EUROPE

September 2009

Navigating Liquidity 3

Making sense of liquidity fragmentation

Recent significant changes in the liquidity landscape
Q CHI-X has become well established as the leading MTF and No. 3
execution venue in Europe. It is still extending its influence in less liquid
stocks.
Q TURQUOISE has succeeded in recovering market share after March
2009 and has kept its key strength: good coverage of indices. Its
performance has also been confirmed by the increase in its percentage of
time at the EBBO. There are concerns about the future ownership and
direction of Turquoise due to current discussions of a potential sale.
Q BATS’s inverted pricing had a major effect on its market share, which
doubled on the CAC 40 in the space of two weeks. However, taking into
account market share, stock coverage and percentage of time at the
EBBO, its best performance has been in UK stocks.
Q NASDAQ OMX has every chance of succeeding as a newcomer.
Indeed, an MTF that can improve its bid-ask spread and BBO presence
compared to others is immediately rewarded with an increase in market
share.
Q Overall, the MTFs’ unique intra-day volume profiles tend to
disappear. This can be attributed to more widespread use of SORs by
investors. Conversely, slight differences at the opening and the close are
likely to continue to exist as long as investors remain uninterested in using
MTFs for fixing auctions.

Key points when routing orders in a fragmented market

Q Focus on child orders: Before market fragmentation, large orders
were only temporally split, but now each child order has to be sent to
different trading destinations according to the liquidity they offer.
Q Use a Smart Order Router (SOR): It has become obvious that
SORs are now the key component when accessing fragmented liquidity.
The best way to use the liquidity available in alternate venues is to avoid
crossing a tick price on the main trading destination if volume is available
at the same price on another venue. Efficient SORs – such as ours at CA
Cheuvreux – now source more than 40% of their liquidity outside primary
markets.
Q Aggregate liquidity to minimise the impact of each order: An SOR
order is like a new order type giving access to an aggregated view of
liquidity. As long as the average size of aggressive orders is larger than
the size of passive orders, any decrease in the size of the former has no
impact on the ATS or on the number of deals.
Q Beware of duplicated liquidity: High frequency proprietary trading
tactics lead to duplicated quantities in order books, which makes the
SOR's task more complex. Adding pegged and iceberg orders, "pre-
trade transparency" is not a very useful pointer for guessing where liquidity
really is. Post-trade transparency gives far more information to investors.
This makes dark pools far more attractive than they seemed some
months ago.

Disclosures available on www.cheuvreux.com www.cheuvreux.com

September 2009 Navigating Liquidity 3

Introduction
Over the past two years, MiFID and the financial crisis have drastically modified the European trading landscape. The
expected fragmentation has occurred, as was the case in the U.S. two years after Reg NMS, but the financial environment
has not been favourable to the creation of numerous MTFs, as some are finding it difficult to compete with the primary
markets.

We can nevertheless observe that Chi-X has succeeded in challenging the traditional exchanges, as it is now the third-
largest trading destination in Europe based on numerous criteria.
At the same time 70% of deals on the U.S. market stem from the constantly-growing population of high-frequency traders
and market makers. These players are coming to Europe and could further modify this already-evolving landscape.

CA Cheuvreux is now monitoring market liquidity for clients in its "Monthly Market Indicators" publication. These indicators
allow you to follow and interpret local fragmentation trends. In contrast, our Navigating Liquidity studies are an invitation to
gain a deeper understanding of crucial mechanisms.

In this third issue, we will comment on market share trends since the publication of Navigating Liquidity 2, and look at mid
cap stocks when needed. This will be followed by a focus on the impact of tick size on fragmentation. We will then analyse
the main principles of smart routing of aggressive orders, and explain the splitting rates obtained by CA Cheuvreux's Smart
Order Router, which gives our clients aggregated and enhanced access to European liquidity.
The last section of this study presents a model explaining the decay in the market share of primary markets resulting from
investors' objective and rational optimisation of their trades.

About the authors

Charles-Albert Lehalle
Currently Head of Quantitative Research at CA Cheuvreux, Charles-Albert Lehalle also lectures at "Paris 6 (EL Karoui) Master of Finance"
(Ecole Polytechnique, ESSEC, Ecole Normale Supérieure) and gives master classes in the Certificate in Quantitative Finance in London.
He has also given lectures at numerous seminars and international conferences at MIT and the University of Edinburgh.
With a Ph.D. in applied mathematics, Charles-Albert is an expert in stochastic processes, information theory and nonlinear control. He has
published international papers on quantitative finance, real-time optimisation of high dimensional processes (with applications to Formula One,
high-mix fabs, large plants, and aerospace), and learning theory.

Romain Burgot
Currently working as a statistician researcher in the CA Cheuvreux Quantitative Research team, Romain previously worked on microstructure
questions for the final paper of his studies (ENSAE 2006), which was entitled "The comparison of VaR measures with high frequency data".

Execution Services Contacts

GENERAL HOTLINES:
Paris: +33 1 41 89 80 88 London: +44 207 621 52 00
New York: +1 212 492 8850

Ian Peacock Jerry Lees

Global Head of Execution Services Head of Alternative Execution Services
+44 207 621 5144 / ipeacock@cheuvreux.com +44 207 621 5281 / jlees@cheuvreux.com

SELL SIDE BUY SIDE

Jonathan Carp Mark Freeman
Head of Alternative Execution Sales Europe Head of Alternative Execution Sales to Buy Side
+44 207 621 5244 / jcarp@cheuvreux.com +44 207 621 5285 / mfreeman@cheuvreux.com

2 www.cheuvreux.com
September 2009 Navigating Liquidity 3

CONTENTS
I— What changes have occurred recently in this fragmented liquidity
landscape? 4
Q The MTF's specificities of intra-day volume profiles are tending
to disappear 4
Q Chi-X continues to do well and to attract more and more flows 6
Q Turquoise has done a very good job and is recovering its market share 8
Q After taking into account BATS' temporary aggressive rebates, its
performance is disappointing 10

II— How can tick size be "optimal"? 12

Q Tick size is clearly linked to an MTF's market share 12
Q The potential dangers and benefits of a tick size reduction 13
Q Each stock needs its own tick size 14

III— How to route orders in a fragmented market? 16

Q Focus on atomic orders 16
Q Using a Smart Order Router (SOR) 17
Q Aggregate liquidity to minimise the impact of each order 18
Q Beware of duplicated liquidity 19

IV— Two results and a guess 21

Q Efficiency of CA Cheuvreux's SOR 21
Q A split driven by Chi-X efficiency 22
Q Fragmentation is a consequence of primary markets' variance 23

Appendices 25
Q Appendix 1: Glossary 25
Q Appendix 2: Methodology 26
Q Appendix 3: Lists of stocks 27
Q Appendix 4: Information on backbone latencies 29
Q Appendix 5: Liquidity metrics 30

3 www.cheuvreux.com
September 2009 Navigating Liquidity 3

I— What changes have occurred recently in

this fragmented liquidity landscape?
Q The MTF's specificities of intra-day volume profiles are
tending to disappear

FIGURE 1: INTRA-DAY PROPORTION OF VOLUME FOR THE DIFFERENT FIGURE 2: INTRA-DAY PROPORTION OF VOLUME FOR THE
EXCHANGES (CAC 40, FEBRUARY 2009) DIFFERENT EXCHANGES (FTSE 100, FEBRUARY 2009)

0.09 0.09
EURONEXT LSE
0.08 CHI-X 0.08 CHI-X
TURQUOISE TURQUOISE
0.07 BATS BATS
0.07

0.06
0.06
0.05
0.05
0.04
0.04
0.03

0.03
0.02

0.01 0.02

0 09 09 10 11 12 12 13 14 14 15 16 16 17 0.01 08 08 09 10 11 11 12 13 13 14 15 15 16
:15 :56 :37 :18 :00 :41 :22 :03 :45 :26 :07 :48 :30 :15 :56 :37 :18 :00 :41 :22 :03 :45 :26 :07 :48 :30

FIGURE 3: INTRA-DAY PROPORTION OF VOLUME FOR THE DIFFERENT FIGURE 4: INTRA-DAY PROPORTION OF VOLUME FOR THE
EXCHANGES (CAC 40, JULY 2009) DIFFERENT EXCHANGES (FTSE 100, JULY 2009)

0.08 0.08
EURONEXT LSE
CHI-X CHI-X
0.07 0.07
TURQUOISE TURQUOISE
BATS BATS
0.06 0.06

0.05 0.05

0.04 0.04

0.03 0.03

0.02 0.02

0.01 09 09 10 11 12 12 13 14 14 15 16 16 17 0.01 08 08 09 10 11 11 12 13 13 14 15 15 16
:15 :56 :37 :18 :00 :41 :22 :03 :45 :26 :07 :48 :30 :15 :56 :37 :18 :00 :41 :22 :03 :45 :26 :07 :48 :30

Source: Crédit Agricole Cheuvreux Quantitative Research

In Navigating Liquidity 2, we highlighted that each trading destination had specific intra-day
volume seasonality (i.e. the volume profile). The main facts worth noting were:
Q Only primary markets are able to provide enough liquidity at the opening;
Q Chi-X and BATS offer their best liquidity when the U.S. market is open;
Q Conversely, Turquoise does not really benefit from higher volumes after the U.S.
markets open.

