D'Hondt C. 2008. Transaction Cost Analysis A-Z. EDHEC. France

An EDHEC Risk and Asset Management Research Centre Publication
Transaction Cost Analysis A-Z

A Step towards Best Execution in the Post-MiFID Landscape
November 2008
Sponsored by
Table of Contents
Foreword ................................................................................................................. 3
Introduction............................................................................................................. 5
I. Transaction Cost Analysis as Part of the Investment Process................. 9
II. Transaction Cost Components and Drivers...............................................13

1. Direct Explicit Transaction Costs.................................................................................................14
2. Indirect Explicit Transaction Costs..............................................................................................15
3. Implicit Transaction Costs..............................................................................................................19
4. Transaction Cost Magnitude.........................................................................................................23
5. Transaction Costs and Execution Methods..............................................................................25
III. Measuring Transaction Costs with Post-Trade Analysis.......................27

1. Benchmark Comparison..................................................................................................................28
2. Implementation Shortfall..............................................................................................................29
3. Common Pitfalls and Shortcomings in Transaction Cost Measurement.......................32
4. Measuring Transaction Costs under MiFID..............................................................................35
IV. Estimating Transaction Costs with Pre-Trade Analysis.........................39

1. Collection and Analysis of Pre-Trade Data..............................................................................40
2. Cost and Risk Estimation................................................................................................................41
3. Optimisation and Efficient Trading Frontier............................................................................53
V. Trading Performance Measurement............................................................57

1. Transaction Costs vs. Trading Performance.............................................................................58
2. Current State of the Industry and Practices...........................................................................58
3. Regulatory Pressure: MiFID and its Best Execution Obligation........................................64
VI. A New Framework: the EBEX Indicators...................................................67

1. General Presentation.......................................................................................................................68
2. Detailed Presentation of the Indicators....................................................................................70
3. Illustration of the Framework.......................................................................................................77
Conclusion..............................................................................................................85
References..............................................................................................................87
About the EDHEC Risk and Asset Management Research Centre............90
About CACEIS . .....................................................................................................94
About NYSE Euronext..........................................................................................96
About SunGard.....................................................................................................98
Printed in France, November 2008. Copyright EDHEC 2008.

The opinions expressed in this survey are those of the authors and do not necessarily reflect those of EDHEC Business
School, Caceis, Euronext or Sungard.
Transaction Cost Analysis A-Z — November 2008
Foreword
After nearly ten years of debate, the final This A-Z is the first step of a number of
implementation of MiFID is radically research initiatives that will make possible
transforming the European capital markets a better understanding of execution risk
landscape. and performance and ultimately provide
tools and technologies that lead to more
New entrants such as Chi-X, Turquoise or efficient trading systems.
Equiduct, new operating models developed
by major brokerage firms or former central Far from being restricted to equity markets
exchanges, along with the development alone, MiFID has so far prompted reaction
of advanced execution technologies such mainly on infrastructure related to trading
as algorithmic trading, form what can be in listed securities; more changes can be
called the MiFID revolution. expected in other markets. As such, our
effort will continue both in the fixed
Core to the change is the obligation of income and the listed derivatives space in
best execution, which is one of the pillars the very near future.
the regulator has imposed for a more
competitive environment. But at the time Finally, I would like to take the opportunity
the European directives were drafted, to thank the partners that have made
there was no consensus on what possible the creation of the ‘MiFID and
constituted best execution; indeed, there Best Execution' research chair hosted by
is still no consensus. the EDHEC Risk and Asset Management
research Centre. CACEIS, NYSE Euronext
Transaction cost analysis (TCA) lies at the and SunGard have committed a significant
very heart of the best execution obligation amount of time as well as financial and
and it is expected to become a tool that no technical support to the team, allowing
intermediary and market participant can us to offer material that, we hope, will be
ignore. The literature on TCA is abundant useful to investment firms involved in the
but it remains difficult to find an overview execution process.
of TCA techniques that allows investment
firms to develop a view on which of the
many approaches could or should be
taken.
Noël Amenc
The objective of this report is to provide a Professor of Finance
Director of the EDHEC Risk and Asset Management
comprehensive view of what TCA is, shed Research Centre
light on the main underlying concepts and
document the tools and techniques that
have been developed in the academic and
professional worlds.
An EDHEC Risk and Asset Management Research Centre Publication 3

About the authors
Catherine D'Hondt is an Associate Professor of Finance at the Louvain School of

Management and the Catholic University of Mons in Belgium. Dr D’Hondt teaches
various graduate courses on financial markets and her primary research area is market
microstructure, with a specific focus on traders’ behaviour and order submission
strategies. Within the EDHEC Risk and Asset Management Research Centre, Catherine
D’Hondt contributes to the Best Execution and Operational Performance programme.
She conducts work on the long-term implications of MiFID for European capital markets
and develops expertise in transaction cost analysis and trading performance measurement.
Her doctoral dissertation received the Euronext – French Finance Association award in
2003 and her research in the area of market microstructure and transaction cost analysis
has been presented at major academic and practitioner conferences and published in
refereed journals such as the Review of Finance, the Journal of Asset Management and
Finance Letters. With Jean-René Giraud, she has co-authored the reference text on
MiFID: MiFID: Convergence towards a Unified European Capital Markets Industry (Risk
Books, 2006) and an influential position paper: MiFID: the (in)famous European Directive?
(EDHEC, 2007). Catherine D’Hondt holds a Doctorate in Management Sciences from the
Catholic University of Mons and the University of Perpignan.
Jean-René Giraud is the Director of Development of the EDHEC Risk and Asset
Management Research Centre and the head of the “MiFID and Best Execution”
Research Chair supported by Caceis, NYSE Euronext, and Sungard. Jean-René Giraud
is in charge of developing the relationship between EDHEC and key industry players.
Before joining EDHEC, he was Managing Principal at Capco (London) with responsibility
for buy-side advisory services. Previously, Jean-René was a Senior Project Manager
with Barclays Capital designing and implementing transaction systems, and prior
to that Head of Consulting at SIP Software. As a Research Associate with EDHEC,
Jean-René Giraud heads the Best Execution and Operational Performance programme
and specialises on hedge fund operational risks and best execution. His work has
appeared in refereed journals such as the Journal of Alternative Investments and the
Journal of Asset Management. He has authored a large number of articles in industry
publications, contributed to Operational Risk, Practical Approaches to Implementation
and co-authored the noted MiFID: Convergence towards a Unified European Capital
Markets Industry (Risk Books, 2006). Jean-René Giraud is a frequent speaker at major
industry conferences on various topics related to operations and risk issues. He has a
Master’s in Information Technology from the Polytechnic School of Engineering at the
University of Nice-Sophia-Antipolis.
4 An EDHEC Risk and Asset Management Research Centre Publication

