $5

> $5

247

245

234

322

307

102

148

7,690

5,496

1,447

1,720

0.401

0.058

0.044

0.031

0.228

0.033

0.029

0.022

-0.173

-0.025

-0.015

-0.009

0.000

Quoted Spread ($)

0.000

Quoted Spread (%)

0.000

Source: LiveVol; OptionMetrics; CRSP.

Note: Each switching option class is matched with a non-switching option class based on underlying market capitalization and option volume. Trades for a given option-class are grouped according to the

bid and ask price. Instances in which the bid and ask fall into separate groups are excluded from the calculation (2.8% of trades). The average daily volume and average daily trade count are calculated on a

per symbol basis and are calculated using the month prior to the month the option class was added to the Pilot (the "pre" month).

Table 9

Univariate Statistics Reported by Expiration Date

Unconstrained Matching

Switching

Expiration Date

< 1 week

1 – 2 weeks

2 – 4 weeks

+4 weeks

247

246

247

Average Daily Average Daily

Trade Count

Volume

Pre

Post

Change

0.090

0.080

0.078

0.110

0.056

0.048

0.047

0.074

-0.034

-0.031

-0.036

Non-Switching

P-Value

Difference-in-Differences

Wilcoxon

Diff-in-Diff T-Test Signed-Rank

Pre

Post

Change

P-Value

0.082

0.073

0.072

0.101

0.083

0.071

0.101

0.001

-0.002

-0.001

0.000

0.838

0.377

0.627

0.938

-0.034

-0.030

-0.036

0.000

0.357

0.260

0.191

0.083

0.364

0.264

0.204

0.087

0.007

0.005

0.014

0.004

0.487

0.594

0.038

0.215

-0.117

-0.130

-0.113

-0.045

0.000

0.104

0.092

0.093

0.130

0.107

0.092

0.091

0.131

0.003

0.000

-0.001

0.001

0.414

0.958

0.512

0.728

-0.049

-0.045

-0.044

-0.053

0.000

0.401

0.300

0.230

0.104

0.422

0.313

0.243

0.109

0.021

0.013

0.005

0.054

0.166

0.059

0.190

-0.152

-0.164

-0.141

-0.063

0.000

Effective Spread ($)

133

146

535

2,242

1,751

2,541

10,412

0.000

Effective Spread (%)

< 1 week

1 – 2 weeks

2 – 4 weeks

+4 weeks

247

246

247

133

146

535

2,242

1,751

2,541

10,412

0.413

0.330

0.246

0.102

0.303

0.205

0.147

0.061

-0.110

-0.125

-0.100

-0.041

< 1 week

1 – 2 weeks

2 – 4 weeks

+4 weeks

247

246

247

133

146

535

2,242

1,751

2,541

10,412

0.118

0.104

0.148

0.071

0.060

0.059

0.095

-0.047

-0.045

-0.052

< 1 week

1 – 2 weeks

2 – 4 weeks

+4 weeks

247

246

247

133

146

535

2,242

1,751

2,541

10,412

0.475

0.388

0.299

0.131

0.344

0.237

0.171

0.073

-0.131

-0.150

-0.128

-0.058

0.000

Quoted Spread ($)

0.000

Quoted Spread (%)

0.000

Source: LiveVol; OptionMetrics; CRSP.

time to an option series' expiration. The average daily volume and average daily trade count are calculated on a per symbol basis and are calculated using the month prior to the month the option class was

added to the Pilot (the "pre" month).

