The Effects of Make and Take Fees in Experimental Markets

Vince Bourke and

David Porter

Economic Science Institute

Chapman University
Abstract: We conduct a series of experiments to examine the effects of the make and take fee structure
currently used by equity exchanges in the U.S. We examine the effects of these fees on measures of
market quality (allocative efficiency, trading volume, book depth, and the bid-ask spread). With the
exception of increased book depth, we document no significant effects of make and take fees relative to a
baseline case in which trading fees are assessed on both sides of a transaction.
Keywords: make and take fees, double auction, experimental economics

JEL Classification: G2, G12, G14

I. Introduction & Literature Review

Formally, the Securities and Exchange Commission (SEC) offer the following description of make and take
fees: non-marketable, resting orders that offer (make) liquidity at a particular price receive a liquidity rebate if
they are executed, while incoming orders that execute against (take) the liquidity of resting orders are charged an
access fee (SEC 2010). Throughout this paper, we refer to the liquidity rebate as the make fee/rebate and the
access fee as the take fee. Traders can be divided into two groups market makers and market takers. A market
maker posts non-marketable (i.e. limit) orders1. These orders are beneficial to an equity exchange because they
create opportunities for other traders to trade on that exchange. A market taker posts marketable (i.e. market)
orders that are executed immediately at the prevailing market price, removing opportunities for other traders to
trade on the exchange. It is clear that for a transaction to occur, both a market maker and a market taker are
required. Market makers receive the make rebate when a market taker hits the makers order. The market taker is
charged the take fee. The take fee is generally greater in absolute value than the make rebate. Thus, the difference

A limit order is considered marketable if is a buy (sell) limit order at a price above (below) the prevailing sell (buy) limit
order.
1

between the two represents revenue for the exchange for facilitating the order. This difference is called the net, or
total fee.
Make and take fees have existed in some form since the late 1990s. When the SEC adopted their Order
Execution Guidelines, Electronic Communication Networks (ECNs) responded by requiring participants to pay to
access order flow. The Island ECN is the first recorded exchange to use make and take fees. They first used them in
1997, and saw their market share of shares listed on the NASDAQ exchange increase from three percent in 1997 to
thirteen percent in 1999 (Kalcheva et al.). In 2007, the SECs implemented the Regulation National Market System
(Reg. NMS). The idea was to have a market composed of multiple trading venues all linked together via rules over
access and trade priority (OHara). This created competition between exchanges for trading volume, and make and
take fees were adopted this is the prevailing trading fee mechanism used by all registered U.S. equity exchanges.
However, the fee structure has seen more widespread attention in the past few years Michael Lewis questioned
the practice of paying market makers to inject liquidity into a system in his book Flash Boys. In a June 17, 2014
hearing of the Senate Permanent Subcommittee on Investigations, NYSE president Thomas Farley said the
exchange is "seeking support for the elimination of maker-taker (fees)."2 NASDAQs CEO, Robert Greifield has
called for a closer examination of the fee structure, and although he does not favor its elimination, he has stated
that the magnitude of fees currently charged is too high.3 In April 2014, the commissioner of the Securities and
Exchange Commission suggested that the commission entertain a pilot program in which maker-taker rebates
would be temporarily prohibited for certain securities.4
Critics of the make and take fee structure point to the misaligned incentives between brokers and their
clients. Specifically, brokers may elect to route orders to exchanges with more a more favorable make and take fee
structure without regard to order execution quality. Angel, Harris, and Spatt (2011) published a critique of the
make and take fee structure in their report on equity trading in the 21st century. They claim that make and take
fees have distorted order routing decisions, aggravated agency problems among brokers and their clients,

http://www.reuters.com/article/2014/06/17/senate-hearing-markets-idUSL2N0OY0I520140617
http://www.reuters.com/article/2014/07/28/nasdaq-omx-congress-marketsrules-idUSL2N0Q31FH20140728
4
http://blogs.barrons.com/focusonfunds/2014/04/03/study-maker-taker-trading-rebates-secs-aguilarsays/?mod=BOLBlog
2
3