4 www.cheuvreux.com
September 2009 Navigating Liquidity 3

Comparing Figures 1 and 3, as well as Figures 2 and 4, shows that these specific aspects
have disappeared. The most significant trend concerns Turquoise, whose volumes were
Intra-day volume almost uniformly distributed throughout the trading day, whereas they now follow the classic
seasonality is now much U-shape with a spike at the preopening period for U.S. markets (which is when most of the
the same on primary news and macro-economic figures are disclosed) and a substantial increase in volumes
markets and MTFs when U.S. cash equity markets open.
Turquoise's liquidity now looks more natural with respect to this seasonality. This is not
surprising, as Figures 1 and 2 have been built using data from a time when there were still
liquidity agreements with non-professional market makers. As this very specific population,
who used to provide more than half of Turquoise's liquidity, is no longer present, this feature
was likely to fade.
The intra-day volume profile of Chi-X has evolved to a lesser extent, it is merely due to the
fact that Chi-X was already a mature MTF with sound liquidity sought after by numerous
investors. Nevertheless, at the U.S. market open, it seemed that Chi-X was more sensitive
to increases in the liquidity flow than the primary markets. This trend has disappeared and
there are now very few differences between Chi-X and primary markets. To be more
precise, there are only two slight differences: the first and the last 15 minutes of the
continuous trading phase.
This is not surprising, and these differences will probably not disappear altogether, as there
is a fundamental reason for this, namely the fixing auctions. These continue to be phases in
which the primary markets still have a monopoly.
Concerning the opening fixing auction, this is evidence that in terms of price discovery,
investors are more confident in the primary markets. The opening often reveals strong
uncertainty about the price, as this is the time when the spread is the highest during the
day. As there is no volume traded at an MTF opening auction, the latter's order book at the
beginning of the continuous phase does not have good depth or a tight spread, and
These slight differences
investors have to wait a while for it to fill with limit orders, to be able to execute on these
at the opening and the
closing are likely to venues.
continue to exist as long These slight differences will continue to exist as long as MTFs do not manage to
as investors are not compete successfully for the flow at fixing auctions. This is not likely to happen until
interested in using MTFs investors become convinced that the primary markets do not host the price formation
for fixing auctions process more efficiently than other execution venues.
In our view, this will be quite a long process, and investor behaviour will not change until
primary markets attract less than half of the non-OTC volume.

5 www.cheuvreux.com
September 2009 Navigating Liquidity 3

Q Chi-X continues to do well and to attract more and more

flows

FIGURE 5: CHI-X MARKET SHARE (EXCLUDING FIXING AUCTIONS) FIGURE 6: CHI-X COVERAGE LIQUIDITY METRIC

25 60
FTSE 100 FTSE 100
FTSE 250 FTSE 250
CAC40 CAC40
DAX 50 DAX
20
AEX AEX
BEL20 BEL20
SMI SMI
40
15

30

10
20

5
10

0 03 03 04 04 05 05 06 06 06 07 07 08 08 0 03 04 04 05 05 06 06 06 07 07 08 08 08
/06 /20 /04 /18 /03 /17 /01 /16 /30 /15 /29 /13 /28 /31 /12 /25 /07 /20 /01 /14 /26 /09 /21 /03 /15 /28
/20 /20 /20 /20 /20 /20 /20 /20 /20 /20 /20 /20 /20 /20 /20 /20 /20 /20 /20 /20 /20 /20 /20 /20 /20 /20
09 09 09 09 09 09 09 09 09 09 09 09 09 09 09 09 09 09 09 09 09 09 09 09 09 09

FIGURE 7: CHI-X PERCENTAGE OF TIME AT BEST BID AND OFFER FIGURE 8: CHI-X PERCENTAGE OF TIME AT BEST BID AND
OFFER WITH GREATEST SIZE

70 20
AEX AEX
BEL20 18 BEL20
60
CAC40 CAC40
DAX 16 DAX
50 FTSE100 FTSE100
14
SLI SLI
EUROSTOXX50 EUROSTOXX50
12
40
10
30
8

20 6

4
10
2

0 03 04 05 06 07 08 0 03 04 05 06 07 08
/20 /20 /20 /20 /20 /20 /20 /20 /20 /20 /20 /20
09 09 09 09 09 09 09 09 09 09 09 09

Source: Crédit Agricole Cheuvreux Quantitative Research

Figures 5-8 above show the main metrics enabling an analysis of how steadily Chi-X is
continuing to spread its influence over Europe. The first metric is the usual market share, for
which fixing auctions have not been taken into account, as there is no competition on these
phases. The second is coverage, as introduced in Navigating Liquidity 2 and monitored in
our Monthly Market Indicators. The last two metrics in these analyses represent:
Q Figure 7: The percentage of time that the best bid and best offer on Chi-X is as
good as the best bid and offer among the primary markets and the other MTFs (the sum
of this metric for all considered venues can be higher than 100 because different venues
can share the best quotations).
Q Figure 8: The percentage of time that Chi-X has the best bid and offer, and has the
greatest size on both sides among the primary markets and the other MTFs (the sum for
all considered venues is less than 100, because venues never have the same best bid
and offer as well as the same sizes at these prices).

6 www.cheuvreux.com
September 2009 Navigating Liquidity 3

With regard to the main European indices (see Figure 5), Chi-X clearly gained market share
from March to July, but the trend slowed down in August. The overall increase of market
share ranges from 4% (CAC 40 and AEX) to 8% (FTSE 250). Furthermore, this market share
increase has been achieved along with increasing coverage, confirming Chi-X's ability to
offer good liquidity on all components of these indices, not just the more liquid stocks.
The analysis of percentage of time at the best EBBO is also flattering for Chi-X which,
exception made of the DAX, shows a clear uptrend. If we displayed the percentage of time
at the EBBO for primary markets, we would see that for the AEX, CAC 40 and FTSE 100,
Chi-X's figures reached the same level as those of the primary markets in July. The DAX is
an exception regarding this metric, as Chi-X has beaten XETRA in this field since April 2009.
The decrease in August is directly linked to the decline in market share, which would
surprise no one, as SOR technology cannot but have such an effect. The analysis of the
percentage of time at the EBBO with the greatest size is a harder task as these figures may
be decreasing for each execution venue. Hence it is important to note that this percentage
of time has also been decreasing for primary markets from March to June.
FIGURE 9: CHI-X MARKET SHARE (EXCLUDING FIXING AUCTIONS) FIGURE 10: CHI-X COVERAGE LIQUIDITY METRIC

20 60
OMX STOCKHOLM 30 FTSE 100
OMX HELSINKI 25 FTSE 250
18
OMX COPENHAGEN 20 CAC40
SBF120\CAC40 50 DAX
16 AMX AEX
MDAX BEL20
SMI
14 40

12
30
10

8 20

6
10
4

2 03 03 04 04 05 05 06 06 06 07 07 08 08 0 03 04 04 05 05 06 06 06 07 07 08 08 08
/06 /20 /04 /18 /03 /17 /01 /16 /30 /15 /29 /13 /28 /31 /12 /25 /07 /20 /01 /14 /26 /09 /21 /03 /15 /28
/20 /20 /20 /20 /20 /20 /20 /20 /20 /20 /20 /20 /20 /20 /20 /20 /20 /20 /20 /20 /20 /20 /20 /20 /20 /20
09 09 09 09 09 09 09 09 09 09 09 09 09 09 09 09 09 09 09 09 09 09 09 09 09 09

Source: Crédit Agricole Cheuvreux Quantitative Research

The progression on less liquid stocks (see Figure 9) is very impressive. Except for the MDAX
and AMX there was clear exponential growth in May and June. However, as noted above,
this increase has also slowed down in the past two months. The overall increase in market
share ranges from 2.2% (AMX and MDAX) to 9%: for the OMX Helsinki. Nevertheless, there
is massive variance in these market shares, as shown by the erratic behaviour over the past
two months. So one might have to be careful: it is unclear whether these market shares will
noticeably increase very soon, or even if they will settle at this level.
Lastly, Chi-X is well established as the leading MTF and the No. 3 execution venue in
Chi-X is well established
Europe. It is still extending its influence in less liquid stocks, as well as other European
as the leading MTF and
indices such as Nordic ones. Nevertheless, this summer's figures in terms of market share
the No. 3 execution venue
and the percentage of time at the EBBO show that primary markets may have found a way
in Europe. It is still
to curb their decline. Whatever the reason for this break in Chi-X's progression in market
extending its influence in
share, it will be interesting to see whether this is just a lull or if there is indeed a trend
less liquid stocks
reversal.