Introduction
A n E D H E C R i s k a n d A s s e t M a n a g e m e n t R e s e a rc h C e n tre Pub l i ca ti on 5
Introduction
From retail to more professional investors investor protection in a marketplace that

and practitioners, all are concerned with is completely open to competition.
transaction costs, as it is an established fact
that lower transaction costs automatically Best execution has consequently become
enable higher returns. To provide their a very fashionable concept. Nevertheless,
clients with competitive portfolio returns, because the regulator has brought neither a
investment firms need to be proactive and clear definition nor a measurable objective
put in place the appropriate means of to make up for the current absence of
effective transaction cost management. consensus, this concept is not always well
However, when implementing their client understood and does not mean the same
1 - MiFID is the second step decisions, investment managers often cope thing to everyone. We can easily find
in the harmonisation of the
European capital markets
with issues regarding how transaction evidence of this phenomenon by simply
industry and aims to adapt costs can be properly identified, measured, reviewing the press and the multiple
the first Investment Services
Directive (ISD 1, issued in forecasted as well as how the quality of occasions where liquidity providers, MTFs,2
1993) to the realities of the
current market structure. Part
execution should be evaluated. Most of the platforms, and technology vendors claim to
of the European Financial time, these fundamental questions remain provide “best execution”, even though there
Services Plan (FSAP), the
“MiFID” (Directive 2004/39/ open because relatively little information is as yet no consensus on what exactly “best
EC, formerly known as
Investment Services Directive
is directly available. In some cases, facing execution” entails or, more importantly,
II) was ratified by the the multiple commercial tools that are while some of those platforms are not yet
European Parliament on April
21st, 2004. The implementing offered, investment managers choose and operational. This ongoing and widespread
Directive and Regulation
were approved by the
apply the most popular indicators in the confusion around best execution—and, to
Parliament over the summer industry, without really knowing if they a larger extent, around TCA as a whole—is
of 2006 and provide detailed
implementation guidelines provide useful and meaningful answers to likely to have serious side effects such as
applicable to all member
states. MiFID came into force
their initial concerns. Some even admit that increased moral hazard, greater adverse
in November 2007. they are sometimes confused, especially selection and a false sense of security given
In a nutshell, MiFID sweeps
away the very concept when they are offered tools based on to end clients. These consequences tend to
of central exchange
and obligation of order
sophisticated models resembling “black emerge when a piece of regulation attempts
concentration currently boxes”. to legislate on elements that may not be
existing in several European
countries, and recognises factually demonstrated.
the need to include
all participants in the
Transaction costs are becoming both
execution cycle and all a topical and relevant issue in Europe To clear up the above-mentioned confusion,
financial instruments under
a consistent regulatory with MiFID—the Markets in Financial we cover in the present work a broad
framework. At the same time,
MiFID offsets the opening of
Instruments Directive1—now in place. In range of material related to TCA and best
the execution landscape to the near future, the role of transaction execution. As understanding transaction
full competition with a set
of obligations whose goal is cost analysis (TCA) is expected to grow costs is crucial to properly assessing
to increase transparency and
investor protection in order
substantially as a result of the best the quality of implementation decisions
to maintain the efficient execution obligation. According to this and complying with the best execution
and fair price discovery
mechanisms as well as the new obligation, investment firms must obligation in the post-MiFID environment,
overall integrity of European
markets in the face of
“take all reasonable steps to obtain the our objectives are the following:
inevitable fragmentation. best possible result when executing orders • provide a state of the art of TCA
2 - MTF: multilateral trading
facility for their clients”. Although so far a mix of fundamentals;
both principle- and rule-based regulation, • do a critical review of existing post-trade
this new obligation is a key element for TCA techniques;

