Can Brokers Have it all?

On the Relation between Make Take Fees & Limit Order Execution Quality.
by Robert Battalio, Shane Corwin, and Robert Jennings

4 December 2013

Average depth when all are at the inside quote: 1st five days of October 2012
14,000

12,000
10,000 Shares 8,000

6,000
4,000 2,000

0
NDAQ Take Fee: $0.30 Make Rebate: $0.29 ARCA $0.30 $0.30 EDGX $0.30 $0.32 BZX $0.29 $0.29 NYSE $0.23 $0.21 BYX - $0.02 -$0.03 EDGA - $0.04 -$0.06 BSX - $0.14 - $0.18

Where are queues the shortest? Where is it cheapest to access displayed liquidity?

Where will an executed limit order generate the highest liquidity rebate?

When were fees explicitly introduced to markets?

1997 Order Handling Rules.
The quotes displayed on computerized limit order books (think ECNs) were incorporated into the public quotation system.

To attract liquidity onto their limit order books, ECNs and Nasdaq began offering liquidity rebates to reward market participants for placing nonmarketable orders on their limit order book. These rebates were funded by charging incoming marketable orders a fee for taking liquidity.
This is the traditional make/take model.

The regulation of take fees.

Rule 610 of Regulation NMS (circa 2005) limits take fees to $0.003/share. Why limit take fees?
Regulation NMS prohibits trading through the quotes of other markets. Incentive to post bid price of $0.00 with a take fee of $1,000,000 per share. Rule 610 is intended to preserve the integrity of the quote.

Why not simply post net prices on the real line???

4

Who cares?
The Investment Company Institutes members, who managed over $11 trillion of assets and held 28% of the value of publicly traded U.S. equity outstanding at the end of 2009.
we are concerned that brokers may refrain from posting limit orders on a particular exchange because it offers lower liquidity rebates than other markets, even though that exchange offers the best possibility of an execution for those limit orders. Practices such as these, in turn, may ultimately harm investors because their limit orders may not be executed. At the same time, it is unclear what benefits liquidity rebates provide to investors. (emphasis added)

What is the Investment Company Institute saying?

1. Brokers may be more concerned about harvesting rebates than they are about obtaining best execution for customer orders.
The classical principal/agent problem.

2. This type of order routing results in inferior limit order execution quality. 3. Competition and transparency cannot fix this problem.

1. Would brokers really do this?

Chris Nagy, Wall Streets most sophisticated order sender Traders Magazine. The group Nagy heads endowed Ameritrade with the ability to separate marketable and non-marketable order flow.
It sends its market orders to wholesalers and receives payments. it began sending its limit orders to exchanges to take advantage of their rebates from maker-taker pricing. In splitting its order flow, Ameritrade could improve its revenues by taking advantage of maker-taker pricing at the exchanges and still get paid for sending order flow to the wholesalers.

2. Are standing limit orders harmed (1 of 4)?

Suppose there are two order books at the National Best Bid (NBB) for Pudgys Chicken common stock.

EDGX
NBB $10.24

EDGA $10.24 - $0.0004

Make Rebate: + $0.0023 Take Fee: + $0.0029 Does the order routing decision matter?

- $0.0005

Are standing limit orders harmed (2 of 4)?

Scenario 1:
Negative earnings news is announced. Market sell orders push the market price for Pudgys Chicken down to $9.99 bid and $10.01 offered.
Both limit buy orders execute.
Both limit buy orders have negative realized spreads (five minute alphas).

In this situation, execution happens regardless of where the order is sent you might as well receive a high rebate.

2. Are standing limit orders harmed (3 of 4)?

Scenario 2:
A fee-sensitive investor routes a marketable sell order to EDGA and trades with the standing limit order. Subsequently, buying pressure pushes the market for Pudgys Chicken to $10.49 bid and $10.51 offered.
Only one limit buy order executes. The executed limit order generates a positive realized spread. In this situation, the limit order resting on the venue with the higher liquidity rebate (and higher take fee) misses out on a profitable trading opportunity.