7 www.cheuvreux.com
September 2009 Navigating Liquidity 3

Q Turquoise has done a very good job and is recovering its

market share

FIGURE 11: TURQUOISE MARKET SHARE FIGURE 12: TURQUOISE MARKET SHARE
(EXCLUDING FIXING AUCTIONS) – MAIN INDICES (EXCLUDING FIXING AUCTIONS) – OTHER INDICES

12 5
FTSE 100 OMX STOCKHOLM 30
FTSE 250 4.5 OMX HELSINKI 25
CAC40 OMX COPENHAGEN 20
10 DAX SBF120\CAC40
4
AEX AMX
BEL20 MDAX
SMI 3.5
8
3

6 2.5

2
4
1.5

1
2
0.5

0 03 03 04 04 05 05 06 06 06 07 07 08 08 0 03 03 04 04 05 05 06 06 06 07 07 08 08
/06 /20 /04 /18 /03 /17 /01 /16 /30 /15 /29 /13 /28 /06 /20 /04 /18 /03 /17 /01 /16 /30 /15 /29 /13 /28
/20 /20 /20 /20 /20 /20 /20 /20 /20 /20 /20 /20 /20 /20 /20 /20 /20 /20 /20 /20 /20 /20 /20 /20 /20 /20
09 09 09 09 09 09 09 09 09 09 09 09 09 09 09 09 09 09 09 09 09 09 09 09 09 09

FIGURE 13: TURQUOISE COVERAGE LIQUIDITY METRIC FIGURE 14: TURQUOISE COVERAGE LIQUIDITY METRIC
– MAIN INDICES – OTHER INDICES

60 50
FTSE 100 OMX STOCKHOLM 30
FTSE 250 45 OMX HELSINKI 25
CAC40 OMX COPENHAGEN 20
50 DAX SBF120\CAC40
40
AEX AMX
BEL20 MDAX
SMI 35
40
30

30 25

20
20
15

10
10
5

0 03 04 04 05 05 06 06 06 07 07 08 08 08 0 03 04 04 05 05 06 06 06 07 07 08 08 08
/31 /12 /25 /07 /20 /01 /14 /26 /09 /21 /03 /15 /28 /31 /12 /25 /07 /20 /01 /14 /26 /09 /21 /03 /15 /28
/20 /20 /20 /20 /20 /20 /20 /20 /20 /20 /20 /20 /20 /20 /20 /20 /20 /20 /20 /20 /20 /20 /20 /20 /20 /20
09 09 09 09 09 09 09 09 09 09 09 09 09 09 09 09 09 09 09 09 09 09 09 09 09 09

Source: Crédit Agricole Cheuvreux Quantitative Research

With regard to the main indices (see Figure 11), there are two cases to highlight:
Q For London and Switzerland, Turquoise has not merely recovered its market share,
but already increased it by approximately 2%. Turquoise's market share on the FTSE
250, which had not really suffered from the end of the liquidity agreements, as it was
near 1%, has seen an impressive increase of more than 5%.
Q For the CAC 40, AEX, BEL20 and DAX, the market share recovery has not been
completed. The worst case has been the AEX, for which the loss is 5%; this was the
index with the highest potential loss, as it represented the best market share. For these
indices, the downward movement began at the end of August, and this trend continued
in the first week of September. More work thus has to be carried out on these segments
to curb this decline, and try to fully recover the lost market share.
With regard to Nordic indices and less liquid stocks, few conclusions can be drawn given
the huge differences in market share. Nevertheless, Turquoise seems to have done a good
job on these segments as well and gained significant market share.

8 www.cheuvreux.com
September 2009 Navigating Liquidity 3

The trend in coverage, The trend in coverage, which has always been one of Turquoise's strengths, is also
which has always been very good. European indices, independently of the execution venues, have experienced
one of Turquoise's such an evolution: looking at coverage on any execution venue, one can see that these
strengths, is also very figures are increasing, which means that activity is expanding in the less active components
good of European indices. Nevertheless, the end of liquidity agreements could have also meant
an end to Turquoise's ability to offer liquidity on the whole index, but this did not happen.
Furthermore, the newly captured market share on the AMX, MDAX, SBF 120\CAC 40 (i.e.,
the SBF 120 excluding the CAC 40, and sometimes known as the SBF 80) and OMX
Stockholm started with the more liquid stocks on these indices, but has quickly spread to all
the components, reaching coverage that is often close to that of the primary markets.

FIGURE 15: TURQUOISE PERCENTAGE OF TIME AT BEST BID AND OFFER

45
AEX
40 BEL20
CAC40
35 DAX
FTSE100
30 SLI
EUROSTOXX50
25

20

15

10

0 03 04 05 06 07 08
/20 /20 /20 /20 /20 /20
09 09 09 09 09 09

Source: Crédit Agricole Cheuvreux Quantitative Research

Turquoise's good Lastly, Turquoise's very good performance has also been confirmed by the increase in
performance has also its percentage of time at the EBBO, which is clearly a key factor in gaining market
been confirmed by the share in an environment where more and more investors are using SORs. Special
increase in its mention has to be made for the Swiss market, where the increase is impressive. There was
percentage of time at a clear jump in Turquoise's market share on the SMI at the end of May: it virtually doubled in
the EBBO two weeks. It is also worth noting that the increase in Turquoise's percentage of time at the
EBBO with the greatest size has increased significantly over the past six months, unlike that
of primary markets and Chi-X.

9 www.cheuvreux.com
September 2009 Navigating Liquidity 3

Q After taking into account BATS' temporary aggressive rebates,

its performance is disappointing

FIGURE 16: BATS MARKET SHARE (EXCLUDING FIXING AUCTIONS) – FIGURE 17: BATS MARKET SHARE (EXCLUDING FIXING
MAIN INDICES AUCTIONS) – OTHER INDICES

8 3
FTSE 100 OMX STOCKHOLM 30
FTSE 250 OMX HELSINKI 25
7 CAC40 OMX COPENHAGEN 20
DAX 2.5 SBF120\CAC40
AEX AMX
6 BEL20 MDAX
SMI
2
5

4 1.5

3
1

0.5
1

0 03 03 04 04 05 05 06 06 06 07 07 08 08 0 03 03 04 04 05 05 06 06 06 07 07 08 08
/06 /20 /04 /18 /03 /17 /01 /16 /30 /15 /29 /13 /28 /06 /20 /04 /18 /03 /17 /01 /16 /30 /15 /29 /13 /28
/20 /20 /20 /20 /20 /20 /20 /20 /20 /20 /20 /20 /20 /20 /20 /20 /20 /20 /20 /20 /20 /20 /20 /20 /20 /20
09 09 09 09 09 09 09 09 09 09 09 09 09 09 09 09 09 09 09 09 09 09 09 09 09 09

FIGURE 18: BATS COVERAGE LIQUIDITY METRIC – MAIN INDICES FIGURE 19: BATS COVERAGE LIQUIDITY METRIC – OTHER
INDICES

50 50
FTSE 100 OMX STOCKHOLM 30
45 FTSE 250 OMX HELSINKI 25
45
CAC40 OMX COPENHAGEN 20
DAX SBF120\CAC40
40
AEX 40 AMX
BEL20 MDAX
35 SMI
35
30
30
25
25
20
20
15

15
10

5 10

0 03 04 04 05 05 06 06 06 07 07 08 08 08 5 03 04 04 05 05 06 06 06 07 07 08 08 08
/31 /12 /25 /07 /20 /01 /14 /26 /09 /21 /03 /15 /28 /31 /12 /25 /07 /20 /01 /14 /26 /09 /21 /03 /15 /28
/20 /20 /20 /20 /20 /20 /20 /20 /20 /20 /20 /20 /20 /20 /20 /20 /20 /20 /20 /20 /20 /20 /20 /20 /20 /20
09 09 09 09 09 09 09 09 09 09 09 09 09 09 09 09 09 09 09 09 09 09 09 09 09 09

Source: Crédit Agricole Cheuvreux Quantitative Research

At the end of March, Turquoise and BATS had similar market share on the main indices,
apart from the FTSE 100 and SMI. With regard to the FTSE 100, Turquoise had 1% more
than BATS, and for the SMI BATS's market share has never reached 1%. For the other
indices, BATS was either slightly ahead or the difference was negligible.
Five months later, BATS has been beaten by Turquoise on all indices. The worst
performances for BATS concern the AEX and FTSE 100 (if SMI is excluded from the
analysis), for which the difference in market share has widened by 3%.
Maybe it is not fair to compare their performances, as their initial situations were not really the
same. Turquoise had very high market share, which subsequently fell, so there were
probably many more investors connected to Turquoise at the end of March than there were
to BATS. Therefore, Turquoise merely had to attract better quotes to its order book, whereas
BATS also had to find a way to convince more people to connect to it.

10 www.cheuvreux.com
September 2009 Navigating Liquidity 3

To achieve this goal BATS chose to use inverted pricing, meaning that for each trade it had
to pay something (receiving 0.2bp from the liquidity taker and paying 0.4bp to the liquidity
BATS's inverted pricing maker). BATS used this inverted pricing for Euronext segments (French, Belgian and
had a major effect on its Dutch stocks) throughout June. This had the expected impact as its market share on
market share, which the CAC 40 reached 8% in mid-June, a very good performance compared to its 4%
doubled on the CAC 40 market share at the end of May. Figure 17 shows that its market share on the
in the space of two SBF 120\CAC 40 also doubled. However, this had a more lukewarm effect on its AEX, AMX
weeks and BEL 20 market share. For example, the market share gain on the AEX during this period
is clearly more abrupt but also rather close to that of the FTSE 100 and DAX.
Perhaps this last point is indeed good news, as it would mean that more investors trading
on the whole European landscape have set up a connection to BATS, and that the growth
in market share has resulted more from these new connections than from already
connected investors directing their flow to BATS in order to benefit from the special pricing.
In any case, it is fairly disappointing to see that BATS's market share at the end of August is
at virtually the same level as at end May. In addition, the stagnating coverage (except for the
FTSE 100) is indeed (as we have said above, it has increased on every other venue) a
regression. This shows that liquidity on BATS is concentrating on the more liquid stocks of
the index (relative to other execution venues, which are gaining market share on less liquid
stocks).