unleveled the playing field among dealers and exchange trading systems, produced fraudulent trades, and
produced quoted spreads that do not represent actual trading costs. Intuitively, consider an exchange in which all
participants are market makers. What would happen on this exchange? The answer is nothing if participants
place nothing but non-marketable orders, no trades will occur. If both participants (market maker and market
taker) are required for a transaction to occur, why is only one side incentivized to trade? Battalio, Corwin, and
Jennings (2013) empirically concluded that several large retail brokerages route their order flow to maximize
order flow payments. Specifically, brokers route limit orders to exchanges that offer the highest rebates, which
may be inconsistent with the their responsibility to obtain the best possible execution for an order. They analyzed
a proprietary data set and the NYSEs Trade and Quote data to demonstrate the negative relationship between limit
order execution quality and the relative level of liquidity fees. We do not comment directly on the controversial
nature of the fee structure rather, we demonstrate that it does not necessarily affect trader welfare in our
experimental setup.
The NYSEs fee structure is complicated and is described in their 21-page Schedule of Fees and Charges for
Exchange Services. More frequent traders receive more favorable considerations. For example, NYSE defines a
Tier 1 trader or firm as one who provide[s] liquidityan average daily share volume per month of 0.70% or
more of the U.S. CADV (defined as the consolidated average daily volume). A Tier 1 trader will receive a make
rebate of 23 30 cents per hundred shares and is assessed a take fee of 28 30 cents per hundred shares,
depending on the tape on which the share is listed. Thus, for every Tier 1 trade, the NYSE receives 0 7 cents per
hundred shares. The several tiers beneath Tier 1 have less favorable considerations. At the Basic Rates level (this
would most likely apply to individual, non-institutional investors), the make rebate is fixed at 20 cents per hundred
shares and the take fee is fixed at 30 cents per hundred shares. Thus, for every Basic Rates trade, the NYSE
receives 10 cents per hundred shares. From January 2014 July 2014, approximately 147 trillion shares were
traded on the NYSE so it is clear that the make and take fee structure provides significant revenue for the exchange.
This paper contributes to the ongoing discussion of this fee structure by answering some of the
fundamental questions regarding its effect. We examine four dependent variables efficiency, trading volume, bidask spread, and book depth in a baseline case and two treatments a trading fee treatment (TF) in which both
3

sides of a transaction are assessed a trading fee and a make and take treatment (MT) in which market makers
receive a rebate and market takers are assessed a fee, compared to a baseline case (B) with no fees assessed to
either side of a transaction. To ensure that our results are endogenously derived, the parameters for our
treatments are such that two times the trading fee in the baseline case is equal to the level of net fee in the make
and take case. We are most interested in the comparison between TF and MT as make and take fees supplanted
trading fees on exchanges. We consider two central questions for each variable that will allow us to paint a picture
of overall market quality.
1. Is there a directional effect on efficiency, trading volume, bid-ask spread, and book depth in the presence
of make and take fees relative to a baseline case in which both sides of a transaction are assessed a fee?
2. In the presence of make and take fees, does the allocation of net fee between market maker and market
taker have an effect?
Trading volume and the bid-ask spread have received some attention in the empirical literature. The bidask spread in particular has been a point of contention. Colliard and Foucault (2012) develop a theoretical model
of the fee structure and show that the bid-ask spread adjusted for take fees (cum-fee) increases in the net fee.
Second, the raw (or quoted) bid-ask spread can increase or decrease when the net fee is reduced, depending on
whether the reduction in net fee is achieved by lowering the make fee (this causes a decrease) or by lowering the
take fee (this causes an increase). Malinova and Park (2013) examined a proprietary data set from a 2005 trial
period on the Toronto Stock Exchange (TSX) and found that under a make and take fee regime, the quoted
(effective) spreads did not change, but the cum-fee effective spreads declined. Finally, Kalcheva et al. (2014)
analyzed a larger sample of all registered equity exchanges from 2008 2010 and were unable to document a
significant relationship between bid-ask spread and make and take fees. For trading volume, Kalcheva et al. (2014)
find that an increase in the level of an exchanges net fee is associated with a reduction in trading volume relative
to other exchanges. That is, if one exchange raises its net fee (through an increase in the take fee, a decrease in the
make rebate, or a combination of both), traders may eschew that exchange in favor of another exchange with a
lower net fee. They also find that an increase in the take fee decreases trading volume more than a decrease in the
make rebate.
4

Given these ambiguities, our design offers a model in which theoretical predictions are non-directional for
three of our dependent variables in TF and MT. For trading volume, our model predicts an equilibrium quantity of
twenty units traded in both TF and MT. Theoretical efficiency is the same in both treatments. The bid-ask spread
should remain equal the following simple example will illustrate. In the absence of any fee structure, a buyer
has a valuation of and a seller has a cost of for one unit of an indivisible good. If we assume that traders
internalize the effects of fees, we would expect to see a relative increase in spreads. For example, suppose an
exchange offers a make rebate of and assesses a take fee of , where || > || and = , where is the net
fee (this is positive from the exchanges perspective and negative from the traders perspective). The hypothetical
buyer enters the market first and places a limit order. She will be willing to post a price greater than her valuation,
since she will receive the make rebate of upon execution. Thus, she will post a price in the interval [0, + ].
The hypothetical seller enters the market second and observes the buyers limit order. To complete a trade, she
must place a market order. However, she is only willing to post a price greater than her cost, since she will incur
the take fee of upon execution. Thus, she will post a price in the interval [ + , ]. When = , this implies
that the spread will never be less than ( + ) ( + ) = . Suppose instead that an exchange charges
trading fees to both sides of a transaction equal to 12. In this case, the buyer will post a price in the interval [0,
] and a seller will post a price in the interval [ + 12, ]. When = this implies that the spread will never be

1
2

less than ( + 12) ( 12) = = , which is equal to the implied minimum spread in the presence of make
and take fees. It is also clear that the bid-ask spread should increase with an increase in the level of net fee.
Foucault et al. (2013) develop a theoretical model of profit maximization and market monitoring intensity.
Their model provides testable implications for book depth and trading volume. In their model,
1. The aggregate monitoring level of both sides increases in the number of participants on either side, and
decreases in monitoring costs and in the fee charged on either side.
2. The trading rate decreases in the monitoring costs and trading fees, and increases in the number of
participants on either side.