10

2. Are standing limit orders harmed (4 of 4)?

Scenario 3:
A marketable sell order is routed to EDGX instead of EDGA because of faster connections and/or the expectation of greater nondisplayed liquidity (e.g., price/depth improvement). Subsequently, buying pressure pushes the market for Pudgys Chicken to $10.49 bid and $10.51 offered.
Only one limit order executes.

The executed limit order generates a positive realized spread.

In this situation, the limit order resting on the venue with the higher liquidity rebate (and higher take fee) received a favorable execution and received the liquidity rebate!

11

3. Why cant Adam Smith fix this?

Brokers only get commissions when limit orders execute. Why wouldnt brokers seek high fill rates rather than high rebates? Big agency problem - Angel, Harris, and Spatt (2010).
Hard for investors to evaluate limit order execution quality. Easy for investors to shop based on commissions. Brokers that seek to maximize order rebates can offer the lowest commissions. Investors cannot see that paying a higher commission might actually make them better off.

12

Our contribution

13

What do BCJ do? (1 of 3)

We use publicly available Rule 606 data to examine whether any brokers appear to be routing orders in a way that maximizes liquidity rebates. Ameritrade, E*Trade, Fidelity and Scott Trade are suspects. We next examine whether this type of order routing is consistent with a brokers fiduciary responsibility to obtain best execution for customer limit orders. More specifically, we examine the relation between take fees (liquidity rebates) and limit order execution quality.
14

What do BCJ do? (2 of 3)

Limit order execution quality metrics for orders
Fill rates & queue lengths On which venue are limit orders executed most frequently? Horseraces For identical orders routed to multiple venues, we examine where orders are executed first.

Conditional on an execution
Execution speed - Where are limit orders executed more quickly?
Realized spreads & good fill ratios Which venue offers limit order executions the largest five minute alpha? On which venue are adverse selection costs the greatest?
15

What do BCJ do? (3 of 3)

We use proprietary limit order data and the NYSEs TAQ data from October and November 2012 to address these issues. Overall, we present strong evidence that venues with high take fees (liquidity rebates) offer inferior limit order execution quality. This suggests brokers cannot have it all.
Brokers must either choose to maximize liquidity rebates or to maximize limit order execution quality.

16

Literature review

17

Literature Review (as it pertains to our paper)

Academic (publicly available data)
Foucault and Menkveld (2008) & Cardella, Cao and Kalcheva (2013)
Fee schedules are a primary determinant of where marketable orders are routed.

Practitioner (proprietary data)

Sofianos and Yousefi (2010), Sofianos, Xiang, and Yousefi (2010), & Bacidore, Otero and Vasa (2011)
Evidence suggests routing decisions (influenced by fee schedules) affects execution quality.
18

Broker order routing in 4Q12.

19

Order Routing (1 of 6)
SEC rule 606 requires brokers to reveal on a quarterly basis the destinations to which they route non-directed orders and whether they receive compensation for their routing choices.
Any venue receiving less than 5% of the brokers orders does not have to be revealed on the brokers 606 report. Thus, percentages need not sum to 100.

We collect 4Q2012 Rule 606 reports for ten national brokerages appearing in either Barrons or Smart Moneys 2012 Broker Surveys from broker websites.
We present results for routing in NYSE-listed securities.
20

Order Routing (2 of 6)
Five brokers route 100% of their order flow to purchasers of order flow: Schwab, Morgan Stanley, Just2Trade, Edward Jones, and LowTrade.
Limit orders routed to any of these brokers have an opportunity to interact with both the brokers and the purchasers marketable order flow. Manning rules are intact. Purchasers pay brokers less for order flow they cannot interact with. Thus, this type of order routing does not, on the surface, appear to be consistent with the objective of maximizing order flow payments.
21

Order Routing (3 of 6)
Venue Make/Take Order Mix Ameritrade E*Trade Fidelity ScottTrade

EDGX $0.32/-$0.30

% Mkt % Lmt

0% 49%

0% 46%

0% 28%

0% 28%

One of three venues charging the highest take fee.

Lava $0.27/-$0.28

% Mkt % Lmt

0% 0%

0% 0%

0% 0%

0% 51%

We think these numbers are conservative.