Introduction
• define a new and complete approach spread, market impact and opportunity
wherein checking whether best execution costs. As variable components, they offer
is achieved lies in the entire transaction the opportunity to improve the quality
cost management process. of execution. This section reviews all the
costs of both categories in detail and
To cover all the aspects related to these gives insights into why they arise when
objectives, we have structured the present investment decisions are made. In addition,
work in six sections. we provide empirical evidence of the
differences in total transaction costs as
Section I provides insights into how well as their composition across regions
transaction costs arise during the investment and trading venues, in Europe in particular.
cycle as well as an introduction to TCA’s
main goals. After a brief description of the Section III is devoted to the two fundamental
investment phases, we first highlight the approaches to measuring transaction costs.
importance of being proactive in managing Benchmark comparison measures the
transaction costs to obtain competitive cost of trading by the signed difference
returns. We then present TCA as a very between the trade price and a specified
valuable decision-making tool since it benchmark price. This method provides
allows monitoring both the transaction cost indicators that depend on whether
costs of different implementation they are built on pre-trade, intraday or
strategies and the trading performance post-trade benchmarks. Implementation
of different intermediaries. At this stage, shortfall defines the cost of trading as the
we briefly introduce the three TCA difference between the actual portfolio
missions—transaction cost measurement, return and its paper return benchmark.
transaction cost estimation and trading This approach has become the standard
performance assessment. We conclude for transaction cost measurement and we
by showing how TCA can effectively document how the primary framework
contribute to enhancing total portfolio may be expanded to fine-tune cost
performance. identification. Before reviewing the most
popular indicators used in the industry, we
Section II provides a thorough investigation emphasise several issues to consider when
into transaction cost components and measuring costs. We conclude with new
drivers. Transaction costs include all costs practical questions and complications that
associated with trading, which are usually emerge with MiFID.
divided into explicit and implicit costs.
Explicit costs include brokerage commissions, Section IV deals with pre-trade analysis,
market fees, clearing and settlement whose primary purpose is to forecast both
costs, and taxes/stamp duties. These costs the transaction costs and the risks associated
do not rely on the trading strategy and with various strategies for a given trade
can be quite easily determined before or a specific trade list not yet executed.
the execution of the trade. By contrast, In this section we refer to the approach
implicit costs represent the invisible developed by Kissell and Glantz (2003) and
part of transaction costs and consist of show that it makes it possible to determine

Introduction
the cost profile of any trading strategy Section VI gives us the opportunity to
through simultaneous estimation of cost introduce a framework developed by
and risk, develop optimisation techniques the EDHEC Risk and Asset Management
to derive an efficient trading frontier and Research Centre team that makes it
devise the most appropriate strategy to possible to address the crucial question
meet the investor’s goal or comply with of best execution evaluation for traders
his preferences. and investment managers. The model is
explained in full and can be easily deployed
We would like to take the opportunity and customised to financial institution’s
to express our sincere thanks to Robert specific needs.
Kissell and Morton Glantz, who have made,
amongst other things, a very significant
contribution to a better understanding and
approach to modeling pre-trade transaction
cost and risk estimates. Our aim here is
only to provide a synthetic overview of the
question; we have therefore summarised
their material, but a good understanding of
the approach cannot be obtained without
reading in detail the full methodology
available in Kissel and Glantz's 2003 Optimal
Trading Strategies.
Section V is dedicated to trading

performance evaluation. After having
exposed how trading performance
measurement differs from transaction
cost measurement, we come back to the
benchmark comparison approach, the
most common practice in the industry,
and we review its major shortcomings
when it comes to measuring quality of
execution. We then focus on the concept
of best execution in general and see that
although it is very fashionable, it is often
misunderstood and does not mean the same
thing to everyone. We conclude with the
MiFID best execution obligation and show
that the regulator has provided neither a
clear definition nor a measurable objective
to make up for the current absence of
consensus in the industry.

I. Transaction Cost Analysis
as Part of the
Investment Process


as Part of the Investment Process
Four distinctive phases are usually must manage transaction costs

identified in the investment decision proactively, because lower transaction
cycle. They may be summarised as costs mean higher portfolio returns.
follows. The asset allocation phase Ineffective transaction cost management
refers to the distribution of funds across may be very damaging in today's
investment classes (equities, bonds, cash, competitive environment, in which the
commodities, derivatives, hedge funds, difference between success and failure
real estate, etc.) with the objective of may be no more than a few basis points,
diversifying risk and targeting a specified or in which tiny spreads offer arbitrage
return. The portfolio construction phase strategies that must be protected from
is the phase wherein decisions about the transaction costs. The aim of transaction
exact instruments to buy or sell in each cost analysis (TCA) is to provide a scorecard
investment class are made. The execution that helps investment managers assess
services phase is the phase during which understand how well their decisions
investment decisions are acted on. It have been acted on and how they can
involves decisions regarding the trading improve their overall performance. On the
strategy: where, when and how to buy/ one hand, as different trading strategies
sell. Finally, the portfolio attribution phase correspond to different risk-cost trade-
involves assessing portfolio performance offs, investment managers need to know
and analysing reasons for missing the the real cost of implementing a given
targeted return. trading strategy. On the other hand,
since bad execution can impact the total
Most financial research is devoted to asset performance of even the best investment
allocation, portfolio construction and decision, investment managers must be
performance attribution. We do not find able to assess the trading performance
the same abundance of literature on the of their intermediaries (brokers, traders or
implementation of investment decisions. even algorithms). Roughly, then, TCA is a
For the total performance of a portfolio, tool for monitoring both the transaction
however, the quality of the implementation costs of trading strategies and the trading
is as important as the decision itself. performance of intermediaries.
The reason is that the implementation
of investment decisions is not free and Going into greater detail, we can attribute
that the associated costs usually reduce to TCA three distinct and specific tasks:
portfolio returns with limited potential • Transaction cost measurement
to generate upside potential. These costs • Transaction cost estimation
can turn high-quality investments into • Trading performance assessment
moderately profitable investments or
low-quality investments into unprofitable It is first very important to understand the
investments. These costs are usually difference between cost measurement and
referred to as transaction costs. cost estimation, because, as we will see,
each requires a specific methodology. In
To provide investors with competitive a nutshell, transaction cost measurement
portfolio returns, investment managers involves assessing the cost of completed