Table 10

Univariate Statistics Reported by Penny Pilot Addition Date Periods

Unconstrained Matching

Option Classes on ETFs

Switching

Average Daily

Average

Trade Count Daily Volume

Addition Dates

All

2007 – 2009

2010 – 2011

2012 – 2015

1,149

1,889

239

766

All

2007 – 2009

2010 – 2011

2012 – 2015

All

2007 – 2009

2010 – 2011

2012 – 2015

All

2007 – 2009

2010 – 2011

2012 – 2015

Pre

Post

Change

39,965

70,848

5,950

13,169

0.115

0.123

0.105

0.111

0.077

0.073

0.061

0.138

-0.037

-0.050

-0.043

0.027

1,149

1,889

239

766

39,965

70,848

5,950

13,169

0.136

0.120

0.146

0.175

0.099

0.088

0.101

0.138

-0.037

-0.032

-0.045

-0.037

1,149

1,889

239

766

39,965

70,848

5,950

13,169

0.154

0.156

0.140

0.183

0.102

0.092

0.078

0.204

-0.052

-0.064

-0.062

0.022

1,149

1,889

239

766

39,965

70,848

5,950

13,169

0.183

0.151

0.192

0.281

0.125

0.104

0.126

0.201

-0.058

-0.046

-0.067

-0.080

Non-Switching

P-Value

Difference-in-Differences

Wilcoxon

Diff-in-Diff T-Test Signed-Rank

Pre

Post

Change

P-Value

0.095

0.099

0.096

0.078

0.107

0.105

0.111

0.106

0.012

0.006

0.016

0.028

0.041

0.420

0.199

0.079

-0.050

-0.056

-0.059

-0.001

0.000

0.983

0.000

0.297

0.117

0.115

0.118

0.121

0.130

0.132

0.119

0.013

0.017

0.014

-0.002

0.002

0.003

0.080

0.858

-0.050

-0.049

-0.059

-0.035

0.000

0.006

0.000

0.031

0.124

0.127

0.112

0.139

0.133

0.146

0.145

0.016

0.008

0.020

0.033

0.032

0.369

0.173

0.087

-0.068

-0.072

-0.082

-0.012

0.000

0.830

0.000

0.297

0.152

0.145

0.154

0.173

0.166

0.162

0.171

0.168

0.014

0.017

0.016

-0.005

0.005

0.004

0.085

0.762

-0.072

-0.064

-0.083

-0.075

0.000

0.009

0.000

0.031

Effective Spread ($)

0.000

0.455

Effective Spread (%)

0.000

Quoted Spread ($)

0.000

0.684

Quoted Spread (%)

0.000

Source: LiveVol; OptionMetrics; CRSP.

prior to the month the option class was added to the Pilot (the "pre" month).

Table 11

Univariate Statistics Reported by Underlying Market Capitalization

Unconstrained Matching

Option Classes on ETFs

Switching

Market Cap

Percentile

Average Daily

Average

Trade Count Daily Volume

Pre

Post

Change

Non-Switching

P-Value

Difference-in-Differences

Wilcoxon

Diff-in-Diff T-Test Signed-Rank

Pre

Post

Change

P-Value

0.081

0.093

0.105

0.085

0.110

0.120

0.004

0.017

0.015

0.519

0.170

0.185

-0.053

-0.058

-0.042

0.016

0.001

0.017

0.000

0.095

0.117

0.131

0.108

0.135

0.141

0.012

0.019

0.009

0.044

0.035

0.176

-0.052

-0.048

0.000

0.104

0.120

0.138

0.105

0.143

0.158

0.001

0.023

0.019

0.888

0.110

0.147

-0.063

-0.080

-0.061

0.007

0.000

0.009

0.000

0.119

0.151

0.173

0.131

0.172

0.183

0.012

0.021

0.010

0.056

0.025

0.283

-0.069

-0.075

-0.072

0.000

Effective Spread ($)

2,213

947

628

103,814

24,294

11,443

0.125

0.102

0.118

0.077

0.061

0.091

-0.049

-0.042

-0.027

2,213

947

628

103,814

24,294

11,443

0.126

0.128

0.149

0.086

0.095

0.110

-0.040

-0.033

-0.039

2,213

947

628

103,814

24,294

11,443

0.158

0.135

0.166

0.096

0.078

0.124

-0.062

-0.057

-0.042

75 +

50 – 74th

Below 50th

0.028

0.000

0.070

Effective Spread (%)

75 +

50 – 74th

Below 50th

0.000

Quoted Spread ($)

75 +

50 – 74th

Below 50th

0.009

0.000

0.040

Quoted Spread (%)

75 +

50th – 74th

Below 50th

2,213

947

628

103,814

24,294

11,443

0.165

0.168

0.206

0.107

0.115

0.144

-0.057

-0.053

-0.063

0.000

Source: LiveVol; OptionMetrics; CRSP.