FIGURE 20: BATS PERCENTAGE OF TIME AT BEST BID AND OFFER

45
AEX
40 BEL20
CAC40
35 DAX
FTSE100
30 SLI
EUROSTOXX50
25

20

15

10

0 03 04 05 06 07 08
/20 /20 /20 /20 /20 /20
09 09 09 09 09 09

Source: Crédit Agricole Cheuvreux Quantitative Research

Moreover, the comparison of percentage of time at the EBBO is once again largely in favour
of Turquoise, and as is the case for market share, these figures have been on a downtrend
Taking into account since June, with the exception of the FTSE 100.
market share, coverage,
and percentage of time Taking into account market share, coverage, and percentage of time at the EBBO, the
at the EBBO, the best best performance for BATS has involved UK stocks. The latter have been the target of its
performance for BATS new inverted pricing in September. During the first week of September there was a
has involved UK stocks significant increase in market share on the FTSE 100, whereas other indices were either
stagnating or decreasing.

11 www.cheuvreux.com
September 2009 Navigating Liquidity 3

II— How can tick size be "optimal"?

In our first issue of Navigating Liquidity, we highlighted the connection between spread and
tick size. To be more precise, the context was: speaking about execution cost, the bid-ask
spread is an important component, and for some stocks, a large tick size can constrain the
bid-ask spread to be large (as the spread can obviously not be smaller than the tick size). As
an example, we highlighted the fact that Alcatel-Lucent had a smaller tick size on Chi-X than
on Euronext, and that this enabled a smaller spread on Chi-X, even smaller during a whole
day than the tick size on Euronext. This has consequences on an execution venue’s market
share, as it is the exact aim of SOR technologies to benefit from such better quotes. We
also illustrated with this specific example that there was no evidence of market depth
decreasing following a reduction in the tick size.
Furthermore, we indicated how to calculate in which cases (which fees for two competing
venues and which tick size) an investor had an incentive to quote a better price on an
alternative venue than the one displayed in the primary market because he would earn the
same amount as on the primary market and because it would also be beneficial for the
participant that would fill his order with a market order.

Q Tick size is clearly linked to an MTF's market share

FIGURE 21: TURQUOISE MARKET SHARE FIGURE 22: BATS COVERAGE MARKET SHARE
(EXCLUDING FIXING AUCTION) (EXCLUDING FIXING AUCTION)

20 14
BARC.L BARC.L
18 BP.L BP.L
LLOY.L 12 LLOY.L
RBS.L RBS.L
16
XTA.L XTA.L

14 10

12
8
10
6
8

6 4

4
2
2

0 03 03 04 04 05 05 06 06 06 07 07 08 08 0 03 03 04 04 05 05 06 06 06 07 07 08 08
/06 /20 /04 /18 /03 /17 /01 /16 /30 /15 /29 /13 /28 /06 /20 /04 /18 /03 /17 /01 /16 /30 /15 /29 /13 /28
/20 /20 /20 /20 /20 /20 /20 /20 /20 /20 /20 /20 /20 /20 /20 /20 /20 /20 /20 /20 /20 /20 /20 /20 /20 /20
09 09 09 09 09 09 09 09 09 09 09 09 09 09 09 09 09 09 09 09 09 09 09 09 09 09

Source: Crédit Agricole Cheuvreux Quantitative Research

Such a situation has recently recurred, and has led to intense debates. This year, MTFs
started working collaboratively in order to find a way to harmonise tick sizes in Europe. The
LIBA (London Investment Banking Association) and the FESE (Federation of European
Securities Exchanges) then joined these discussions. This led to the FESE tables, a list of
four different tick size regimes that should become the standards in European trading.
At the beginning of June, Chi-X decided to implement smaller tick sizes for Danish,
Norwegian, Spanish and Swedish stocks. The successful result of this change can easily
be read on Figure 9. Then, on 8 June, Turquoise chose to reduce the tick sizes for five UK
blue chip stocks. One week later, Turquoise implemented smaller tick sizes for five more UK
stocks and five Italian stocks. On the same day, BATS followed in the steps of Chi-X and
Turquoise, and implemented the same tick sizes as Chi-X (for Danish, Norwegian, Spanish
and Swedish stocks) and Turquoise (for the UK and Italian stocks). The consequences, in
terms of market share, of the step taken by Turquoise and BATS can easily be seen in
Figures 21 and 22:

12 www.cheuvreux.com
September 2009 Navigating Liquidity 3

The venue that offers a The venue that offers a tick size that best fits a stock is immediately rewarded with an
tick size that best fits a increase in market share.
stock is immediately LIBA and FESE organised a conference call on 16 June, during which MTFs and exchanges
rewarded with an agreed not to implement smaller tick sizes until a meeting on 30 June. On 22 June, the LSE,
increase in market share Chi-X, Turquoise, BATS Europe and Nasdaq OMX Europe standardised the tick sizes where
differences had appeared.
On 30 June, a FESE press release announced that the MTFs were now committed to
adopting the same tables as the domestic venues, and to make no further moves in their
tick size regimes outside the timetable agreed for each market.
This led to the harmonisation of tick sizes on the BEL 20, AEX 25, AMX 25, CAC 40, CAC
next 20, SBF 120 and PSI 20 (all of these indices are traded on Euronext) on 16 July.
Q The potential dangers and benefits of a tick size reduction
Reducing tick size, for Indeed, there has been quite extensive academic literature in the past decade on the
an intensively traded subject of optimal tick size, and studies have rarely reached the same conclusion. The only
stock that has a spread
consensus is that reducing tick size, for an intensively traded stock that has a spread
mostly equal to one tick,
mostly equal to one tick, will reduce the realised spread. And this is a good thing,
will reduce the realised
because the realised spread is part of execution costs. Some observers have therefore
spread
predicted that such a reduction in the tick size would increase market turnover as execution
costs were lowered.
But there are arguments against tick size reduction, among which:
Q It would trivialise time priority and thus lead to higher negotiation costs, as there
would be a higher range of possible prices;
Q It would reduce market depth, meaning that for a large order, it would be more
expensive to execute in one go. This is very hard to assess because of the lack of good
measures of 'resiliency', i.e. the market’s ability to quickly renew the liquidity that has just
disappeared.
It is also frequently stated that reducing the tick size will reduce the profitability of limit market
orders. It is true that if the tick size is large and you are relatively sure you can fill your order,
you might have an incentive to post a limit order. But conversely, as everyone will be thinking
the same way, you will have huge volume on the best bid and offer, so the probability of
your order being filled in a reasonable time will be reduced. What will the investor really gain
Reducing tick size may from using limit orders?
lead to the use of more
hidden orders, making Furthermore, many studies have shown that it was unclear whether market depth was
the potential depth loss lowered, but it was clear that more hidden orders such as icebergs were used in small
even harder to evaluate tick size contexts.
This is a part of the very difficult question raised by tick size: dynamic and hidden liquidity
versus static displayed liquidity. The former definitely requires more research and monitoring
to be exploited.
But even more so, some studies have concluded that lowering tick size (where appropriate)
would reduce volatility, some have said that it would increase the profitability of front-running
strategies, and finally some people have warned about more executions increasing
execution costs or the complexity of gathering data.
Many people have ideas on the subject, but no one has given a satisfactory answer to this
difficult question. Markets have thus globally reduced tick size in the past decade in an
attempt to lower the spread, but are careful about the potential detrimental effects on
liquidity.

13 www.cheuvreux.com
September 2009 Navigating Liquidity 3

Q Each stock needs its own tick size

FIGURE 23: LSE RELATIVE VWAS ON FTSE 100 (AUGUST 2009) FIGURE 24: EURONEXT RELATIVE VWAS ON CAC 40 (AUGUST
2009)

1
10

1
10

0
10
2 3 1 2
10 10 10 10

FIGURE 25: BORSA ITALIANA RELATIVE VWAS ON MIB 30 FIGURE 26: XETRA RELATIVE VWAS ON DAX (AUGUST 2009)
(AUGUST 2009)

1
10

1
10
0
10

1 1 2
10 10 10

Source: Crédit Agricole Cheuvreux Quantitative Research

In Figures 23 to 26, the VWAP is plotted on the x-axis, and the VWAS (volume weighted
average spread) divided by the VWAP (and expressed in basis points) is plotted on the
y-axis, with both axes on a log scale. Furthermore, one point stands for one stock on one
day of August, and we have used colour to indicate the stock’s turnover on that particular
day: the redder the point, the higher the turnover. We also added the tick size for the various
price ranges divided by the price; this is the black line that sets the lower boundary of the
spread.
What is very clear from these charts is that exchanges do not agree on optimal tick size.
Euronext has the lowest one, followed by XETRA, and then the London Stock Exchange
Group (the Italian stock exchange is part of the London stock exchange group).
The consequence is also fairly clear: the London Stock Exchange Group’s tick policy
constrains the spread to be higher than it would be otherwise. Once again, we are not
claiming that this is a bad thing for overall liquidity, but rather that if one would like to reduce
spreads then there is a way to do so.