The aggregate monitoring level decreases in monitoring costs. We make two assumptions. First, the
aggregate monitoring level is a proxy for order flow as placing orders is assumed to be costless in our model. There
are no exogenous monitoring costs, but it is clear that there are monitoring benefits for both market makers and
market takers in the presence of make and take fees. For market makers, we expect to see an increase in
monitoring intensity (and order flow) with MT as makers compete to earn the make rebate. This competition
between makers provides more opportunities for market takers to trade. Market takers will not necessarily place
more orders, but they are incentivized to monitor the market more closely to take advantage of the makers price
setting competition. We also have a theoretical foundation for the empirical results from Kalcheva et al. for trading
volume that is, the trading rate decreases in trading fees. While this hypothesis tells us nothing about trading
volume in MT vs. TF, it does imply trading volume will decrease with an increase in the level of net fee. We
parameterize the net fee accordingly (discussed in Section II) to directly test this hypothesis.
The paper proceeds as follows. Section II describes the model and experimental design, including a detailed
description of the treatments, how each of the statistics were calculated, and our hypotheses. Section III contains
results and a discussion. Section IV concludes and offers some interesting avenues for extensions to our model and
future research.
II. Model & Experimental Design
Sixteen experimental sessions were conducted at a major North American university. In each eight-subject
session, four subjects were assigned to the role of buyer and four subjects were assigned to the role of seller.
Subjects were seated at computer terminals in a laboratory and were given role-specific instructions5. The
instructions were self-paced and required subjects to complete examples of the tasks that they would perform
during the experimental trading periods6. Subjects earned an average of $30.60 for a 1.5 hour session (including a
$7.00 show-up bonus). At the end of the session, subjects completed a short questionnaire that asked them to
describe the strategies that they used within and between market periods.

5
6

Refer to Appendix B for a copy of the instructions.

Refer to Appendix C for a diagram of the user interface.

A nonstationary double auction environment similar to the baseline model in Smith et al. (1992) was
selected to model this institution. The double auction is used to model stock markets given its operational
simplicity, efficiency, and its capacity to respond quickly to changing market conditions (Gjerstad and Dickhaut
(1998)). In each period, a value for an random variable was drawn uniformly from [-50, -40, -30, , 50].
Buyers had resale values randomly drawn from the following inverse demand function:
140 + ,
130 + ,
120 + ,
() = 110 + ,
100 + ,
90 + ,
{ 80 + ,

0<
5<
10 <
15 <
20 <
25 <
30 <

5
10
15
20 (50, 50).
25
30
35

Similarly, sellers had costs randomly drawn from the following inverse supply function.

60 + ,
70 + ,
80 + ,
90 + ,
() =
100 + ,
110 + ,
{120 + ,

0<
5<
10 <
15 <
20 <
25 <
30 <

5
10
15
20 (50, 50).
25
30
35

Thus, buyer has a vector of induced values = (1 , 2 , , ) that are monotonically decreasing such
that < 1 . Similarly, seller has a vector of induced costs = (1 , 2 , , ) that are monotonically
increasing such that > 1 . In the absence of fees (B), the profit for trade (1 , 2 , , ) that matches a
buyer and seller is:
=

In the presence of trading fees (TF), the profit for trade (1 , 2 , , ) that matches a buyer and seller is:

= 12

= 12
7

In the presence of make and take fees (MT), the profit for trade (1 , 2 , , ) that matches a market making
buyer and market taking seller is:

= +

Figure 1 illustrates the normalized ( = 0) structure of supply and demand:

Dollars
180

170
160
150
140
130
120
110
100

CE Price (0 = 100)

90
80
70
60
50
40

D
5

10

15

20

25

30

35

Units

Figure 1: supply and demand curves for x= 0

We linearly perturbed the inverse demand and inverse supply curves between periods. This prevented
subjects from discovering the underlying equilibrium prices. For example, if = 30, all of the demand and supply
steps in Figure 1 would be shifted up by 30, making the competitive equilibrium price 130. Table 1 below displays
values of , the random variable drawn for each trading period.
Period

1
-30

2
50

3
-40

4
50

5
0

6
-40

7
8
9
10
-10 -20 40 -20
Table 1: values of x

11
-50

12
-50

13
40

14
20

15
10

16
-20

17
-30

18
-50

Subjects participated in multiple trading periods of various treatments:

Baseline: This is a nonstationary double auction with no fee structure exogenously imposed on subjects.
TF1/2 (DA w/ Trading Fees Treatment #1/2): This is a nonstationary double auction with a fee structure
exogenously imposed on subjects. Both sides of each transaction are assessed an identical fee. The levels of
net fee per transaction match those in the make and take fee treatments, which enables direct comparison
between the treatments.

MT1/2/3 (DA w/ Make & Take Fees - Treatment #1/2/3): This is a nonstationary double auction with a
make & take fee structure exogenously imposed on subjects. MT2 and MT3 increase the level of net fee

relative to MT1 the distinction lies in how the increase is achieved. In MT2, the rebate is lowered. In MT3,
the fee is increased. This creates a method to compare whether the allocation of fee between maker and
taker has an effect on our dependent variables. The fees for each treatment are given below in Table 2.