Note: Each switching option class is matched with a non-switching option class based on underlying market capitalization and option volume. Matched pairs of switching and non-switching option classes

month the option class was added to the Pilot (the "pre" month).

Table 12

Univariate Statistics Reported by Option Class Volume

Unconstrained Matching

Option Classes on ETFs

Switching

Option Volume

Percentile

Average Daily

Average

Trade Count Daily Volume

Pre

Post

Change

Non-Switching

P-Value

Difference-in-Differences

Wilcoxon

Diff-in-Diff T-Test Signed-Rank

Pre

Post

Change

P-Value

0.107

0.075

0.105

0.126

0.081

0.119

0.018

0.006

0.015

0.185

0.361

0.228

-0.055

-0.040

-0.056

0.076

0.000

0.005

0.000

0.104

0.121

0.124

0.116

0.136

0.012

0.015

0.012

0.076

0.038

0.138

-0.048

-0.045

-0.058

0.000

0.133

0.102

0.137

0.155

0.108

0.158

0.022

0.006

0.021

0.200

0.454

0.138

-0.062

-0.060

-0.080

0.096

0.000

0.007

0.000

0.131

0.158

0.162

0.143

0.170

0.180

0.012

0.017

0.087

0.142

0.094

-0.063

-0.067

-0.085

0.000

Effective Spread ($)

2,350

1,118

233

111,135

18,709

5,034

0.128

0.114

0.106

0.091

0.079

0.065

-0.036

-0.035

-0.041

2,350

1,118

233

111,135

18,709

5,034

0.115

0.141

0.149

0.078

0.111

0.103

-0.037

-0.030

-0.046

2,350

1,118

233

111,135

18,709

5,034

0.159

0.162

0.141

0.119

0.108

0.083

-0.040

-0.055

-0.059

95 +

90 – 94th

Below 90th

0.173

0.001

0.000

Effective Spread (%)

95 +

90 – 94th

Below 90th

0.000

Quoted Spread ($)

95 +

90 – 94th

Below 90th

0.215

0.000

Quoted Spread (%)

95 +

90th – 94th

Below 90th

2,350

1,118

233

111,135

18,709

5,034

0.148

0.197

0.196

0.097

0.143

0.128

-0.051

-0.054

-0.067

0.000

Source: LiveVol; OptionMetrics; CRSP.

Note: Each switching option class is matched with a non-switching option class based on underlying market capitalization and option volume. Matched pairs of switching and non-switching option classes

universe of option classes on common stock in the LiveVol data. The average daily volume and average daily trade count are calculated on a per symbol basis and are calculated using the month prior to

the month the option class was added to the Pilot (the "pre" month).

Table 13

Penny Pilot Regression Results[1]

Impact on Effective Spreads

Option Classes on Common Stock

[2]

Dec. 2006 – Dec. 2009

[3]

Jan. 2010 – Aug. 2011

Dec. 2011 – Feb. 2016

Dollar

Percent

Dollar

Percent

Dollar

Percent

Dollar

Percent

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

-0.088**

-0.074**

-0.034**

-0.062**

-0.080**

-0.084**

-0.005

-0.006

(0.000)

(0.821)

(0.434)

0.831**

0.587**

0.410**

0.286**

0.737**

0.305**

0.958**

0.630**

(0.000)

-0.015**

-0.012**

-0.006**

-0.005**

-0.010**

-0.006**

-0.019**

-0.011**

(0.000)

0.004*

-0.005**

0.007**

-0.002

0.010**

0.005**

0.007**

-0.003*

(0.011)

(0.000)

(0.222)

(0.000)

(0.009)

(0.001)

(0.033)

1,331,185

292,670

411,698

626,768

3,833

2,625

2,358

3,078

Full Sample

Parameter

Post-Addition Indicator

[6]

Equity Volatility[7]

Log of Option Volume[8]

Log of Underlying Volume[9]

Number of Class/Days

Number of Classes

[4]

[5]

Sources: LiveVol; CRSP; OptionMetrics; ISE Penny Pilot Addition Dates.