Purchasers $0.00/<$0.00

% Mkt % Lmt

96% 45%

98% 51%

97% 57%

66% 21%

Mostly marketable limit orders?

22

Order Routing (4 of 6)
Venue Take Fee Order Mix

Uses a smart order router and passes fees/rebates onto customers

Interactive Brokers

Nasdaq $0.30 NYSEs Arca Exchange $0.30

% Lmt

7%

% Lmt

23%
Why no inverted routing?

Bats Z Exchange $0.29

% Lmt

14%

NYSE (vintage) $0.23

% Lmt

47%
23

Traditional venue with lowest take fee

Order Routing (5 of 6)
Fidelitys non-directed limit order routing in 3Q2013
Retail Limit Orders Direct Edge Nasdaq BATS NYSE Citadel (wholesaler) Goldman 30.45% 0.00% 0.00% 0.00% 9.72% 3.37% Institutional Limit Orders 9.91% 11.28% 15.02% 22.49% 0.08% 1.71%

NFS (wholesaler)
Knight (wholesaler) UBS (wholesaler)

19.19%
21.36% 5.72

27.04%
1.12% 0.75%

Two Sigma (wholesaler)

2.79%

0.00%
24

Order Routing (6 of 6)
Revisiting an earlier slide
Ameritrade began sending its limit orders to exchanges to take advantage of their rebates from maker-taker pricing.
In splitting its order flow, Ameritrade could improve its revenues by taking advantage of maker-taker pricing at the exchanges and still get paid for sending order flow to the wholesalers.

Rule 606 data suggest the limit order routing decisions of Ameritrade, E*Trade, Fidelity, and Smart Trade are heavily influenced by make rebates.

Does this adversely impact limit order execution quality?

25

Analysis of proprietary limit order data.

26

Data
We obtain limit order data from a major broker-dealers smart order routing system for October and November 2012.
The orders in the sample are from customers, primarily institutional investors. Orders included are those routed through the broker-dealers algorithmic trading system and orders entered directly by customers. 25th percentile and median order size is 100 shares. 75th percentile order size is 300 shares. The median (75th percentile) order displays 0 (100) shares.
27

Where are orders sent?

Venue Take Fee Min/Max $0.29/$0.30 $0.30/$0.30 $0.30/$0.30 $0.29/$0.29 Displayed Order Make Rebate +$0.23/+$0.32 +$0.20/+$0.29 +$0.21/+$0.30 -$0.25/+$0.29 Hidden Order Make Rebate +$0.23 +$0.10 +$0.21 -$0.10 % of Displayed Orders
(N=10,392,317)

% of Hidden Orders
(N=17,829,197)

EDGX NDAQ ARCA BZX

29.72% 23.84% 12.61% 5.26%

31.97% 39.76% 11.36% 2.15%

NYSE
EDGA BSX

$0.23/$0.23
-$0.04/-$0.04 -$0.14/-$0.14

+$0.15/+$0.21
-$0.06/-$0.05 -$0.18/-$0.15

+$0.15
-$0.06 -$0.18

26.72%
0.20% 1.65%

14.46%
0.00% 0.30%

Unfortunately, not many orders are routed to the inverted venues.

28

Order outcomes

Fill Status No Fill Complete Partial Fill

# of Orders 17,455,297 10,231,469 534,748

Rejected 1,681 0 1

Replaced 117,895 8 107

Cancelled 15,853,841 53,083 524,757

One Fill n.a. 7,964,586 349,482

Expired 1,481,880 n.a. 9,883

29

Order aggressiveness
Limit Order Type Behind-the-quote At-the-Quote Inside-the-Quote Marketable Buy Orders 14.90% 30.74% 5.48% 0.65% Sell Orders 12.32% 29.76% 5.24% 1.04%

Display Choice
Limit Order Type Hidden Behind-the-quote At-the-Quote Inside-the-Quote Marketable 32.21% 58.84% 7.19% 1.88% Partially or Fully Displayed 18.56% 63.34% 16.77% 1.38%
30