14 www.cheuvreux.com
September 2009 Navigating Liquidity 3

A good tick size not Regardless of the methods used to determine optimal tick size, it is likely that no one can
only depends on the give a universal grid (dependent on the price range, anyone would agree on this point).
price of the stock, Indeed, it has to depend on the liquidity of the stock (the higher the turnover, the lower
but also on the the spread), and it would need to be reviewed on a frequent basis, as volatility is
liquidity of the stock determinant for the spread and would inevitably change the optimal tick size.
and its volatility
In addition, the higher the tick size, the more crucial it is to be part of the queuing process
on each price tick. Thus, it is worth waiting when tick size is high, whereas it is efficient to
simply take the liquidity aggressively when the spread is very low because of a small tick
size. From this point of view, "Fill or kill" or similar orders are more suited to small tick stocks
than to large tick stocks. This means that penalising such very fast order types (for instance
via a fee schedule) is equivalent to increasing the tick size.

15 www.cheuvreux.com
September 2009 Navigating Liquidity 3

III— How to route orders in a fragmented

market?
Q Focus on atomic orders
FIGURE 27: MARKET IMPACT VS. MARKET RISK FIGURE 28: OPTIMAL TRADING CURVES
Exec of 50,000 shares for FTE.PA
5

4.5 Market risk

Market impact
4 Overall risk

3.5

3
Risk cost (bp)

2.5

1.5

0.5

0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Pct of the day

Source: Crédit Agricole Cheuvreux Quantitative Research

Most large orders are now temporally split by traders using decision support tools and
trading algorithms. It is well known that this split has to follow a few rules:
Q On the one hand, the market impact of orders has to be estimated (no temporal split
leading to a large order that would impact the market and depreciate the price obtained);
Q On the other hand, the market risk has to be considered: waiting too long can
expose the order to opposite moves in the price.
A subtle balance must be found between these two effects. Practitioners and academics
have proposed using a parameter called "risk aversion", whose meaning is similar to that
used in portfolio allocation: if you fear risk (high risk aversion), you will accept paying a high
market impact to reduce your exposure to future risk. If you are ready to accept risk (low risk
aversion), you will accept being exposed to market risk and try to use time to reduce your
market impact. Figure 27 shows how the market impact and the market risk vary with the
time you take to trade a given number of shares (horizontal axis). The black line is the
combined risk: for this risk aversion parameter, we see the line decreasing first (it is worth
waiting) before increasing (the market risk after 10% of the day does not balance the
expected gain in market impact).
Because the market risk and the market impact are not linear functions of the time and the
quantity to be traded, and because volume and volatility experience the usual intra-day
variation, solving this optimisation leads to optimal trading curves that are not straight lines,
like the one used by CA Cheuvreux’s algorithms. Figure 28 shows three trading curves
corresponding to three different market conditions (see "Rigorous Strategic Trading:
Balanced Portfolio and Mean-Reversion", The Journal of Trading, Vol. 4, No. 3. (2009), pp.
40-46, by Charles-Albert Lehalle).
Before market fragmentation, large orders were only temporally split according to such
optimisation schemes, but now each temporal slice itself (each "atomic order", also called
"child order", "slice" or "occurrence") has to be split and sent to different trading destinations
according to the liquidity they offer. This part of the study is dedicated to an explanation of
the mechanisms that take place during this spatial split.

16 www.cheuvreux.com
September 2009 Navigating Liquidity 3

Q Using a Smart Order Router (SOR)

FIGURE 29: FUNCTIONAL DIAGRAM OF A SMART ORDER ROUTER

Venue
1
SOR 5. Log for post
trade
Incoming Venue
orders 2
1. Listen to 4. Take exec
market data into account
Venue
N
2. Take the 3. Send
decision orders

Source: Crédit Agricole Cheuvreux Quantitative Research.

Assuming that each "slice" is sent at the better instant given the information available to the
trader and his decision support systems, assuming also that this slice has a well-defined
"limit price" and exact quantity, it is possible to focus on the best way to split the quantity
among available trading venues. It is obvious that such a split will be optimal as:
Q It obtains a price better than or equal to the asking price;
Q It finishes the order as soon as possible (because the temporal aspect has been
solved at a higher level).
An optimal split will The best way to use the liquidity available in alternate venues is to avoid crossing a tick price
accelerate the on the main trading destination if volume is available at the same price on another venue: it
execution, improve the will accelerate the execution, improve the price and minimise the footprint of the
price and minimise its overall execution in the market.
footprint

The standard device to achieve such an optimisation is named a Smart Order Router (SOR).
Other tactics such as multi-destination icebergs, liquidity seekers, etc., can also be used
given that users of these tactics are aware of the overall strategy that led to this order being
A SOR order is a new sent at this time and at this price. The specific feature of an SOR is that it is an agnostic
type of order giving device: it takes a limit or market order and splits it with advanced knowledge of the state of
access to an the liquidity on the trading destinations that are ready to trade. An SOR order is like a new
aggregated view of a type of order giving access to an aggregated view of a large set of trading venues.
set of trading venues
Crédit Agricole Cheuvreux’s SOR can be customised for different uses, mainly via a
selection of trading destinations and some split options. The better the capabilities of each
trading destination are understood, the better an SOR will be able to comply with clients'
needs. The knowledge shared in the Navigating Liquidity studies has been used
substantially to fine-tune our SOR.

17 www.cheuvreux.com
September 2009 Navigating Liquidity 3

An SOR is not only a "fire and forget" device: it also needs to be able to prove that every
decision made is in accordance with the agreed-upon "execution policy". To comply with
MiFID, the choices made by the SOR and the relevance of the execution policies must also
be reviewed. Such a trading tool thus needs to be properly connected to large data
warehouse recording policies, snapshots of the state of the market, decisions made and
executions obtained.
Figure 29 is a simplified view of the components of an SOR: the decision-making part has to
fit in with the market listeners, and the SOR has to be able to adjust its decisions with
respect to feedback from the market.

Q Aggregate liquidity to minimise the impact of each order

FIGURE 30: INDICATORS OF THE AVERAGE TRADING SIZE (ATS) FIGURE 31: INDICATORS OF THE AVERAGE TRADING SIZE (ATS)
AND OF THE DAILY NUMBER OF TRADES FOR THE FTSE 100 AND OF THE DAILY NUMBER OF TRADES FOR THE CAC 40
Indicator of the ATS Indicator of the ATS
700 60
600 50
500 40

400 30

300 20

200 10

Indicator of the no of deals Indicator of the no of deals

15 25

20
10
15

10
5
5

0 0 3/ 06 08 11 01 03 06 08 11 01 04 06 08 0 0 3/ 06
/0 7
08
/0 7
11
/0 7
01
/08
03
/08
06
/0 8
08
/0 8
11
/08
01
/09
04
/09
06
/09
08
/09
07 /0 7 /0 7 /0 7 / 08 /0 8 /0 8 /0 8 / 08 /0 9 /0 9 / 09 /0 9 07

Source: Crédit Agricole Cheuvreux Quantitative Research

Figures 30 and 31 show that while the average trading size (ATS) of an order has decreased
overall since early 2007, the daily number of deals has increased regularly. This is a well-
known consequence of the development of electronic trading.
While the acceleration of the decrease of the ATS due to MiFID and the financial crisis can
As long as the size of easily be seen, the slope of the increase of the daily number of deals did not change with
aggressive orders is MiFID. This probably stems from the fact that the current average trading size is conditioned
larger than the size of by the size of passive orders. As long as the average size of aggressive orders is larger
passive orders, any than the size of posted orders, any decrease in the size of aggressive orders has no
decrease in the size of impact on the ATS or on the number of deals.
the former has no
impact on the number For instance, if an aggressive order of 150 consumes two posted orders of 100 and 50 at
the same limit price (say 10.00), two deals will occur and the ATS for these two transactions
will be 75, which is the average size of the two passive orders.
Therefore, these charts have to be read as indicators of the behaviour of passive orders:
with market fragmentation, the ATS goes down, and moreover, passive orders are now
spread over more trading destinations. An SOR’s main task is thus to be able to build a
consolidated view of all these small quantities to avoid a dramatic decrease in execution
quality.
Going back to the previous example: if the two posted orders are no longer on the same
trading venue, if the sender of the marketable order (say it is a buy) does not use an SOR,
he will need to take quantity at a higher price after 100.