[1] This table reports results from a panel regression estimating the effects of switching to trading in pennies on effective spreads in the options market. The sample includes data for all

options on common stock that traded in the calendar month before and calendar month after the Penny Pilot addition dates. The dependent variable is the average effective spread. The

average spread is calculated by (1) calculating the volume weighted spread for an option series-day and then (2) equally weighting across option series to calculate the average spread for

an option class-day. Canceled trades and trades that were executed during a locked or crossed market are excluded. Options with non-standard settlement types are excluded. Only trades

with a condition code in 0, 18, 95, and 106 are included. Condition codes 0, 18, 95, and 106 correspond to regular trades and intermarket sweep orders. Effective spreads that are greater

than $5 when the price is less than $100 and effective spreads that are greater than $10 otherwise are excluded from the average effective spread calculation. If the underlying stocks traded

below $1 or above $1,000 on any day for which an average effective spread is calculated, the entire option class is removed from the sample. Date and option class fixed effects are used.

Standard errors are estimated by clustering on option class and date. The "Permno" in CRSP is used to identify option classes 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] The model is estimated over the time period from December 1, 2006 to February 29, 2016. The sample includes 38 calendar months of data.

[3] Pilot addition dates 1/26/07, 9/28/07, 3/28/08, and 11/2/09 are included in the December 2006 – December 2009 period.

[4] Pilot addition dates 2/1/10, 5/3/10, 8/2/10, 1/4/11, and 7/6/11 are included in the January 2010 – August 2011 period.

[5] Pilot addition dates 1/4/12, 7/3/12, 1/3/13, 7/3/13, 1/3/14, 7/3/14, 1/5/15, and 7/6/15 are included in the December 2011 – February 2016 period.

[6] An indicator variable that equals one in the post-addition period and zero otherwise.

[7] The underlying stock volatility as measured by the high price minus the low price divided by the closing price.

[8] The natural log of the option class volume. The option class volume excludes canceled trades and only includes trades with a condition code in 0, 18, 95, and 106.

[9] The natural log of the volume for the underlying stock.

Table 14

Penny Pilot Regression Results[1]

Impact on Quoted Spreads

Option Classes on Common Stock

[2]

Dec. 2006 – Dec. 2009

[3]

Jan. 2010 – Aug. 2011

Dec. 2011 – Feb. 2016

Dollar

Percent

Dollar

Percent

Dollar

Percent

Dollar

Percent

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

-0.161**

-0.137**

-0.053**

-0.091**

-0.122**

-0.119**

-0.037

-0.049**

(0.000)

(0.239)

(0.000)

0.907**

0.705**

0.441**

0.331**

0.812**

0.360**

1.048**

0.769**

(0.000)

-0.026**

-0.021**

-0.008**

-0.015**

-0.012**

-0.034**

-0.023**

(0.000)

0.009**

-0.005**

0.008**

-0.002

0.014**

0.005*

0.014**

-0.005*

(0.000)

(0.001)

(0.000)

(0.204)

(0.000)

(0.025)

(0.000)

(0.023)

1,331,123

292,668

411,691

626,715

3,832

2,625

2,358

3,077

Full Sample

Parameter

Post-Addition Indicator

[6]

Equity Volatility[7]

Log of Option Volume[8]

Log of Underlying Volume[9]

Number of Class/Days

Number of Classes

[4]

[5]

Sources: LiveVol; CRSP; OptionMetrics; ISE Penny Pilot Addition Dates.