Unconditional statistics for displayed at-the-quote limit orders

Venue EDGX NDAQ Take Fee/Rebate $0.30 $0.30 # Orders 2,525,513 1,500,240 Fill Rate 53.96% 47.96% Fill Speed (seconds) 111 59 Realized Spread -$0.0021 -$0.0020 Good Fill Ratio 49.32% 49.19%

ARCA
BZX NYSE EDGA BSX

$0.30
$0.29 $0.23 -$0.04 -$0.14

791,087
486,375 1,088,393 20,529 170,736

53.11%
54.66% 56.75% 56.85% 74.51%

61
76 40 105 33

-$0.0019
-$0.0028 -$0.0011 $0.0011 $0.0011

49.44%
50.42% 50.39% 57.36% 54.81%

31

Probit analysis of limit order fill rates (1 of 2)

Fill = 0 + 1 Sell + 2 Short Sell + 3 Display + 4 Moneyness + 5 Trading Intensity + 6 Price + 7 Volatility + 8 Seconds from Mid-day + 9 Mean Response Time + 10 Order Size + 11 Take Fee + 12 (Display)(Take Fee) +

Caveats:

Control variables are created within sample from actual orders. We can/will do a much better job of this in next version of paper. We dont think things will change.
Hypothesis: The probability that a displayed limit order fills is decreasing in an exchanges take fee.
(11 + 12) < 0
32

Probit analysis of limit order fill rates (2 of 2)

Results:
A limit order is more likely to execute if it is displayed, small, aggressively priced, and if it is placed on a venue with faster connections to our broker. in a less volatile, less active stocks. Endogeneity???

displayed on a venue with a lower take fee.

33

OLS analysis of limit order time-to-execution (1 of 2)

Speed = 0 + 1 Sell + 2 Short Sell + 3 Display + 4 Moneyness + 5 Trading Intensity + 6 Price + 7 Volatility + 8 Seconds from Mid-day + 9 Mean Response Time + 10 Order Size + 11 Take Fee + 12 (Display)(Take Fee) +

Caveats: Control variables are created within sample from actual orders. We can/will do a much better job of this in next version of paper. We dont think things will change. Hypothesis: The time-to-execution for displayed venues is increasing in the size of the take fee.

(11 + 12) < 0

34

OLS analysis of limit order time-to-execution (2 of 2)

Results:
Time-to-execution is shorter for limit orders that are displayed, small, aggressively priced, and if it is placed on a venue with faster connections to our broker. in more volatile, more active stocks. routed to venues with lower take fees.

35

Horseraces (1 of 2)
We identify order-pairs that have the same stock symbol, order date, order side (buy or sell), limit price, order time (to within one millisecond), but different destination venues.
By construction, these paired orders control for stock characteristics and market conditions.

For each pair of identical orders routed to different venues, we conduct a horserace to determine which venue (if either) seems to perform better.

36

Horseraces (2 of 2)
The primary definition of winner is the venue that fills the order first while the competing venue still holds the paired order.
If both orders receive an execution, the venue that fills the order first wins. If one order (partially) fills, and the second subsequently is cancelled or replaced, the venue filling the order wins. Horseraces in which both orders in an order pair are filled within 500 microseconds (1/2 a millisecond) of one another are considered ties.

37

EDGX vs. the NYSE (1 of 3)

All Order Pairs 1st Venue 2nd Venue (take fee) (take fee) EDGX ($0.30) NYSE ($0.23) NYSE ($0.23) EDGX ($0.30) # of Order Pairs % Both Fill Difference in Time to 1st Fill

258

14.34

1.01

461

9.33

57.91

38

EDGX vs. the NYSE (2 of 3)

Order Pair 1st Venue 2nd Venue (take fee) (take fee) EDGX ($0.30) NYSE ($0.23) NYSE ($0.23) EDGX ($0.30) # of Order Pairs % 1st Venue Trades First % 2nd Venue % Tie Trades First

258

10.08

82.56

7.36

461

91.32

3.47

5.21

Takeaway: Roughly 5% of the horseraces end in a tie (it didnt matter where the order was routed). When it did matter, the NYSE offered the first fill much more frequently.
39

EDGX vs. the NYSE (3 of 3)

Order Pair 1st Venue (take fee) 2nd Venue (take fee) Good Fill Ratio 1st Venue 2nd Venue

EDGX ($0.30) NYSE ($0.23)

NYSE ($0.23) EDGX ($0.30)

51.79

58.78

65.72

55.36

Takeaway: A higher percentage of executions produce positive realized spreads on the NYSE. However, the median realized spread is positive on both the NYSE and on EDGX.
40

Horseraces involving all orders that do not end in ties.