0
0.5
1
3
3
4

0
0
0
2
1
2

0
0
0
1
2
2

0
1
2
1
2
2

Table 2: fee structure for all treatments

The fee levels that we selected for our experiments is similar to the fee structures in place on U.S. equity
exchanges. As discussed earlier, the fees assessed by exchanges are on the order of 20 30 cents per hundred
shares traded (or, 0.2 0.3 cents per share). In our experiment, a take fee of 3 represents a fee of three U.S. cents
(conversion rate = $0.01 per experimental dollar earned). Thus, for purposes of generalizability, we assume that
each share traded in our experiment reflects a block of 10 shares.
To illustrate the supply and demand schedules in MT, refer to Figure 2. Assume that in this market, all
buyers are market makers and all sellers are market takers without loss of generality. Recall that a market making
buyer is willing to post a price in the interval [0, + ]. Thus, each step on the demand curve is shifted up by .
Similarly, a market taking seller is willing to post a price in the interval [ + , ]. Accordingly, each step on the
9

supply curve is shifted up by . Figure 2 plots the shifted demand and supply curves (D and S, respectively) and
the original demand and supply curves (D and S). In D and S, the equilibrium price is 100 and the equilibrium
quantity is contained in the interval [20, 25]. D and S now intersect vertically, so the equilibrium price is
contained in the interval [100 + , 100 + ] and the equilibrium quantity is 20.

Dollars
180
170
160
150
140 + m
140
130 + m
130
120 + m
120
110 + m
110
100 + t

S'

100 + m

D'

100

S/D

90+t
90
80+t
80
70+t
70
60+t
60
50
40

10

15

20

25

30

Units

Figure 2: supply and demand curves for x= 0 with MT

We used a crossover design that controlled for treatment order effects. Table 3a illustrates the crossover
design that was used during the first twelve sessions. Data were collected from the grayed periods. Since there are
three treatments (MT1, MT2 and MT3), we permuted the orderings so that each treatment was in each of the
blocks twice across the 12 experiments. The baseline (B) treatment in this case was a double auction with no
trading fees.
10

Period
Treatment

1
B

2
B

3
B

4
MT

5
MT

6
MT

7
B

8
B

9
B

10
MT

11
MT

12
MT

13
B

14
B

15
B

16
MT

17
MT

18
MT

Table 3a: treatment order (MT & B sessions)

Table 3b illustrates the design of the remaining four sessions, where we implemented the relevant baseline case
(TF1 and TF2). Again, data were collected from the grayed periods.
Period
Treatment

1
B

2
B

3
B

4
TF

5
TF

6
TF

7
TF

8
TF

9
TF

10
B

11
B

12
B

13
TF

14
TF

15
TF

16
TF

17
TF

18
TF

Table 3b: treatment order (TF & B sessions)

We now summarize our hypothesis for our two research questions.
1. Is there a directional effect on efficiency, trading volume, bid-ask spread, and book depth in the presence of
make and take fees relative to a baseline case in which both sides of a transaction are assessed a fee?
Our null hypothesis is that there will be no significant difference in trading volume, bid-ask spread, or
system efficiency between MT and TF. We do expect an increase in book depth in MT relative to TF. Our
hypothesis are testable from results derived from the empirical literature and our theoretical predictions.
2. In the presence of make and take fees, does the allocation of net fee between market maker and market taker
have an effect?
Our null hypothesis is that there will be directional effects for trading volume and bid-ask spread. For
trading volume, the empirical literature suggests that efficiency and trading volume will decrease with an
increase in the level of net fee (MT1 MT2 & MT3). Further, an increase in the take fee (MT3 MT1)
should decrease trading volume more than a decrease in the make rebate (MT2 MT1). The bid-ask
spread should increase with an increase in the level of net fee. We do not expect a significant difference in
book depth or efficiency in the MT treatments. The following describes the methodology for each of the
statistics captured during the sessions.
11

Bid-Ask Spread/Trading Volume

Bid-Ask spread was calculated as a time-weighted average in each period:

=1 ( )

100+

where is the bid-ask

spread since the last bid, ask, or contract (this is undefined when no bid or ask is posted), is the time (in
seconds) elapsed since the last bid, ask, or contract, is the duration of the market period (in seconds), is the
value of the random variable drawn in each period, and is the number of contracts formed per market period (in
addition, it is our statistic for trading volume). To match the empirical literature, we will calculate the bid-ask
spread in two ways the quoted spread (simply the difference between the best bid and best ask), and the true
economic net-of-fees spread (the spread that an investor pays on a liquidity-demanding round trip transaction in
the presence of fees calculated as the quoted spread plus twice the take fee per share).

Efficiency

In each period, efficiency is calculated as =

= . Realized Surplus is the total value

minus costs of trades made in the market. The Maximum Surplus is the maximum theoretical gains from exchange
(values minus costs) in the system.