[1] This table reports results from a panel regression estimating the effects of switching to trading in pennies on quoted spreads in the options market. The sample includes data for all

options on common stock that traded in the calendar month before and calendar month after the Penny Pilot addition dates. The dependent variable is the average quoted spread. The

average spread is calculated by (1) calculating the volume weighted spread for an option series-day and then (2) equally weighting across option series to calculate the average spread for

an option class-day. Canceled trades and trades that were executed during a locked or crossed market are excluded. Options with non-standard settlement types are excluded. Only trades

with a condition code in 0, 18, 95, and 106 are included. Condition codes 0, 18, 95, and 106 correspond to regular trades and intermarket sweep orders. Quoted spreads that are greater

than $5 when the price is less than $100 and quoted spreads that are greater than $10 otherwise are excluded from the average quoted spread calculation. If the underlying stocks traded

below $1 or above $1,000 on any day for which an average quoted spread is calculated, the entire option class is removed from the sample. Date and option class fixed effects are used.

Standard errors are estimated by clustering on option class and date. The "Permno" in CRSP is used to identify option classes 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] The model is estimated over the time period from December 1, 2006 to February 29, 2016. The sample included 38 calendar months of data.

[3] Pilot addition dates 1/26/07, 9/28/07, 3/28/08, and 11/2/09 are included in the December 2006 – December 2009 period.

[4] Pilot addition dates 2/1/10, 5/3/10, 8/2/10, 1/4/11, and 7/6/11 are included in the January 2010 – August 2011 period.

[5] Pilot addition dates 1/4/12, 7/3/12, 1/3/13, 7/3/13, 1/3/14, 7/3/14, 1/5/15, and 7/6/15 are included in the December 2011 – February 2016 period.

[6] An indicator variable that equals one in the post-addition period and zero otherwise.

[7] The underlying stock volatility as measured by the high price minus the low price divided by the closing price.

[8] The natural log of the option class volume. The option class volume excludes canceled trades and only includes trades with a condition code in 0, 18, 95, and 106.

[9] The natural log of the volume for the underlying stock.

Table 15

Penny Pilot Regression Results[1]

Impact on Effective Spreads

Unconstrained Matching[2]

Full Sample[3]

Parameter

Post-Switch Indicator

[7]

Equity Volatility[8]

Log of Option Volume[9]

Log of Underlying Volume[10]

Number of Class/Days

Number of Classes

Dec. 2006 – Dec. 2009[4]

Jan. 2010 – Aug. 2011[5]

Dec. 2011 – Aug. 2015[6]

Dollar

Percent

Dollar

Percent

Dollar

Percent

Dollar

Percent

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

-0.036**

-0.058**

-0.034**

-0.059**

-0.040**

-0.060**

-0.011*

-0.032**

(0.000)

(0.036)

(0.001)

0.360**

-0.200

0.286**

-0.194**

0.391**

-0.251

0.362**

0.158

(0.000)

(0.110)

(0.000)

(0.009)

(0.000)

(0.196)

(0.002)

(0.206)

0.002**

-0.001

0.002**

0.000

0.002*

-0.002

0.001

0.016**

(0.004)

(0.439)

(0.008)

(0.959)

(0.022)

(0.234)

(0.653)

(0.005)

0.008**

0.009

0.010**

0.009*

0.006**

0.010

0.018**

-0.018

(0.000)

(0.074)

(0.000)

(0.027)

(0.001)

(0.148)

(0.001)

(0.236)

21,025

7,368

12,369

1,288

494

170

292

Sources: LiveVol; CRSP; OptionMetrics; ISE Penny Pilot Addition Dates.

[1] This table reports results from a panel regression estimating the effects of switching to trading in pennies on effective spreads in the options market. The sample includes data for options

on common stock that traded in the calendar month before and calendar month after the Penny Pilot addition dates. The dependent variable is the average effective spread. The average

spread is calculated by (1) calculating the volume weighted spread for an option series-day and then (2) equally weighting across option series to calculate the average spread for an option

class-day. Canceled trades and trades that were executed during a locked or crossed market are excluded. Option classes on common stock are identified using share code 10 and 11 in the

CRSP data. Options with non-standard settlement types are excluded. Only trades with a condition code in 0, 18, 95, and 106 are included. Condition codes 0, 18, 95, and 106 correspond to

regular trades and intermarket sweep orders. Effective spreads that are greater than $5 when the price is less than $100 and effective spreads that are greater than $10 otherwise are

excluded from the average effective spread calculation. If the underlying stocks traded below $1 or above $1,000 on any day for which an average effective spread is calculated, the entire

option class is removed from the sample. Date and option event fixed effects are used. Standard errors are estimated by clustering on option class-event and date. The "Permno" in CRSP is

used to identify option classes 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] Each switching option class is matched with a non-switching option class based on underlying market capitalization and option volume.