100% 90% 80% 70% 60% 50%

99.74%

% of horseraces won by low fee venue

68.35% 68.78%

93.79%

64.02%
60.01%

54.45%

40% 30%
20% 10% 0%
BZX vs ARCA BZX vs NDAQ BZX vs EDGX NYSE vs BZX NYSE vs ARCA NYSE vs NDAQ NYSE vs EDGX BSX vs NDAQ Fee Diff = $0.01 Fee Diff = $0.01 Fee Diff = $0.01 Fee Diff = $0.06 Fee Diff = $0.07 Fee Diff = $0.07 Fee Diff = $0.07 Fee Diff = $0.44 N = 1,433 N = 9,811 N = 24,309 N = 2,637 N = 550 N = 51,240 N = 676 N = 3,128

23.59%

41

Horseraces involving all displayed orders that do not end in ties.

100%
90% 80% 70% 60%

85.30%

81.04% 71.95%

% of horseraces won by low fee venue

65.77% 51.27% 69.88%

50% 40% 30% 20% 10% 0%

BZX vs ARCA BZX vs NDAQ BZX vs EDGX NYSE vs BZX NYSE vs ARCA NYSE vs NDAQ Fee Diff = $0.01 Fee Diff = $0.01 Fee Diff = $0.01 Fee Diff = $0.06 Fee Diff = $0.07 Fee Diff = $0.07 N = 170 N = 327 N = 2,934 N = 310 N = 149 N = 1,604
42

What does it mean to win a set of horseraces?

43

Horseraces involving all orders that do not end in ties.

12%

11.03%
10%

9.15%

8%

Difference in good fill ratios. Difference is positive if low fee venue has a higher good fill ratio

6%

4%

2.92%
2.15%
2%

2.73%

2.72%

0.26%
0%

-0.39%
-2%
BZX vs ARCA BZX vs NDAQ BZX vs EDGX NYSE vs BZX NYSE vs ARCA NYSE vs NDAQ NYSE vs EDGX BSX vs NDAQ Fee Diff = $0.01 Fee Diff = $0.01 Fee Diff = $0.01 Fee Diff = $0.06 Fee Diff = $0.07 Fee Diff = $0.07 Fee Diff = $0.07 Fee Diff = $0.44 N = 1,433 N = 9,811 N = 24,309 N = 2,637 N = 550 N = 51,240 N = 676 N = 3,128

44

Summary of results obtained from proprietary data. (1 of 2)

Univariate statistics suggest a negative relation between take fees and fill rates and between take fees and realized spreads. Multivariate probit analysis reveals that for displayed orders, there is a negative relationship between take fees and limit order fill rates. OLS regression analysis confirms that limit orders resting on venues with high take fees require more time to fill than those on venues with lower take fees. Our examination of identical order pairs simultaneously routed to different venues under identical market conditions reveals that for take fee differences exceeding $0.0001 per share, the lower take fee venue has higher measured limit order execution quality.
45

Summary of results obtained from proprietary data. (2 of 2)

Our results suggest the decision to route the bulk of ones limit orders to a single venue offering the highest liquidity rebate is inconsistent with a brokers fiduciary responsibility to obtain best execution.

46

Shortcomings of our analysis of proprietary data.

The executed limit orders in our dataset are from a single broker and they comprise about 1.5% of average daily volume. Our data do not span the thirteen U.S. stock exchanges that utilize maketake or inverted make-take fee schedules. Specifically, our data provider uses inverted venues somewhat sparingly. Our proprietary data do not allow us to examine the types of stocks/situations in which the routing of limit orders to venues with relatively high take fees does not harm limit order execution quality.