Book Depth
Book depth refers to the number of unique bids and asks in a system at a given point in time and will serve
as a proxy for liquidity in the system. We calculate it as a time-weighted average of the number of unique active
bids and asks. Mathematically, we can observe a sequence of transactions (bids, asks, and contracts) within each
market period. Let = (1 , 2 , , ) represent this sequence where 1 represents the first bid or ask within a
market period. Let = (1 , 2 , , ) represent a sequence corresponding to where is a count of the bids and
asks in the order book at . Here is a simple example: = (, , , , , ). Thus, =
(1,2,3,1,2,3). Finally, define = (1 , 2 , , ) as a sequence of timestamps within a market period, where 1 is the
time at which 1 was posted.
Thus, 2 1 represents the time elapsed between 2 and 1 . Book depth is calculated as follows within each
period:

12

=
1

(+1 )

III. Results & Discussion

III.I Trading Volume

In the baseline treatment, 22.44 shares were traded per market period on average, in line with the
equilibrium prediction of [20, 25] shares traded. In both MT and TF, our model predicted an equilibrium quantity
of 20 shares traded but we observed significant overtrading in both cases. In MT, 22.07 shares were traded per
market period on average compared to 22.25 shares per market period in TF. In both MT and TF, trading volume
declined relative to the baseline, but still fall well above our equilibrium prediction. Table 4a reports summary
statistics for each treatment.

()
()

27
36
16

22.44
22.07
22.25

. .
0.974
1.299
1.366

20.5
19.5
20

24
26
25

Table 4a: summary statistics for trading volume (all treatments)

A one-way ANOVA test (0 : = = : 0 false) confirms that our result, although present, is
not statistically significant ( = 0.74 = 0.4807). Clearly, there was a pronounced degree of overtrading in MT
and TF of approximately two trades per market period. This result is curious, as only twenty trades per market
period in these two treatments were profitable for subjects. That is, there were more than two trades per market
period in which at least one subject was made worse off by engaging in these trades. Next, we consider whether
the allocation of net fee between market maker and market taker has an effect on trading volume (0 : 2 =
3 : 2 > 3 ). Recall that we have three distinct make and take fee treatments. For purposes of
allocation, we compare the change in MT2 vs. MT1 to the change in MT3 vs. MT1 (see Table 4b). We can simply
difference out MT1 and directly compare MT2 and MT3.

13

. .

12
12
12

22
22.29
21.92

0.603
1.514
1.621

21
20
19.5

23
26
25

Table 4b: summary statistics for trading volume (MT treatments)

The difference was not statistically significant ( = 0.5855 & = 0.2821) and we actually saw an increase
in trading volume in MT2 relative to MT1. However, there was a decrease in trading volume in MT3 relative to both
MT1 and MT2, which matches the empirical conclusion from Kalcheva et al. (2013) that is, an increase in the take
fee decreases trading volume more than a decrease in the make rebate.

III.II Efficiency
In the baseline treatment, subjects achieved 98.52% efficiency. In MT, subjects achieved 94.24% efficiency
compared to 94.26% efficiency in TF. There is an efficiency loss in MT and TF relative to the baseline since trading
fees were exogenously imposed on subjects transactions. Recall that in MT and TF, either one or two experimental
dollars was transferred from the subjects to our fictional exchange per contract formed. Table 5 reports summary
statistics for each treatment.

. .

()
()

27
36
16

98.52%
94.24%
94.26%

1.40%
2.43%
3.19%

95.0%
86.8%
86.1%

100.0%
97.9%
97.9%

Table 5: summary statistics for efficiency (all treatments)

A one-way ANOVA test (0 : = = : 0 false) confirms that this efficiency decrease is
statistically significant ( = 30.15 = 0.0000) but we are more interested in comparing MT and TF (0 : =
: ) (See Table 5b). Efficiency in these two treatments was almost identical (94.24 % and
94.26%). This confirms our hypothesis of no significance in efficiency between MT and TF ( = -0.0152 & =
0.5060).

14

III.III Book Depth

In the baseline treatment, the average time-weighted book depth was 11.68. That is, there were 11.68
unique bids and asks in the order queue on average. We observed a decrease in book depth in TF to 10.72 and an
increase in book depth in MT to 12.99. In MT, subjects continually improve their bids and asks by small amounts
when employing rebate capture strategies. In the post-experiment questionnaires, subjects indicated a propensity
to employ these strategies. For example, one subject wrote You should not accept very many offers to sell to avoid
trading fees, so its worth it to just up their offer by one and put it on the board in response to the question Was
your strategy different in periods with trading fees vs. periods without trading fees? Table 6a reports summary
statistics for each treatment.

. .

()

()

27
12
12
12
36
8
8
16

12.49
12.73
12.88
13.36
12.99
10.43
11.01
10.72

2.49
2.54
2.26
2.69
2.45
1.42
2.13
1.77

7.27
8.74
9.54
9.00
8.74
8.12
8.59
8.12

18.10
17.42
16.19
17.35
17.42
12.51
14.55
14.55

Table 6a: summary statistics for book depth (all treatments)

A one-way ANOVA test (0 : = = : 0 false) confirms that there is a significant difference in
book depth between treatments ( = 5.24 = 0.0074). A two-sample t-test (0 : = : > )
confirms the statistically significant difference between MT and TF ( = 3.76 = 0.0003), which verifies our
hypothesis.