trades and hence occurs ex post. By measure the transaction costs of past
contrast, transaction cost estimation trades, estimate the transaction costs
attempts to forecast the cost of proposed of future trades and allow comparisons
trades. Basically, transaction cost measures across trading strategies for a specific
and estimates differ in two main ways. trade list, as well as measure the trading
First, transaction cost measures result in performance of the intermediaries. All the
single point values expressed in either information collected in this phase must
monetary units or basis points per share, then be used in the other phases to avoid
while transaction cost estimates are misallocation of funds, inefficient portfolio
expressed as a probabilistic distribution mixes and ineffective intermediaries.
defining both an expected cost and a risk TCA thus contributes to enhanced total
parameter. Next, estimation is essentially performance over the entire investment
based on cost components and drivers, decision cycle.
while identifying and measuring each cost
component is not so obvious ex post. To allow the reader to view this report in
light of hard figures, Karceski, Livingston
When dealing with TCA, it is also of great and O'Neal (2004) established that
importance to understand how trading actively managed equity portfolios bear a
performance evaluation differs from total of 0.98% transaction costs per year.
transaction cost measurement, even To a significant degree, those costs are
if both are done ex post. Performance implicit and are therefore not reported in
evaluation attempts to assess how well total expense ratios.
intermediaries perform when executing
trades. The ultimate aim is to determine
the most effective intermediaries by
market, instrument and trading strategy.
With this information at hand, investment
managers can considerably reduce the time
required to select the best intermediary
for the execution of a specific trade. While
transaction cost measurement focuses on
determining how large transaction costs
are and where they arise, the analysis of
trading performance involves determining
whether the costs are justified or result
from poor implementation decisions and
could have been avoided.
By nature, all the tasks assigned to TCA must

be performed within the execution services
phase of the investment decision process.
It is there that quantitative procedures
and approaches can be developed to



II. Transaction Cost
Components and Drivers
II. Transaction Cost Components

and Drivers
Understanding exactly what transaction Brokerage commissions are paid to

costs are and why they arise is the first and intermediaries for executing trades. They
fundamental step when dealing with TCA. can be expressed on a per share basis or
Transaction costs have nine components, based on a total transaction value, most
usually categorised as implicit or explicit, of the time in basis points and subject to
as shown in figure 1. In this section, volume discount. Although they differ
we describe the components of both from one intermediary to another, they
categories in detail and give insights into are a fixed and visible transaction cost
why these costs arise when implementing component.
investment decisions.
Market fees are paid to trading venues
Figure 1: Typology of transaction costs
for executing trades on their platforms.
Brokerage Commissions
They are usually bundled into brokerage
Market Fees
commissions for investors. These fees
Clearing and Settlement Costs
vary. On average, higher volume markets
Taxes/Stamp Duties
have the lowest costs. In recent years,
Bid-Ask Spread
Market Impact
competitive pressure has led to a significant
Operational Opportunity Costs
reduction in these explicit costs. Like
Market Timing Opportunity Costs brokerage commissions, market fees are
_ _
Missed Trade Opportunity Costs considered a fixed and visible transaction
cost component.
Explicit Costs Implicit Costs
Clearing and settlement costs are related
Those costs can be considered direct when to the process whereby the ownership
related to individual orders/transactions, of securities is transferred finally and
or indirect when accounted for globally irrevocably from one investor to another.
as part of the provision of the transaction When the trading venue owns the clearing
services. To provide a full picture, we and settlement system, these costs,
also introduce the concept of direct and which are a fixed and visible transaction
indirect explicit transaction costs. cost component, are usually included in
market fees. Like the latter, clearing and
settlement costs differ from one trading
1. Direct Explicit Transaction Costs venue to another. This is illustrated in table
Brokerage commissions, market fees, 1, which exhibits statistics about the cost
clearing and settlement costs, and taxes/ per execution (in €) in the main European
stamp duties are explicit costs. They are stock exchanges.
said to be explicit because they do not
Table 1: Cost per execution in Europe (€)
depend on the trade price and are usually
Cost per Mean Maximum Minimum
documented separately from it. As these Execution
costs do not rely on the trading strategy, Exchange 2.85 4.94 1.57
they can be known in advance, before the Clearing 1.24 2.57 0.38
execution of the trade. Settlement 1.22 2.18 0.52
Total cost 5.31 7.00 4.01
Source: Various public sources (2007)


and Drivers
Taxes/stamp duties are visible and variable Capital requirements related to these costs
transaction cost components. They are are currently assessed on a global basis;
visible because tax rates or specific stamp allocating these costs to specific trading
duties are known in advance but variable activities remains very ambitious even
because they often vary by type of return though these costs are often estimated
or trade. to support strategic organisational
decisions.
In theory, explicit costs could be
determined before the execution of the 2.a Indirect costs related
trade. In practice, their measurement to operational risks
is not so obvious because brokerage Basel II defines operational risk as the risk
commissions are often paid for bundled of loss resulting from inadequate or failed
services, not only for order execution. internal processes, people and systems, or
Research, analytics and trading technology from external events. From the perspective
are often bundled services provided of managing transactions in financial
by intermediaries. A trend towards instruments, operational risks relate to
unbundling is being observed in Europe the risk of loss resulting from inadequate
and the US. New unbundling of or failed internal processes, people and
commission regulations, which separates systems in the handling of the transaction
the payment for deal execution and the cycle.
payment for broker research, will surely
facilitate the measurement of explicit The definition provided in the Basel II
costs. capital requirement framework encom-
passes situations such as:
• Internal fraud—misappropriation of
2. Indirect Explicit Transaction Costs assets, tax evasion, intentional mismarking
In addition to direct costs, explicit costs of positions
also include a number of indirect costs • External fraud—theft of information,
having to do with the processing supporting hacking damage, third-party theft and
the execution and the counterparties forgery
involved. Even though the determination • Business disruption and systems
of these costs in today’s environment is failures—utility disruptions, software
very difficult (and perhaps impossible on failures, hardware failures
a trade-by-trade basis), their importance • Execution, delivery, and process
should not be underestimated when m a n a g e m e n t — d a t a - e n t r y e r ro rs,
venues or particular types of transaction accounting errors, failed mandatory
are opted for. reporting, negligent loss of client assets.
These indirect explicit transaction costs The complex nature of the entire trade
encompass: processing cycle, from execution to
• capital costs related to operational risks settlement, therefore has a direct impact
• capital costs related to counterparty on the operational risks borne by the
(and credit) risks financial institution.