18 www.cheuvreux.com
September 2009 Navigating Liquidity 3

Q Beware of duplicated liquidity

FIGURE 32: SIMPLIFIED DIAGRAM OF TRAJECTORIES OF MESSAGES

Priority / queuing
Market data management

Incoming
orders Broker Trading
systems venue Matching
(including Orders gateway Engine
Outgoing
executions SORs)
Executions

Backbone / hardware

Source: Crédit Agricole Cheuvreux Quantitative Research

The very first issue of Navigating Liquidity placed the emphasis on market makers (or traders
with an activity very similar to market makers, such as high frequency arbitrageurs) that
become heavily very involved in MTFs when they are created.
High-frequency prop Their typical activity is to be present simultaneously at the bid and ask of the same security
trade tactics lead to on several trading destinations. They agree to take on some market risk (because they
duplicated quantities in maintain an inventory rather than immediately unwinding their positions), and are "rewarded"
order books that makes by a fraction of the market bid-ask spread. The maker / taker fee schedule of many MTFs
the SOR’s task more helps high frequency traders be profitable. These HF traders invested substantially in
complex technology, and each one generally focuses on a small subset of instruments whose
behaviour is compliant with their trading robots.
Once market makers have posted liquidity on the same side of the order books on several
trading venues, their inventory-keeping strategies often imply that as soon as one of those
orders is executed, all the remaining ones are immediately cancelled. This leads to
"duplicated" quantities in the order books, which makes the SOR’s task more
"Pre-trade transparency" complex.
is not so useful to guess Adding pegged and iceberg orders, it is now clear that so-called "pre-trade
where liquidity really is transparency" is not very useful to guess where liquidity really is. Post-trade
transparency (i.e. trade reports) give far more information to investors. This makes small dark
pools (with no pre-trade transparency but clear post-trade transparency) more attractive than
some months ago.
Figure 32 gives keys to understanding how latency is an issue in such a context: it is a
simplified view of an order’s trajectory. An SOR’s goal is to obtain an execution that is in line
with the information used to take its decision:
Q The first step is to build an aggregated view of the market data; this view is usually
built on the fly at the rhythm of the updates coming from the market;
Q Then the SOR has to compute its decision according to its tactic. This is usually very
fast because the only limiting factor is the clock of the server it runs on.
Q After that, the orders sent have to go from the SOR to the trading venues. If the
trading venues are close to each other, the SOR can be positioned in the same area,
but this is not always the case and physical distance cannot be crossed instantaneously.

19 www.cheuvreux.com
September 2009 Navigating Liquidity 3

For instance, it takes light around 1.14ms to travel the 350km distance between Paris
and London; the Eurostar’s route would take 1.6ms at the speed of light, while in a usual
backbone, a message needs around 4ms to cover the same distance (see Figure A4 in
Appendix 4 for more information on the latency of worldwide roundtrips).
Q Then the orders have to reach the matching engines of the trading destinations.
Depending on its type, an order will not have the same priority (for instance, inserting a
new order will have a better rank than a modified order).
Q Finally, the execution reports have to be returned to the SOR.
All these steps give an idea of the different "latency arbitrages" that can occur during the life
The efficiency of an of an order.
SOR cannot be stated
in terms of latency but Nevertheless, the efficiency of an SOR cannot be stated in terms of latency but in
in terms of the results terms of the results obtained: for well-defined orders and given that the price obtained is
obtained: the higher better than or equal to the one that could have been obtained on the main market only, the
the quantities executed higher the quantities executed on MTFs, the more efficient the SOR.
on MTFs, the more
efficient the SOR

20 www.cheuvreux.com
September 2009 Navigating Liquidity 3

IV— Two results and a guess

Q Efficiency of CA Cheuvreux's SOR
FIGURE 33: RANGE OF FIGURE 34: COMPARISON OF MARKET SHARES FIGURE 35: COMPARISON OF MARKET SHARES
OBTAINED PRICE (UPPER) WITH THE SPLIT OBTAINED (LOWER) FOR (UPPER) WITH THE SPLIT OBTAINED (LOWER) FOR
IMPROVEMENTS (IN BP) THE FTSE 100 THE CAC 40
Market share Market share
100 100
2.2
80 80
BATS BATS
2
60 CHI-X 60 CHI-X
LSE EURONEXT
1.8 40 40
NASDAQ OMX NASDAQ OMX
1.6 20 TURQUOISE 20 TURQUOISE

0 0
1.4

1.2 Cheuvreux' aggressive split Cheuvreux' aggressive split

100 100
1
80 80

0.8 60 60

0.6 40 40

20 20
0.4
0 01 / 18 05 23 12 30 17 04 22 09 26 14 01 0 01/ 18
/0 1
05
/0 2
23
/0 2
12
/0 3
30
/0 3
17
/0 4
04
/0 5
22
/0 5
09
/0 6
26
/0 6
14
/0 7
01
/0 8
/0 1 /0 2 /0 2 /0 3 /0 3 /0 4 /0 5 /0 5 /0 6 /0 6 /0 7 /0 8 01
1 01 /0 9 /0 9 /0 9 /0 9 /0 9 /0 9 /0 9 /0 9 /0 9 /0 9 /0 9 /0 9 /0 9
/09 /09 /09 /09 /09 /09 /09 /0 9 /0 9 /0 9 /0 9 /0 9 /0 9

Source: Crédit Agricole Cheuvreux Quantitative Research

As explained above, splitting an aggressive order on several trading destinations enables

access to the same level of price for a larger volume, instead of going to the next tick of
price on the primary trading destination.
The efficiency of such a split can be quantified by the part of the execution coming from
alternative venues as long as a price improvement is obtained.

The median of the Figure 33 gives the range of price improvement obtained by CA Cheuvreux's SOR on all its
improvement is higher than aggressive orders from 1 January to end August 2009 on the DAX, CAC 40, AEX and FTSE
1 basis point; an 100 indices. This shows that the median of the improvement is higher than 1 basis
improvement of 1.8bp is point and the probability of an improvement between 0.8bp and 1.4bp is high. Client
not rare orders can even obtain a 1.8bp improvement. This is a realistic and very good
performance, as it has to be compared with less than half of the effective tick size. An order
that would consume two price ticks should have a size of around 5 ATS (~EUR50k on the
CAC 40), and if the entire quantity needed could be found on the first limit in an alternative
venue, the price improvement for a stock of EUR20 with a 0.005 tick size would be 1.25bp.
Moreover, the SOR’s efficiency can be seen in its split rate: the SOR finds more than
The SOR’s efficiency can 50% of its liquidity outside primary markets, meaning that being more concentrated on
be seen in its split rate: the primary venues would have a negative impact on the price. One of the explanations of the
SOR finds more than 50% split rate obtained is that because more agents are posting on cheap venues, consuming
of its liquidity outside the liquidity on such venues for one order increases the probability of being able to find liquidity
primary markets again for the next one.

21 www.cheuvreux.com
September 2009 Navigating Liquidity 3

Q A split driven by Chi-X efficiency

FIGURE 36: EXPLANATION OF CA CHEUVREUX'S AGGRESSIVE SPLIT

100

90

80

70

60

50

40

30

20 CAC 40 outside EURONEXT

Predictor
10 FTSE 100 outside LSE
Predictor
0 16 03 18 02 17 02 17 01 16 01 16 31 15
/0 2 /0 3 /0 3 /04 /0 4 /0 5 /0 5 /06 /0 6 /0 7 /0 7 /07 /0 8
/0 9 /09 /0 9 /0 9 /0 9 /09 /09 /0 9 /0 9 /09 /0 9 /0 9 /0 9

Source: Crédit Agricole Cheuvreux Quantitative Research

Figure 36 shows the performance of a model explaining the rate of execution by our SOR
outside the primary market for the CAC 40 and FTSE 100 indices.
Q The model obtained is linear and uses as inputs the Cheuvreux-TAG market monthly
indicators;
Q The less the LSE is present on the BBO with the greatest quantity, the fewer
executions occur on primary markets;
Q The smaller the Chi-X bid-ask spread, the fewer executions go to the primary
markets;
The performance of Q The more Chi-X is present on the bid-ask spread, the fewer executions go to the
Chi-X and the primary markets.
decreased efficiency of
the LSE is a sufficient The split rate obtained is very close to the effective execution rate. This means that the
information about the month-by-month performance of Chi-X and the decreased efficiency of the LSE give
fragmentation sufficient information about the fragmentation to be able to explain the splitting rate
outside primary markets.

22 www.cheuvreux.com
September 2009 Navigating Liquidity 3

Q Fragmentation is a consequence of primary markets' variance

FIGURE 37: SIMULATION OF A PRIMARY MARKET WITH SMALL FIGURE 38: SIMULATION OF A PRIMARY MARKET WITH LARGE
VARIANCE VARIANCE

100 100
Primary ( 76) Primary ( 76)
90 Main Altern. ( 17) 90 Main Altern. ( 17)
Altern. 2 ( 7) Altern. 2 ( 7)
80 80

70 70

60 60
Split rate

Split rate
50 50

40 40

30 30

20 20

10 10

0 0
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2 0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2

Source: Crédit Agricole Cheuvreux Quantitative Research

The preceding sections analysed the optimal split for aggressive orders, and have shown
that a natural consequence of the availability of quantities at the best bid and offer on
alternative venues was that the bigger the quantity sent to MTFs, the more efficient the
execution. For passive orders, the question is more complex, because the prices and
quantities of many of these are often reassessed. In such a context, it would be nonsensical
to produce aggregated statistics for passive orders, because such data would mix orders
that have overly divergent goals.
Nevertheless, we obtained a very interesting result on splitting passive orders that is worth
sharing. To give a clear idea of what was obtained, we will use a simulated instrument and
say that it is traded on three venues:
Q Its primary venue, concentrating 76% of the flows;
Q A main alternative venue, with 17% of the flows;
Q And another "small" alternative venue, obtaining 7% of the flows.
The illustrations come from large-scale simulations and are backed by theoretical
calculations.
The important point here is that the average size of the investor we will study is lower than
the average size traded on the market.
Figure 37 shows how the optimal split can be learned from the data: the horizontal axis is
time (in the real world, the more investors use SORs and rational tactics, the faster this time
line will be covered) and the vertical line is the optimal split rate. A "learning phase", during
which investors are trying to guess how to split optimally, can be observed. This come from
the fact that pre-trade transparency is not really good on current trading venues (iceberg,
duplicated and pegged orders), so execution reports have to be used: the most reliable
The most reliable
source of information is the messages you get back from the venues when you send
source of information is
them an order. It takes time the learn the optimal split from them.
trade reports
As expected, in such a low variance context, the optimal split for an investor using an
efficient SOR, as seen at the end of the time line of Figure 37, respects the market shares of
the various venues.