47

Using the NYSE TAQ database to make inferences regarding across-venue limit order execution quality.

48

What can be done with TAQ? (1 of 3)

TAQ data can be used to determine whether fees influence where marketable orders (as evidenced by trades) execute when multiple venues have the best posted quote.
Based on the analysis in the prior section, we expect that venues with lower or negative take fees receive a larger share of marketable orders than venues with high take fees when all venues display the best quote. Unless the three venues charging take fees of $0.30 per hundred always have the shortest queues when they are at the best quote or the market always exhausts the depth at posted quotes, such a finding would suggest that brokers seeking to obtain best execution for customer limit orders should not route 100% of their limit orders to a single high take fee venue.

49

What can be done with TAQ? (2 of 3)

TAQ data can be used to compute the realized spreads of executions that take place at the quoted price when each of the relevant trading venues is posting the best quote.
Assuming these quotes are set by limit orders, we can examine the association between take fees (make rebates) and the execution quality of executed limit orders. Do limit orders resting on venues with higher take fees face higher adverse selection costs because they tend to be the last to trade at a price?

50

What can be done with TAQ? (3 of 3)

We can also examine the extent to which across-venue differences in limit order execution quality dissipate when fees/rebates are incorporated into realized spreads.
We might expect the all in cost of limit order trading (e.g., the implementation shortfall) to equilibrate across venues. Investors must chase the market to ultimately satisfy their trading interests when limit orders are unfilled. Implementation shortfall captures this cost. This does not imply that fee-adjusted realized spreads should equilibrate across venues.
51

How restrictive is our focus on trades executed at the quote when all venues are at the quote?
12.25% of all non-sweep trades in NYSE-listed securities occur at the quote when each of the relevant trading venues is at the inside quote.
The median NYSE stock has 0.24% of its non-sweep trades executed at the quote when all are at the quote. Ford is the stock with the highest percentage, 52.23%, of non-sweep trades executed at the quote when all are at the quote. Ford traded between $9.71 and $10.16 during the 1st week of October.

Over 99% of the trades executed at the quote when all venues are at the inside quote occur when the width of the NBBO is equal to $0.01.

52

Market share of trades in NYSE-listed securities.

(No sweep trades) Unconditional
18.00% 16.00% 14.00% 12.00% 10.00% 8.00% 6.00% 4.00% 2.00% 0.00% NDAQ $0.30 ARCA $0.30 EDGX $0.30

All at inside quote & trade at inside quote

Internalization?

The NYSE is always at the inside quote others are not.

BZX $0.29

NYSE $0.23

BYX - $0.02

EDGA - $0.04

BSX - $0.14

53

Average depth when all are at the inside quote

14,000
12,000 10,000 8,000 6,000

4,000
2,000 0 NDAQ $0.30 ARCA $0.30 EDGX $0.30 BZX $0.29 NYSE $0.23 BYX - $0.02 EDGA - $0.04 BSX - $0.14 54

Interpretation
On average, inverted venues have shorter queues and are at least as likely to receive marketable orders as the traditional venues. Hard to rationalize routing all/most of your limit orders to a single traditional venue offering the highest liquidity rebate (charging the highest take fee) if fees/rebates are not passed through to investors.

55

Unconditional realized spreads in NYSE-listed securities. (No sweep trades)

5 4 3 2
Avg. realized spread (bps.)

1
0 -1 -2 -3 EDGX $0.30 ARCA $0.30 NDAQ $0.30 BZX $0.29 NYSE $0.23 BYX - $0.02 EDGA - $0.04 BSX - $0.14
56

Realized spreads for trades at the quote when all are at the quote in NYSE-listed securities. (No sweep trades)
Avg. realized spread (bps.)
8 6 4

Avg. realized spread when all are at inside (bps.)

Order flow becomes more toxic!

2
0

-2
-4 EDGX $0.30 ARCA $0.30

Order flow becomes less toxic!