and Drivers
Basel II and various supervisory bodies and qualitative criteria. Use of the AMA is
have prescribed a number of soundness subject to supervisory approval and makes
standards for operational risk management it possible to reduce the minimum capital
for banks and similar financial institutions. requirements significantly.
To complement these standards, Basel
II has provided guidance on three broad Under this last regime, widely opted for by
methods of capital calculation for major institutions, the actual operational
operational risk: risks measured by means of quantitative
• Basic indicator approach and qualitative analysis have a direct
• Standardised approach impact on capital requirements.
• Advanced measurement approach (AMA)
The following factors are likely to make a
Banks using the basic indicator approach direct impact on the internal measure of
must hold capital for operational risk operational risks:
equal to the average over the previous • Nature of the execution process
three years of a fixed percentage (denoted • Extent of trade processing automation/
alpha) of positive annual gross income. manual interaction
• Nature of interfaces with third parties
In the standardised approach, bank (paper/electronic/fax)
activities are divided into eight business • Extent of operations outsourcing
lines: corporate finance, trading and
sales, retail banking, commercial banking, The nature of the execution process (direct
payment and settlement, agency services, trading, algorithmic trading, phone/
asset management, and retail brokerage. electronic interface) provides an important
Within each business line, gross income is input to the extent of operational risks. As
a broad indicator that serves as a proxy a result, the choice of execution venue
for the scale of business operations and and the maturity of the interface between
thus the likely scale of operational risk the financial institution and the liquidity
exposure within each of these business pool are key factors behind operational
lines. The capital charge for each business risks and hence capital requirements.
line is calculated by multiplying gross
income by a factor (denoted beta) assigned The extent to which the processing of
to that business line (18% for sales and trades is automated and the nature
trading, 12% for retail banking, 18% for of interfaces with third parties are
payments and settlement, 12% for asset determined by the maturity of the markets
management and retail brokerage). on which transactions occur. For example,
there are significant differences between
In the advanced measurement approach, emerging/developing and historic markets
the regulatory capital requirement and exchanges. Similarly, transactions
will equal the risk measure generated executed on exchanges and those that
by the bank’s internal operational risk take place over-the-counter (OTC) are not
measurement system using quantitative equally automated. In the latter situation,


and Drivers
financial institutions may not benefit as well as by back-office intermediaries

from the advanced and proven market such as custodian banks.
infrastructure that makes it possible to
streamline and automate transactions. So it would be a mistake to believe
that the choice of execution venue or
As a result, in the post-MiFID environment, changes in execution methods have no
financial institutions must address the implications on the back-end of the
consequences of choosing alternative processing cycle. Custodian banks have
trading venues not only in relation to the understood the importance of post-
execution process but, more importantly, trade processing in the search for best
in light of the impact on the entire trade execution perfectly well and, through
processing cycle as well. The choice better integration of alternative trading
of alternative platforms implying OTC cycles in their operational processes with
transactions (e.g., dark liquidity pools) the aim of both reducing operational
might present significant advantages constraints and risks and enhancing overall
in terms of execution efficiency and efficiency, are well positioned to support
quality but is likely to impose operational the new requirements created by the
constraints that need to be addressed fragmentation of liquidity venues.
to avoid possibly significant capital
implications. 2.b Indirect costs related to
counterparty (and credit) risks
Likewise, the extent to which operations Counterparty risk exists when a financial
are outsourced is likely to modify the institution transacts with another firm
firm’s overall operational risk profile. Basel that may default on its obligation to settle
II does not allow firms to eliminate capital a transaction or deliver securities related
requirements by transferring operations to a transaction. This situation can arise
to third-party providers, but those third when the counterparty is in financial
parties may boast more modern and distress (we are then confronted with
scalable operations tailored to processing credit risk) or when the counterparty faces
transactions in the most efficient manner, a pure operational or cash management/
implicitly reducing operational risks. stock management glitch.
When a firm outsources its transaction
processing cycle to a third party, it is As a consequence, counterparty risk can
usually to a partner that specialises in occur when a financial institution lends
providing this service; the firm can thus cash or securities, but it could also be
benefit from economies of scale and gain influenced by occasional problems in the
access to state-of-the-art infrastructure. settlement cycle. Failure to collect payment
or deliver securities can have a damaging
The quest for operations efficiency and impact on a financial institution, as it
the associated reduction in operational- may cause the firm to default on other
risk costs account for much of the transactions and create a cascade of
recent success of outsourcing offerings events that can, if not dealt with in due
developed by pure back-office providers time, lead to default.