23 www.cheuvreux.com
September 2009 Navigating Liquidity 3

Figure 38 is far more informative: the market shares of the three simulated trading
destinations are the same as in Figure 37, but with a larger variance. The variance quantifies
the regularity of the flow that a venue can guarantee. For instance, let’s assume one venue
can ensure you a flow of execution between 60 and 120 (with an average of 90), while the
range of the flow for another lies between 20 and 25 (the average is 22.5). Given that the
size of your orders is around 80, your best split is 60 for venue A and 20 for venue B and
you are sure to have the order filled. This leads to a rate of 75% for venue A when its market
share is 80%.
In Figure 37, the best split for an investor with orders that are on average smaller than the
market is 50% for the exchange that has a 76% market share, 40% for the alternative one
and 10% for the smallest: this obviously transfers flows from the main venue to MTFs.
This brief illustration backs the simulation, which is in line with theoretical results: investors
whose average order size is lower than or equal to the market average will fear
variance of the flows. All studies have shown that variance is proportional with market
share, so the main markets automatically have more variance than mid-players like Chi-X or
Turquoise. As a result, most of the investors send Chi-X more than its market share.
Investors whose
average size is lower An important remark is that given the shape of the distribution of the volumes going to the
than or equal to the market, a significant percentage of investors have an average size lower than the overall
market average will fear market. The only situation where no investor has an average size lower than the overall
variance of flows; this average is when all investors have exactly the same order size… Another important point is
automatically leads to a that investors with an average size larger than the market have no reason to send more than
situation of eroding their market share to primary markets: this automatically leads to a situation of eroding
market share for the market share for the main players (some investors send less than their market share,
primary markets others send the market share: the combination of these factors is a decrease in market
share for the next month).
This explanation fits perfectly with the observations: the main market loses market share
each month, especially when the flow it provides is volatile (as is the case for the LSE).

24 www.cheuvreux.com
September 2009 Navigating Liquidity 3

Appendices
Q Appendix 1: Glossary
Price Formation Process
The price formation process covers all events occurring during trading that result in a constantly evolving market price. Such
events are, for instance: insertion of new orders or cancellation of orders, or matching of opposite orders.
Walrasian Equilibrium An equilibrium between consumers and producers to set a perfect match between
supply and demand.
Trading Destinations
MiFID removes domestic exchange concentration rules and recognises three trading destinations: Regulated Markets (RMs),
Multilateral Trading Facilities (MTFs) and Systematic Internalisers (SIs). Everything else is over-the-counter (OTC).
Primary Market Also called new issue market. Securities are issued for the first time in this market.
In this study, EURONEXT PARIS, the LONDON STOCK EXCHANGE and XETRA
are examples of primary markets.
Multilateral Trading Facility (MTF) A multilateral system operated by an investment firm or a market operator which
brings together multiple third-party buying and selling interests in financial
instruments in a way that results in a contract. Chi-X, Turquoise and BATS are the
MTFs studied here.
Dark Pools A trading destination not disclosing its order book. Trades are often reported with a
delay. These venues are said to be adapted to trade large quantities without
leaving a footprint in the market.
Smart Order Router A device routing an order across a given set of trading destinations according to a
disclosed execution policy. It can split an order into smaller ones to spray all
available destinations if needed.
Execution Costs
Execution costs are a mixture of fees, bid-ask spread, price impact, market impact, opportunity risk and market risk.
Market Risk Measurement of the uncertainty in the evolution of the price.
Price impact Impact of the volume of an aggressive order on the obtained average price.
Market Impact Possibly persistent impact of the volume of an aggressive order on the market
price.
Opportunity Risk Price of missing a transaction or obtaining a deteriorated price by not having
placed an order on the adequate trading destination.
Tick Size The minimal difference allowed between two different prices. It is defined by the
trading rules of each trading destination.
VWAS: Volume Weighted Average Spread The mean of the bid-ask spread at each trade, weighted by the volume of the
trade.
Execution Benchmark
Benchmarks define the target to be achieved during the trading of a large order.
Implementation Shortfall The average price of the order is compared to the market price before the beginning
of trading.
PVOL: Percentage of Volume The volume curve of the order compared to that of the market for the same stock.
VWAP: Volume Weighted Average Price The mean of traded prices weighted by traded volume. Cross trades are often
excluded from this computation.

25 www.cheuvreux.com
September 2009 Navigating Liquidity 3

Q Appendix 2: Methodology
Once again, the scope of this study is broader than that of the previous one. We have added another trading destination:
NASDAQ OMX.
We have also included certain indices when needed to provide some clues about how fragmentation is spreading to mid caps.
We decided to use data from March to August to be in line with our monthly "market liquidity indicators", unless explicitly stated.
Our data comes from our tick by tick database containing all market deals synchronised with the best bid and offer (prices and
quantities) on a large set of stocks, and five limit-order books on a subset of those stocks.
This database is very comprehensive as it stores information about the trading phase, the trading destination and the type of
deal (lit or dark).

What are we looking at?

Electronic markets have two main kinds of trading rules:

Rules for fixing auctions: after a pre-auction period, during which all agents post orders as liquidity providers, a
"Walrasian" equilibrium is instantaneously found to fix a price. All sellers that post a price lower than the fixing price and all
buyers that post a higher price will have their orders executed.
Rules for continuous auctions: a Limit Order Book (LOB) consists of orders posted by liquidity providers that did not
find a counterpart. The highest buying price of the LOB and the lowest selling price are called best bid and best ask
prices. Liquidity consumers have to post buy orders higher than the best ask, or sell orders lower than the best bid, thus
generating a trade.

Source: Crédit Agricole Cheuvreux Quantitative Research

When talking about liquidity, investors can be divided into two subsets: liquidity providers, who are waiting for a counterpart
and take on the risk of a reversal of markets (i.e. the "market risk") and liquidity consumers, who want to trade now, without
taking any market risk. The latter will buy their need for certainty from the former, paying them the "price impact" of their trades.
Market data natively provides an indicator for each kind of behaviour: the turnover over the previous 15 minutes is a good proxy
for the flow of liquidity consumers (i.e. investors seeking to trade immediately) and the bid-ask spread is an indicator of the
price of this urgency because this is the amount they agree to pay to trade immediately.
Another indicator is market risk. It is known that there is no way to estimate this with perfect accuracy, but a common
measurement of the uncertainty part of price moves is volatility. As its measurement is based on traded prices, it characterises
liquidity consumers. It also influences the behaviour of liquidity providers since the premium they demand has to take into
account the market risk they have to assume.

26 www.cheuvreux.com
September 2009 Navigating Liquidity 3

Q Appendix 3: Lists of stocks

Composition of indices used in the whole study:

CAC 40 Stocks
ACCOR ACCP.PA SAINT GOBAIN SGOB.PA
AIR LIQUIDE AIRP.PA SCHNEIDER ELECTRIC SCHN.PA
ALCATEL-LUCENT ALUA.PA VINCI (EX. SGE) SGEF.PA
AXA AXAF.PA STMICROELECTRONICS STM.PA
BNP PARIBAS ACT. A BNPP.PA SOCIETE GENERALE SOGN.PA
BOUYGUES BOUY.PA TOTAL SA TOTF.PA
CAPGEMINI CAPP.PA UNIBAIL-RODAMCO UNBP.PA
CARREFOUR CARR.PA VALLOUREC VLLP.PA
DANONE DANO.PA AIR FRANCE –KLM AIRF.PA
ESSILOR INTL. ESSI.PA DEXIA SA DEXI.BR
FRANCE TELECOM FTE.PA SANOFI-AVENTIS SASY.PA
L'OREAL OREP.PA ARCELORMITTAL ISPA.AS
LAFARGE LAFP.PA EADS EAD.PA
LAGARDERE S.C.A. LAGA.PA VEOLIA ENVIRONNEMENT VIE.PA
LVMH LVMH.PA VIVENDI VIV.PA
MICHELIN MICP.PA CREDIT AGRICOLE CAGR.PA
PERNOD RICARD PERP.PA GDF SUEZ GSZ.PA
PEUGEOT PEUP.PA ALSTOM ALSO.PA
PPR PRTP.PA EDF EDF.PA
RENAULT RENA.PA SUEZ ENVIRONNEMENT SEVI.PA
Source: Crédit Agricole Cheuvreux Quantitative Research