NDAQ $0.30
BZX $0.29 NYSE $0.23 BYX -$0.02 EDGA -$0.04 BSX -$0.14

57

Good fill ratios for trades at the quote when all are at the quote in NYSE-listed securities. (No sweep trades)
Good Fill Ratio
60%
50% 40%

30%
20% 10%

0%
EDGX $0.30 ARCA $0.30 NDAQ $0.30 BZX $0.29 NYSE $0.23 BYX - $0.02 EDGA - $0.04 BSX - $0.14
58

Interpretation
On average, at-the-quote limit orders receive more favorable executions on inverted venues (and the NYSE). At-the-quote limit orders resting on inverted venues face lower adverse selection costs.
Indeed, over half of the executed at-the-quote limit orders on the inverted venues and the NYSE have positive realized spreads!

59

Unconditional, conditional, and adjusted realized spreads for trades in NYSE-listed securities. (No sweep trades)
8
6 4 2 0 -2 -4 EDGX $0.30 ARCA $0.30 Avg. realized spread (bps.) Avg. realized spread when all are at inside (bps.) Adjusted realized spreads when all are at the inside (bps.)

NDAQ $0.30

BZX $0.29

NYSE $0.23

BYX - $0.02

EDGA - $0.04

BSX - $0.14

60

Unconditional, conditional, and adjusted realized spreads for trades in NYSE-listed securities. (No sweep trades)
Avg. realized spread (bps.) Avg. realized spread when all are at inside (bps.)

8
6 4 2 0 -2 -4 EDGX $0.30 ARCA $0.30

IBs limit order routing 22% 7% 14% 47%

NDAQ $0.30
BZX $0.29 NYSE $0.23 BYX - $0.02 EDGA - $0.04 BSX - $0.14

61

Unconditional, conditional, and adjusted realized spreads for trades in NYSE-listed securities. (No sweep trades)
Avg. realized spread (bps.) Avg. realized spread when all are at inside (bps.)

8
6 4 2 0 -2 -4 EDGX $0.30 ARCA $0.30

Fidelitys institutional limit order routing 10% 11% 15% 22%

NDAQ $0.30
BZX $0.29 NYSE $0.23 BYX - $0.02 EDGA - $0.04 BSX - $0.14

62

Interpretation
Conditional on an execution, investors who are responsible for the fees/rebates that their trades generate earn the highest adjusted realized spreads on the NYSE, a traditional venue. Apparently, the liquidity on the inverted venues does not absorb all of the good fills.

Relative to the NYSE, for executed limit orders, the benefit of gaining net price priority on the BSE is, on average, outweighed by the differential in make rebates (which is at least $0.003/share). Is why the smart routers used by brokers that pass on fees/rebates do not use the inverted venues that frequently?
Table VIII reveals the inverted venues spend much less time at the inside quote than traditional venues.
63

Realized spread regressions

Routing choice is more important when queue lengths are long.

Routing choice is more important for low priced stocks.

Because the minimum tick is binding Because queue lengths are longer.

Focusing on the model with fee dummies that examines trades executed when all venues are at the relevant quote. Its not!
Assuming the probability of execution is constant across venues, the average at-thequote limit order execution would have to earn a make fee of $1.16 per hundred shares for NYSE-listed securities to compensate for the lower realized spreads generated by atthe-quote trades on the venues charging a take fee of $0.30 per hundred shares to be as well off as executing on the BSX (the most aggressive inverted venue). Moreover, the make fee would have to be directly passed through to investors.

64

Conclusions (1 of 2)
Several large, national brokerage are making order routing decisions that appear to be consistent with the goal of maximizing order flow rebates. Each of these brokers routes limit orders either to a market maker or to a single traditional venue that pays the highest rebates.

Proprietary data suggests this type of order routing results in lower fill rates and increased adverse selection costs. We find similar results using TAQ data.
We have identified situations in which expected fill rates are higher on inverted venues

Adverse selection costs are highest on the traditional venues.

65

Conclusions (2 of 2)
Our evidence suggests brokers who do not pass fees/rebates onto customers cannot have it all.
Maximizing liquidity rebates on executed orders is inconsistent with obtaining best execution for non-marketable limit orders.

66