[3] The model is estimated over the time period from December 1, 2006 to August 31, 2015.

[3] Pilot addition dates 1/26/07, 9/28/07, 3/28/08, and 11/2/09 are included in the December 2006 – December 2009 period.

[4] Pilot addition dates 2/1/10, 5/3/10, 8/2/10, 1/4/11, and 7/6/11 are included in the January 2010 – August 2011 period.

[5] Pilot addition dates 1/4/12, 7/3/12, 1/3/13, 7/3/13, 1/3/14, 7/3/14, 1/5/15, and 7/6/15 are included in the December 2011 – August 2015 period.

[7] An indicator variable that equals one in the post-switch period and zero otherwise.

[8] The underlying stock volatility as measured by the high price minus the low price divided by the closing price.

[9] The natural log of the option class volume. The option class volume excludes canceled trades and only includes trades with a condition code in 0, 18, 95, and 106.

[10] The natural log of the volume for the underlying stock.

Table 16

Penny Pilot Regression Results[1]

Impact on Quoted Spreads

Unconstrained Matching[2]

[3]

Dec. 2006 – Dec. 2009

[4]

Jan. 2010 – Aug. 2011

Dec. 2011 – Feb. 2015

Dollar

Percent

Dollar

Percent

Dollar

Percent

Dollar

Percent

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

-0.052**

-0.078**

-0.048**

-0.075**

-0.056**

-0.082**

-0.036**

-0.062**

(0.000)

0.393**

-0.195

0.274**

-0.228**

0.439**

-0.247

0.405*

0.313*

(0.000)

(0.161)

(0.002)

(0.003)

(0.000)

(0.253)

(0.011)

(0.021)

0.000

-0.004*

0.001

-0.001

0.000

-0.004*

-0.002

0.003

(0.698)

(0.019)

(0.256)

(0.452)

(0.795)

(0.023)

(0.759)

(0.600)

0.011**

0.011*

0.014**

0.011**

0.008**

0.012

0.029**

-0.009

(0.000)

(0.032)

(0.000)

(0.009)

(0.000)

(0.105)

(0.000)

(0.321)

21,025

7,368

12,369

1,288

494

170

292

Full Sample

Parameter

Post-Switch Indicator

[7]

Equity Volatility[8]

Log of Option Volume[9]

Log of Underlying Volume[10]

Number of Class/Days

Number of Classes

[5]

[6]

Sources: LiveVol; CRSP; OptionMetrics; ISE Penny Pilot Addition Dates.

[1] This table reports results from a panel regression estimating the effects of switching to trading in pennies on quoted spreads in the options market. The sample includes data for options

on common stock that traded in the calendar month before and calendar month after the Penny Pilot addition dates. The dependent variable is the average quoted spread. The average

spread is calculated by (1) calculating the volume weighted spread for an option series-day and then (2) equally weighting across option series to calculate the average spread for an option

class-day. Canceled trades and trades that were executed during a locked or crossed market are excluded. Option classes on common stock are identified using share code 10 and 11 in the

CRSP data. Options with non-standard settlement types are excluded. Only trades with a condition code in 0, 18, 95, and 106 are included. Condition codes 0, 18, 95, and 106 correspond to

regular trades and intermarket sweep orders. Quoted spreads that are greater than $5 when the price is less than $100 and quoted spreads that are greater than $10 otherwise are excluded

from the average quoted spread calculation. If the underlying stocks traded below $1 or above $1,000 on any day for which an average quoted spread is calculated, the entire option class is

removed from the sample. Date and option event fixed effects are used. Standard errors are estimated by clustering on option class-event and date. The "Permno" in CRSP is used to identify

option classes 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.