FTSE 100 Stocks

DIAGEO PLC DGE.L STANDARD CHARTERED PLC STAN.L
BARCLAYS PLC BARC.L HAMMERSON PLC HMSO.L
LLOYDS BANKING GROUP PLC LLOY.L LIBERTY INTERNATIONAL PLC LII.L
MARKS & SPENCER GROUP PLC MKS.L LONMIN PLC LMI.L
UNILEVER PLC ULVR.L REXAM PLC REX.L
BRITISH AIRWAYS PLC BAY.L SCHRODERS PLC SDR.L
BP PLC BP.L SERCO GROUP PLC SRP.L
PRUDENTIAL PLC PRU.L SEVERN TRENT PLC SVT.L
RSA INSURANCE GROUP PLC RSA.L SHIRE PLC SHP.L
VODAFONE GROUP PLC VOD.L WM MORRISON SUPERMARKETS MRW.L
CADBURY PLC CBRY.L INTERNATIONAL POWER PLC IPR.L
KINGFISHER PLC KGF.L OLD MUTUAL PLC OML.L
BAE SYSTEMS PLC BAES.L PENNON GROUP PLC PNN.L
WPP PLC WPP.L ANGLO AMERICAN PLC AAL.L
ASTRAZENECA PLC AZN.L SCHRODERS PLC-NON VOTING SDRt.L
LEGAL & GENERAL GROUP PLC LGEN.L AMEC PLC AMEC.L
PEARSON PLC PSON.L ANTOFAGASTA PLC ANTO.L
TESCO PLC TSCO.L CAIRN ENERGY PLC CNE.L
SMITHS GROUP PLC SMIN.L TULLOW OIL PLC TLW.L
CABLE & WIRELESS PLC CW.L COBHAM PLC COB.L
HSBC HOLDINGS PLC HSBA.L AMLIN PLC AML.L
LAND SECURITIES GROUP PLC LAND.L ICAP PLC IAP.L
NEXT PLC NXT.L BALFOUR BEATTY PLC BALF.L
ROYAL BANK OF SCOTLAND GROUP RBS.L AUTONOMY CORP PLC AUTN.L
BG GROUP PLC BG.L CARNIVAL PLC CCL.L
NATIONAL GRID PLC NG.L GLAXOSMITHKLINE PLC GSK.L
RIO TINTO PLC RIO.L COMPASS GROUP PLC CPG.L
THOMSON REUTERS PLC TRIL.L FRIENDS PROVIDENT PLC FP.L
SAINSBURY (J) PLC SBRY.L BT GROUP PLC BT.L
AVIVA PLC AV.L XSTRATA PLC XTA.L
SMITH & NEPHEW PLC SN.L FOREIGN & COLONIAL INVEST TR FRCL.L
BRITISH SKY BROADCASTING GRO BSY.L INTERTEK GROUP PLC ITRK.L
SCOTTISH & SOUTHERN ENERGY SSE.L ALLIANCE TRUST PLC ATST.L
INVENSYS PLC ISYS.L RANDGOLD RESOURCES LTD RRS.L

27 www.cheuvreux.com
September 2009 Navigating Liquidity 3

FTSE 100 Stocks

RECKITT BENCKISER GROUP PLC RB.L INTERCONTINENTAL HOTELS GROU IHG.L
REED ELSEVIER PLC REL.L VEDANTA RESOURCES PLC VED.L
BHP BILLITON PLC BLT.L G4S PLC GFS.L
BRITISH AMERICAN TOBACCO PLC BATS.L ADMIRAL GROUP PLC ADML.L
ROLLS-ROYCE GROUP PLC RR.L ROYAL DUTCH SHELL PLC-A SHS RDSa.L
SAGE GROUP PLC/THE SGE.L INMARSAT PLC ISA.L
IMPERIAL TOBACCO GROUP PLC IMT.L ROYAL DUTCH SHELL PLC-B SHS RDSb.L
JOHNSON MATTHEY PLC JMAT.L PETROFAC LTD PFC.L
ASSOCIATED BRITISH FOODS PLC ABF.L KAZAKHMYS PLC KAZ.L
BRITISH LAND CO PLC BLND.L DRAX GROUP PLC DRX.L
BUNZL PLC BNZL.L STANDARD LIFE PLC SL.L
CAPITA GROUP PLC CPI.L EXPERIAN PLC EXPN.L
CENTRICA PLC CNA.L HOME RETAIL GROUP HOME.L
MAN GROUP PLC EMG.L THOMAS COOK GROUP PLC TCG.L
UNITED UTILITIES GROUP PLC UU.L TUI TRAVEL PLC TT.L
WHITBREAD PLC WTB.L EURASIAN NATURAL RESOURCES ENRC.L
SABMILLER PLC SAB.L FRESNILLO PLC FRES.L
Source: Crédit Agricole Cheuvreux Quantitative Research

DAX Stocks
ADIDAS AG ADSG.DE LINDE AG LING.DE
ALLIANZ SE ALVG.DE MAN AG MANG.DE
BASF SE BASF.DE MERCK KGAA MRCG.DE
BAYER AG BAYG.DE METRO AG MEOG.DE
BAYERISCHE MOTOREN WERKE AG BMWG.DE MUENCHENER RUECKVER AG-REG MUVGn.DE
BEIERSDORF AG BEIG.DE SALZGITTER AG SZGG.DE
COMMERZBANK AG CBKG.DE RWE AG RWEG.DE
DEUTSCHE BANK AG -REG DBKGn.DE SAP AG SAPG.DE
DEUTSCHE LUFTHANSA-REG LHAG.DE SIEMENS AG SIEGn.DE
DEUTSCHE TELEKOM AG-REG DTEGn.DE E.ON AG EONGn.DE
FRESENIUS AG -VZO- FREG_p.DE VOLKSWAGEN AG VOWG.DE
FRESENIUS MEDICAL CARE AG & CO FMEG.DE DAIMLER AG DAIGn.DE
HANNOVER RUECKVERSICHERUNGS HNRGn.DE THYSSENKRUPP AG TKAG.DE
HENKEL AG+CO.KGAA VZO HNKG_p.DE DEUTSCHE POST AG-REG DPWGn.DE
K+S AG SDFG.DE DEUTSCHE BOERSE AG DB1Gn.DE
Source: Crédit Agricole Cheuvreux Quantitative Research

BEL 20 Stocks
BEKAERT NV BEKB.BR ACKERMANS & VAN HAAREN ACKB.BR
COLRUYT SA COLR.BR BEFIMMO S.C.A. BEFB.BR
DELHAIZE GROUP DELB.BR COFINIMMO COFB.BR
DEXIA SA DEXI.BR CNP -CIE NATL A PORTEFEUILLE NAT.BR
FORTIS FOR.BR OMEGA PHARMA SA OMEP.BR
SOLVAY SA SOLB.BR ANHEUSER-BUSCH INBEV NV INTB.BR
UCB SA UCB.BR GROUPE BRUXELLES LAMBERT SA GBLB.BR
UMICORE UMI.BR BELGACOM SA BCOM.BR
KBC GROEP NV KBC.BR GDF SUEZ GSZ.PA
MOBISTAR SA MSTAR.BR TELENET GROUP HOLDING NV TNET.BR
Source: Crédit Agricole Cheuvreux Quantitative Research

28 www.cheuvreux.com
September 2009 Navigating Liquidity 3

Q Appendix 4: Information on backbone latencies

FIGURE A4: INDICATIVE TIMES FOR DATA ROUNDTRIPS

Source: Hibernia Atlantic (courtesy of Roderick Beck)

29 www.cheuvreux.com
September 2009 Navigating Liquidity 3

Q Appendix 5: Liquidity metrics

MARKET SHARE LIQUIDITY METRIC, AUGUST 2009

Index Primary market Chi-X Turquoise BATS
FTSE 100 63.4% 22.5% 9.3% 4.7%
CAC 40 69.8% 18.1% 7.3% 4.7%
DAX 70.8% 19.0% 6.0% 4.2%
AEX 70.1% 18.8% 7.3% 3.9%
BEL 20 80.5% 12.8% 3.7% 3.0%
SMI 81.0% 10.8% 8.2% N/A
VWAS AND VWAS EFICIENCY LIQUIDITY METRICS, AUGUST 2009
VWAS (BASIS POINTS)
Index Primary market Chi-X Turquoise BATS
CAC 40 6 bp 5.8 bp 13 bp N/A
FTSE 100 7.1 bp 7.3 bp 15.9 bp N/A
DAX 6.8 bp 4.4 bp 11 bp 20.4 bp
AEX 7 bp 6.4 bp 12.7 bp 20.5 bp
FIXING AUCTION LIQUIDITY METRIC, AUGUST 2009
AEX BEL 20 CAC 40 DAX FTSE 100
6.3 5.3 9.2 5.5 5.2
COVERAGE LIQUIDITY METRIC, AUGUST 2009
Index Primary market Chi-X Turquoise BATS
FTSE 100 42% 41% 47% 43%
AEX 48% 41% 51% 35%
CAC 40 59% 53% 60% 46%
DAX 50% 46% 53% 52%

30 www.cheuvreux.com
September 2009 Navigating Liquidity 3

31 www.cheuvreux.com