()
()

. .

36
16

12.99
10.72

2.45
1.77

8.74
8.12

17.42
14.55

Table 6b: summary statistics for book depth (MT & TF)

15

III.IV Bid-Ask Spread - Quoted

As described earlier, we calculated bid-ask spread with respect to the equilibrium price (100) a
uniformly drawn constant within each market period. The quoted bid-ask spread is defined as the bid-ask spread

exclusive of fees. The average quoted bid-ask spread was almost identical in MT and TF 0.165 and 0.164,
respectively. The quoted bid-ask spread increased in both cases relative to the baseline case of 0.157. Recall that
our hypothesis was non-directional in nature and our model was designed such that any treatment effects were
endogenously determined that is, the bid-ask spread should have been equivalent theoretically and our results
confirm that intuition. Table 7a reports summary statistics for each treatment.

. .

()
()

27
36
16

. 139
. 165
. 164

. 082
. 057
. 091

. 053
. 035
. 066

. 377
. 280
. 404

Table 7a: summary statistics for quoted bid-ask spread (all treatments)
A one-way ANOVA test (0 : = = : 0 false) confirms that there is no statistically significant
effect on the quoted-bid ask spread between treatments. ( = 1.07 = 0.349). This confirms the empirical
conclusion of Malinova and Park (2013) recall that they examined a proprietary data set from a 2005 trial period
on the Toronto Stock Exchange (TSX) and found that under a make and take fee regime, the quoted spreads did not
change and Kalcheva et al. (2014) who analyzed a larger sample of all registered equity exchanges from 2008
2010 and were unable to document a significant relationship between bid-ask spread and make and take fees.
Next, we consider whether the allocation of net fee between maker and taker has an effect (See Table 7b).

12
12
12

. 169
. 168
. 159

. .
. 043
. 063
. 066

. 109
. 070
. 035

. 246
. 280
. 269

Table 7a: summary statistics for quoted bid-ask spread (MT treatments)
We again fail to reject the null hypothesis (0 : 2 = 3 = 1 : 2 = 3 > 1 ) that there is
a statistically significant difference between MT2 and MT3 ( = 0.3511 = 0.7236). The empirical conclusion
from Colliard and Foucualt (2012) that is, the quoted bid-ask spread can increase or decrease when the net fee is
16

reduced, depending on whether the reduction in net fee is achieved by raising the make rebate this causes a
decrease or by lowering the take fee this causes an increase is partially true here. Although we increased the
level of net fee in our experiments in MT2 and MT3 relative to MT1, we would expect to see an increase in the
quoted bid-ask spread upon lowering the make rebate and a decrease in the quoted bid-ask spread upon raising
the take fee. We observed a slight decrease in the quoted bid-ask spread in MT3 (net fee = 2) relative to MT1 (net
fee = 1), but there was no change in the quoted bid-ask spread in MT2 relative to MT1.

III.V Bid-Ask Spread - Effective

The effective bid-ask spread is defined as the bid-ask plus two times the take fee. This represents the round
trip cost of trading on a liquidity-demanding transaction. In this case, we are limited to examining the changes
between different levels of net fee. Table 8 reports summary statistics for the MT treatments.

. .

()

12
12
12
36

. 268
. 276
. 282
. 275

. 053
. 064
. 083
. 066

. 189
. 175
. 105
. 105

. 366
. 383
. 418
. 418

Table 8: summary statistics for effective bid-ask spread (MT treatments)

In both MT2 and MT3, we observed an increase in effective bid-ask spread relative to MT1. However, a one
way ANOVA test (0 : = = : 0 false) confirms that the increase was not statistically significant.
This result does not confirm the empirical result from Malinova and Park (2014) in which the effective spreads
declined with an increase in the level of net fee.

III.VI Level of Net Fee

In this section, we consider the level of net fee in the MT periods in relation to the equilibrium price. Make
and take fees are assessed based on volume without regard to share price. That is, investors pay identical fees
whether they are buying 100 shares of a high price stock or a penny stock. Here, we take the equilibrium price
(shown in Table 9) and divide by the level of net fee (either 1 or 2) to give us a measure of the effects of net fees

17

relative to the predicted price. We examine the relationship between the magnitude of the fees relative to the
predicted price and our dependent variables.
Period

P*

5
100

6
60

11
50

12
50

17
120

18
110

Table 9: equilibrium prices in each period

For example, the equilibrium price in Period 12 is 50. Suppose the level of net fee in this period is equal to 2
(as in certain market periods in the MT2 and MT3 treatments). Then, the fee magnitude is equal to 1/25 of the
equilibrium price. If the equilibrium price was greater than 50, the magnitude would be lower than 1/25, and we
may expect less of an effect of imposing fees on our dependent variables. The charts in Appendix D show the
results of a linear regression analysis. We observe that the only statistically significant relationship between fee
magnitude and our dependent variables is the positive association between the magnitude and efficiency. That is,
efficiency is higher on average during periods in which the level of net fee is low relative to the equilibrium price
and vice versa. There does not appear to be a statistically significant relationship between fee magnitude and
trading volume, book depth, or bid-ask spread.