and Drivers
We will not deal here with the question have huge consequences on the economic
of credit risk, i.e., the assessment of the profitability of trading activity.
ability of the firm’s counterparties to meet
their obligation to repay/deliver, as this So, to benefit from the economies of
element is part of normal banking activity. scale and state-of-the-art infrastructure
The risk incurred by firms as a result of the offered by firms specialising in post-trade
settlement process of transactions should processing, financial institutions have
nonetheless be taken extremely seriously begun outsourcing part of their back-
in light of the changes resulting from the office responsibilities to third parties.
implementation of MiFID.
Three broad benefits are to be expected:
The direct consequence of the • A decrease in the investments required to
development of MTFs and systematic cover a growing number of liquidity pools
internalisation is the likely significant with regards to post-trade processing.
increase in OTC transactions that involve a As specialist firms build infrastructure
direct settlement to the counterparty of the centrally for a large number of clients,
trade rather than the clearing of transactions they indirectly let their clients benefit
through a clearing house and a central from economies of scale that allow them
counterparty. In most established markets, to cover a larger number of pools than
central counterparties and clearing individual financial institutions would
houses have made it possible to reduce otherwise be able to.
the complexity of the settlement process • A significant increase in processing
significantly (thanks to the netting of efficiency, leading to better control of
transactions) as well as the counterparty explicit direct costs (settlement costs,
risk (thanks to the use of a single central costs of handling the clearing process,
counterparty managing the individual costs of reporting, streamlined processes
counterparty risks through appropriate for managing mistakes)
processes such as delivery versus • A significant decrease in operational
payment). risk and application of the most advanced
management processes, making it possible
Once again, the choice of venue and the to reduce counterparty risk (or at the very
selection of a mode of execution will least improve the management of breaks
have a direct impact on the final level of and intraday defaults) and the capital set
risks and costs incurred by the financial aside as part of the Basel II framework.
institution.
But third-party back-office providers and
2.d Increased importance of custodians have also realised that their
back-office providers status as recipients of transaction data
Allocating the exact indirect explicit costs of puts them in an ideal position to develop
operational and counterparty risks to single and propose a set of value-add services
transactions is obviously too ambitious, directly related to transaction cost
but these costs do represent significant analysis. Much as with the development of
expenditures or capital requirements that risk and performance allocation services


and Drivers
delivered by back-office providers on the analysis of transaction costs is—because it

basis of portfolio and transaction data is a prerequisite to the smooth running of
collected to manage positions on behalf a fragmented execution market—likely to
of clients, these intermediaries can now become one of the essential value-added
offer transaction cost analysis services services provided by custodian banks to
based on transaction data transmitted for financial institutions managing third-
settlement purposes. party assets.
The firms that attempt to provide services

for the monitoring of best execution will 3. Implicit Transaction Costs
face two main challenges: Transaction costs are more than just
• Access to data and consolidation of brokerage commissions, market fees and
external market tick and transaction data taxes. Implicit costs are the transaction
• Enrichment of existing front-to-back costs that, invisible, cannot be known
data flows to include front-office trading in advance because they are included
information (timestamps) in the trade price. They depend mainly
on the trade characteristics relative to
The second challenge is under the control the prevailing market conditions. In
of the third-party provider (though we other words, they are strongly related
do not underestimate the operational to the trading strategy and, as variable
and technical complexity of handling costs, provide opportunities to improve
trade data in addition to data usually the quality of execution. These implicit
transmitted for settlement and reporting costs can be broken down into their
purpose only), but creating an adequate components: spread, market impact and
repository of market and transaction opportunity costs.
data that will make it possible to perform
such advanced forms of transaction cost (1) Spread
analysis as peer group analysis will require This component is compensation for the
heavy investment and proper coordination costs incurred by the liquidity provider.
throughout the industry, as MiFID failed When taking liquidity (by buying at the
to deliver an industry model that would best ask or selling at the best bid), we
simplify the collection and re-distribution pay the spread. In the microstructure
of public information on markets and literature, three kinds of cost are usually
transactions. associated with the bid-ask spread. The
order processing cost is compensation
All the same, it is clear that back- for supplying an immediacy service
office intermediaries are the only firms to the market (Demsetz 1968). The
currently involved in the transaction cycle ability to trade immediately rather
in a position to offer an independent than to have to wait for the opposite
assessment of transaction costs on behalf trade provides certainty for market
of the final investor. Like independent asset participants. Liquidity is thus provided at
valuation, likely to be in demand following that cost.
the recent banking crisis, independent


and Drivers
The inventory control cost is compensation quantity available at the best opposite
for the risk of bearing unwanted quote. The spread cost is therefore
inventories (Ho and Stoll 1981). particularly sensitive to the timing of
Accommodating other market execution.
participants’ trades makes liquidity
providers deviate from their optimal Figure 2 illustrates the variation of the
inventory based on their own risk-return volume-weighted bid-ask spread across
preference. To restore their optimal several major European stock exchanges.
position, they adjust their bid-and-ask For a given trading venue, spreads
prices to attract and/or avoid some trades. fluctuate across stocks. In general, spreads
The adverse selection cost is compensation are negatively associated with market
for the risk of trading with informed traders capitalisation and liquidity and positively
(Copeland and Galai 1983). Informed associated with volatility and information
traders have a certain amount of private asymmetry.
information that allows them to know or
better estimate the true value of a security. (2) Market impact
As liquidity providers lose when they trade Market impact is the price to pay for
with informed traders, they widen their consuming the liquidity available on
bid-ask spread for all market participants the market beyond the best quote: to
to cover their potential losses. complete their “large” orders, buyers must
pay premium prices and sellers must offer
The spread represents the implicit cost of discounts. In other words, market impact
a round-trip for a small trade and, as such, is the price shift that is due to the trade
may be measured from market data by the size. Its main determinants are the trade
simple difference between the best ask size and the market liquidity available
and best bid quotes. The ease of obtaining at the time of the trade. Accordingly,
measures for spread makes some people market impact can be viewed as a positive
consider it a visible implicit cost. In any function of the trade size and a negative
case, this cost component is variable function of the liquidity available. For
since spreads vary over time according a given level of liquidity, market impact
to trading conditions. In limit order book increases with the trade size. For a given
systems, spreads mechanically widen after trade size, market impact increases with
a large trade consuming more than the the lack of liquidity.
Figure 2: Weighted average bid-ask spreads in %