0.01
13.88
0.04
1.08
0.23

0.9222
0.0004
0.8512
0.3018
0.6304

0.0141
0.1535
0.0138
0.0012
0.0109

**significant at = 0.01

IV. Conclusion
This paper investigated the effects of imposing an exogenous make and take fee structure on a
nonstationary double auction model in an experimental setting. We compared the effects of make and take fees
relative to a model in which trading fees were assessed to both sides of a transaction equally and a baseline case in
which fees were absent. We examined trading volume and proxy measures of market quality and liquidity
efficiency, bid-ask spread, and book depth. There have been very few empirical studies of make and take fees given
18

their recent widespread implementation. The results from the empirical literature were mixed for example, two
of the papers demonstrated opposite effects of make and take fees on the bid-ask spread. In addition, the strength
of our paper is the controlled and randomized setting and elimination of some of the noise of the empirical data.
We conclude that make and take fees appear to have few tangible statistically significant effects on market
quality. In particular, we cannot document a statistically significant relationship between make and take fees and
efficiency or bid-ask spread (quoted and effective) nor does the presence of make and take fees have a significant
effect on trading volume. We do observe a statistically significant increase in book depth in the presence of make
and take fees relative to trading fees. If nothing else, this illustrates that subjects (and possibly brokers tasked with
potentially dichotomous tasks revenue maximization and order execution guidelines) pay attention to the
presence of rebates and fees. However, in future research, we plan to extend the model developed in this paper and
create an environment in which subjects assume one of three roles buyers, sellers, and fee setters. This research
will feature multiple markets operating simultaneously and have fee setters who will act as an equity exchange.
This will enable us to directly observe the effects of changes in the level of fees in a controlled setting and produce
an equilibrium prediction for the optimal level of net fee from the exchanges perspective. Further, we can
introduce brokers into the model to highlight the agency problem described in the empirical literature.

19

Bibliography
Angel, James J., Lawrence E. Harris, and Chester S. Spatt. "Equity trading in the 21st century." The Quarterly Journal
of Finance 1.01 (2011): 1-53.
Battalio, Robert, Shane A. Corwin, and Robert Jennings. "Can Brokers Have it all? On the Relation between Make
Take Fees & Limit Order Execution Quality." On the Relation between Make Take Fees & Limit Order Execution
Quality (December 13, 2013) (2013).
Cardella, Laura, Jia Hao, and Ivalina Kalcheva. "Make and Take Fees in the US Equity Market." April 29 (2013):
2013.
Colliard, Jean-Edouard, and Thierry Foucault. "Trading fees and efficiency in limit order markets." Review of
Financial Studies 25.11 (2012): 3389-3421.
Foucault, Thierry. Pricing Liquidity in Electronic Markets. Foresight Driver Review
Foucault, Thierry, Ohad Kadan, and Eugene Kandel. "Liquidity cycles and make/take fees in electronic markets."
The Journal of Finance 68.1 (2013): 299-341.
Gjerstad, Steven, and John Dickhaut. "Price formation in double auctions." Games and economic behavior 22.1
(1998): 1-29.
Malinova, Katya, and Andreas Park. "Subsidizing liquidity: The impact of make/take fees on market quality."
November 23 (2011): 2011.
McCabe, Kevin A., Stephen J. Rassenti, and Vernon L. Smith. "Designing call auction institutions: is double Dutch the
best?." The Economic Journal (1992): 9-23.
NYSE Arca Equities, Inc. Schedule of Fees and Charges for Exchange Services. Last updated September 2014
<https://www.nyse.com/publicdocs/nyse/markets/nyse-arca/NYSE_Arca_Marketplace_Fees.pdf>
O'Hara, Maureen. "High frequency market microstructure." Journal of Financial Economics 116 (2015): 257-270.

20

Appendix B Subject Instructions (buyer)

Note: While completing the instructions, subjects were asked to perform tasks to demonstrate that they
understood the experimental procedures (i.e. enter a bid). In the post-experiment questionnaire given to all
subjects, the vast majority of subjects indicated that the instructions were clear. Seller instructions are very similar
and are available from the author upon request.
____________________________________________________________________________________________________________________________________
Introduction
If you have any questions during the instructions or difficulties that arise after the experiment has begun, raise
your hand and someone will assist you.
The experiment will consist of several 3-minute trading periods. Some of the participants in this room are buyers
and some are sellers. For the duration of the experiment, you will be assigned to the role of buyer. Your role will
not change between periods.
Buyer Basics
You will be given values (in Experimental Dollars) for one or more shares in each period in the holdings table
displayed on the bottom right corner of the screen. You will attempt to purchase shares by placing bids to buy or
accepting sellers offers to sell.
To place a bid, enter the price at which you are willing to buy a share and click the Submit Bid to Buy button. Your
bid will be listed in the bid queue and will be marked with an asterisk.
Try this now. Type 100 into the box next to the Submit Bid to Buy button and click the button. Your bid is now
listed in the market along with another bid from another buyer for 110.
If a seller accepts your bid, a contract is formed.
You may retract your bid at any time by selecting your bid in the bid queue and clicking the Retract Bid button, as
long as yours is not the standing bid (the highest price currently posted by the buyers).
Try this now. Select your bid of 100 by clicking the bid in the bid queue and press the Retract Bid button.
There are two ways to accept the standing offer (the lowest price currently posted by the sellers):
1. Click the Buy@ button.
2. Place a bid that is higher than the standing offer.
In both cases, a contract will be immediately formed at the standing offer price. Contracts are formed in order from
highest value to lowest value.
Try this now. Click the Buy@ button to observe the changes to your holdings table. You bought a share that you
valued at 170 at a price of 120 for a profit of 50.

21

You cannot place more bids in the bid queue than you have remaining share values for. As you can see, you started
with four share values in your holdings table. You bought one share, so you can now place up to three bids.
You may place bids at any time. Bids must be made in whole numbers (no decimals or fractions).
Trading Mechanics
In some trading periods, you will either be assessed a fee or receive a rebate upon contract formation.
You will be assessed a fee if you accept the standing offer by clicking the Buy@ button or by placing a bid that is
higher than the standing offer. In either case, a contract will be immediately formed at the standing offer price.
Try this now. Click the Buy@ button to observe the changes to your holdings table. You bought a share that you
valued at 170 at a price of 120, but you were assessed a fee of 4 for a profit of 46.
You will receive a rebate if you enter a bid that is lower than the standing offer and wait for a seller to accept your
bid. If no seller accepts your bid, you will not receive a rebate.
Try this now. Type 100 into the box next to the Submit Bid to Buy button and click the button. Your bid is now
listed in the market along with another bid from another buyer for 90.
On the next page, you will observe what happens when a seller accepts your bid.
Your holdings table has been updated as a seller accepted your bid of 100. You bought a share that you had valued
at 150 at a price of 100, but you received a rebate of 2 for a profit of 52.
You received a rebate since you entered a bid that was lower than the standing offer and waited for a seller to
accept it.
Calculating Your Earnings
You will earn $1 U.S. Dollar for every 100 Experimental Dollars that you earn during the experiment. Your earnings
in each period are independent and do not carry over into the next period. Your earnings for the experiment will be
the sum of your earnings from each market period.
If you do not execute a trade in the period, you are guaranteed to earn zero profit. However, it is possible to earn
negative profit if you do not trade wisely. Specifically, you are not required to buy all of the shares that you have
values for in your holdings table.
Your profit for each contract is given by the following formula:
(Value Contract Price) Fee + Rebate = Profit

Example One: You have a value of 50 for one share. The standing offer in the market is 40. You accept the standing
offer. You would be assessed a fee, which we will assume to be 5 for this example. Your profit for this contract is:
(50 40) 5 = 5
(Value Contract Price) + Rebate = Profit

Example Two: You have a value of 100 for one share. The standing offer in the market is 110. You place a bid in the
22

market for 97. A seller later accepts your bid. You will receive a rebate, which we will assume to be 5 in this
example. Your profit for this contract is:
(100 97) + 5 = 8
(Value Contract Price) + Rebate = Profit
As you have observed, the profit calculations are done for you in the holdings table once you form a contract.
At the top of your screen, the following information is provided in the Relevant Information box;

Buyer ID #
Market Fee & Rebate

Your earnings for the experiment are displayed on the top right corner of the screen. This information is updated
at the end of each trading period.
You have completed the instructions. The experiment will begin once all participants have completed the
instructions. You may review the instructions now by using the navigational buttons at the bottom of the screen.
Printed copies of the instructions should be available at your computer terminal. Once you are ready, click the Start
button to proceed.
As a reminder, if you have any questions or difficulties that arise after the experiment has begun, raise your hand
and someone will assist you.
___________________________________________________________________________________________

23

Appendix C Subject Interface (buyer)

24

Appendix D Graphs

Trading Volume
30

Trading Volume

25
20

y = -0.0009x + 22.107
R = 0.0001

15
10
5
0
0

10

20

30

40

50

60

70

80

90

100

110

100

110

Eq. Price / Net Fee

Efficiency
100.0%

y = 0.0005x + 0.9178
R = 0.1654

Efficiency

95.0%

90.0%

85.0%

80.0%

75.0%
0

10

20

30

40

50

60

70

80

90

Eq. Price / Net Fee

25

Book Depth
20
18
16

y = 0.0029x + 12.866
R = 0.0005

Book Depth

14
12
10
8
6
4
2
0
0

10

20

30

40

50

60

70

80

90

100

110

100

110

Eq. Price / Net Fee

Quoted Bid-Ask Spread

25

Quoted B/A Spread

20

y = 0.0245x + 9.0529
R = 0.0152

15

10

0
0

10

20

30

40

50

60

70

80

90

Eq. Price / Net Fee

26

Effective Bid-Ask Spread

30
25
y = 0.0117x + 16.291
R = 0.0033

Effective
B/A Spread

20
15
10
5
0
0

10

20

30

40

50

60

70

80

90

100

110

Eq. Price / Net Fee

27