Euronext Portugal
Sweden
United Kingdom
Italy
Euronext Brussels
Switzerland
Spain
Germany
Euronext Paris
Euronext (Total)
Euronext Amsterdam
US Nasdaq
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.40
Source: Various public sources, EDHEC-Risk Advisory


and Drivers
Trades affect market prices for two main Figure 4 focuses on the relationship
reasons. First, if the trade is large relative between market impact and liquidity. Here,
to the available liquidity, the trade causes market liquidity is measured by turnover,
a market supply-demand imbalance and defined as the ratio of the average daily
mechanically shifts the market price volume to the total number of shares
towards less favourable prices because outstanding. As we can see, the lack of
it needs to attract additional liquidity liquidity becomes expensive when turnover
(consumption of several quotes in the falls below 0.026% in the sample.
order book). This mechanical market
impact lasts until liquidity is replenished.
Figure 3: Average costs by trade relative size
Second, trades can affect prices when
they are perceived as motivated by new Market Impact
0.25
information. When the trade brings new
information to the marketplace, a market
0.20
correction occurs and the related market
impact is permanent. 0.15
Market impact is one of the most costly 0.10

implicit components because it generates
mainly adverse price movements and 0.05
therefore reduction in portfolio returns.
It is an invisible and variable transaction 0.0
1 2 3 4 5 6
cost: invisible because it cannot be easily
Relative Size
determined before the execution of the
trade and variable because it depends * The classes are defined as follows:
greatly on the trade size, the available Class 1: Relative Size < 0.05%;
Class 2: 0.05% =< Relative Size <0.2%;
liquidity and the specified trading Class 3: 0.2% =< Relative Size <0.4%;
strategy. Class 4: 0.4% =< Relative Size <1%;
Class 5: 1% =< Relative Size <5%;
Class 6: Relative Size >= 5%
Figures 3 and 4 illustrate the relationship
Source: Boussema et al (2002)
between market impact and trade size
or liquidity. They are built on statistics
based on the Sinopia Asset Management
databases spanning sixteen developed
countries and refer to trades executed
during the period from 1999 to 2000.
Figure 3 shows the relationship between
market impact and the trade relative size.
The latter is defined as the size of trade
(expressed in number of shares) divided
by the daily average volume for the stock.
Market impact increases greatly from class
four—trade size approaching or in excess
of 1% of average daily volume—onwards.

and Drivers
Figure 4: Average costs by turnover At each stage between the investment

Market Impact decision and the execution of the trade,
0.15
there are potential delays and thus
0.10 price movements that could positively
or negatively affect portfolio returns.
0.05 Opportunity costs are consequently
invisible and variable implicit costs that
0.0 depend greatly on the speed of execution.
1 2 3 4 5 6
Turnover
As the source of the time between
* Share turnover = average volume traded during the
previous month/Number of shares outstanding. trade decision and execution can vary,
The classes of liqudity are defined as follows: opportunity costs can be broken down
Class 1: Turnover < 0.019%; into three components:
Class 2: 0.019% =< Turnover <0.026%;
Class 3: 0.026% =< Turnover <0.034%;
• operational costs
Class 4: 0.034% =< Turnover <0.046%; • market timing costs
Class 5: Turnover >= 0.046
• missed trade costs
Source: Boussema et al. (2002)
Operational opportunity costs arise when
the time required to trade is operational
(3) Opportunity costs
and unintended (transmission delays
The decision to trade and the actual trade
between buy and sell sides, for example).
do not usually take place at the same time.
When the delay results from market
Market prices can move for or against the
timing under the control of the broker
proposed trade. The costs related to price
(for example, the broker splits the order
fluctuation during the time required to act
into small lots over a period to minimise
on an investment decision are opportunity
market impact), we refer to market
costs. They arise when prices move between
timing opportunity costs. Operational
the time the trading decision is made and
and market timing costs are sometimes
the time the order is executed.
both considered delay costs. We prefer
the distinction as it enables identification
In general, two elements are responsible
of the party responsible for the costs
for changes in opportunity costs during
incurred: the investor or his intermediary.
the time required to complete the trade:
Finally, missed trade opportunity costs
price appreciation and timing risk. Price
are incurred when investors fail to fill
appreciation represents the natural price
their orders. Some trades may not be fully
trend of the security. Timing risk has to do
completed either because price movements
with the uncertainty associated with the
have led to the cancellation of the initial
price movement. In other words, the first
trading decision or because there is no
element has to do with expected price
more security available (lack of liquidity).
movements and the second has to do with
If a predetermined trading strategy is not
unexpected price movements.
completed, the resulting opportunity cost
can be high. Failing to trade can be costly
for the investor, who will have missed the


and Drivers
opportunity to make an investment in the This fall is reported by Domowitz et al.

security requested. (2001) in an analysis of equity trading
costs in a sample of forty-two countries.
The contribution of opportunity costs to Boussema et al. (2001) also document
total implicit costs is not independent of this fall in transaction costs. For a sample
market impact. Attempting to reduce one of trades included in the Sinopia Asset
can lead to the increase of the other. For Management database, the authors find
example, splitting large orders over time that total transaction costs amounted to
to reduce market impact can lead to larger about 0.38% from 1996 to 1999 while
opportunity costs and vice versa. This issue they fell to 0.18% from 1999 to 2000.
is summarised in figure 5. Minimising
market impact and reducing opportunity We observe a similar finding in Munck
costs form a set of conflicting objectives (2005), a more recent study devoted
usually referred to as “the trader’s dilemma”. to transaction costs at the larger stock
Balancing these conflicting costs is a real exchanges in Europe and in the north.
challenge. Munck shows that total transaction costs
have dropped in recent years. He attributes
3 - For a discussion about Figure 5: Relationship between market impact and opportunity
the long-term impact of costs
this drop to the significant drop in explicit
MiFID on the European costs. Figure 6 shows that the explicit
trading landscape, see
D’Hondt and Giraud (2007). costs of trading on the OMX exchanges,
Euronext and Deutsche Börse have fallen
over the past eight years and are fairly
clustered. The higher explicit costs on the
London Stock Exchange are the result of
a special stamp duty of fifty basis points
on all buy trades. Figure 7 shows the
pattern of implicit costs over the same
Source: Giraud (2004)
period. Although there are fluctuations,
it is clear that costs at these exchanges
4. Transaction cost magnitude are clustered. In late 2004, implicit costs
Several empirical studies have examined ranged from ten to fifteen basis points.
the relative importance of explicit and
implicit transaction costs. Most have Today, with MiFID just in place, it is expected
focused on the costs of equity trading that execution fees will keep falling in
for institutional investors. We can draw Europe. The impact of the Directive on
several conclusions from these empirical total transaction costs, however, is not so
findings. obvious, as implicit costs depend on the
efficiency of market structures, on which
First, transaction costs have fallen the long-term implications of MiFID are
dramatically in recent years, essentially as yet unknown.3
due to technological innovations and
increased competition among trading
venues and among intermediaries.


and Drivers
Figure 6: Explicit transaction costs variable for both explicit and implicit costs.
He finds that both explicit and implicit
costs fall as system activity increases.
Figure 8 illustrates this relationship for
total transaction costs.
Figure 8: Total transaction costs vs. market system turnover
Source: Munck (2005)
Figure 7: Implicit transaction costs
Third, the composition of transaction costs

can vary across trading venues. Figure
9, excerpted from Munck (2005), breaks
down total transaction costs in 2004 for
Second, transaction costs vary from
the European exchanges in his sample.
region to region and even within regions.
Except on the OMX Helsinki market,
Although they are falling, they remain
explicit costs account for the lion’s share
economically significant, especially
of transaction costs, with an average of
in emerging markets. According to
approximately sixty basis points. This cost
Domowitz et al. (2001), over the third
composition is unlike that in US trading
quarter of 2000, total transaction costs
venues, where implicit costs tend to be
ranged from twenty-two basis points in
higher.
the Netherlands to 184 basis points in
Venezuela, with a cross-country mean of Figure 9: Composition of transaction costs
about sixty basis points. Boussema et al. 100
(2001) also compare transaction costs in

Share of total trading costs
80
developed and emerging markets. In their
sample of trades, explicit costs (including 60
only brokerage commissions) average

40
0.15% in developed markets and 0.61%
in emerging markets. This phenomenon is 20
similar for implicit costs (market impact

0
and delay costs): they are about 0.23%
n (I
)
gen t (II
)
olm Osl
o fur
t
do ha ex ckh nk elsi
nki
Lon Copen Euron X Sto Fra XH
in developed markets and approximately X OM OM
OM
0.58% in emerging markets. Checking
Implicit trading costs
for a possible correlation between costs Explicit trading costs
and the use of trading systems, Munck Average
(2005) identifies market system turnover
as a statistically significant explanatory


and Drivers
5. Transaction Costs and In principal trading, the broker provides

Execution Methods the investor with guaranteed execution
Transaction costs also depend on execution of the trade list at the market prices at a
methods. There are two kinds of techniques specific point in time (usually the close).
for executing an order: agency trading or Here, the entire market risk is transferred
principal trading. The two methods differ to the broker because he acquires the
in terms of the risk sharing between the investor’s position and all associated risk.
investor and the executing broker. In As a consequence, brokerage fees are
fact, the choice between agency trades higher than for agency trades and depend
and principal trades involves a trade-off on the risk associated with the execution
between a low certain explicit cost and an of the trade list. These specific higher
uncertain implicit cost. commissions are known as the principal
bid premium and serve as an insurance
Agency trading is the most common policy for the broker.
execution method. When sending orders
to the market through an agency, the The two methods and their implications
investor bears all the risks associated with are summarised in table 2.
the trade. The agency assumes no market
risk, so all the implicit costs fall on the
investor. In practice, the investor sends an
order to the broker, specifying the name of
the security and the number of shares to
buy or sell. The agency then executes the
order and the speed of execution depends
mainly on the order size and the available
liquidity. In some cases, the investor can
give the agency price (closing price, daily
VWAP, etc.) or volume (percentage of the
daily volume, for example) constraints.
Average brokerage fees are relatively low.
Table 2: Execution methods and transaction costs

Execution method Characteristics Advantage Disadvantage
Agency trade Release of an order Low fees Market risk
specifying the security (implicit costs)
and the quantity
Execution on the market

with or without a target
(closing price, VWAP, etc.)
Principal trade Determination of a basket: No market risk Higher fees
size, securities, etc.
Commitment of the broker

to trade the basket at a
predetermined price


and Drivers

III. Measuring Transaction Costs
with Post-Trade Analysis

III. Measuring Transaction Costs with

Post-Trade Analysis
As Harris (2003) notes, it is hard to have been observed if the trade had not
manage effectively what you cannot taken place. In this case, the difference
accurately measure. Proper measurement between this price and the trade price
of transaction costs is thus crucial for could be pinned entirely on the trade.
quality decision-making, since it permits
assessments of portfolio profitability and Different transaction cost indicators
final performance. More broadly, accurate are used for different benchmarks. The
transaction cost measures are also useful indicators can be based on pre-trade,
for building and testing transaction cost intraday or post-trade benchmarks.
estimation models. Furthermore, in each of these three
categories, an additional distinction can
In this section, we discuss the two be made between absolute indicators
currently used fundamental approaches and time-related indicators. Absolute
to measuring transaction costs. indicators are easier to compute because
These approaches are the benchmark they do not require consideration of time
comparison and the implementation stamping. In other words, these indicators
shortfall. We then identify common do not take into account the exact time at
pitfalls to look out for when measuring which the trade is executed. By contrast,
costs and offer a critical review of the most time-related indicators are dependent
popular indicators. The section concludes on the exact time at which the trade is
with a discussion of complications brought completed. They rely on a benchmark
about by MiFID. price that is computed around or at the
execution time.
1. Benchmark Comparison (1) Indicators built upon pre-trade

This method involves comparing the benchmarks
monetary value difference between These indicators use benchmark prices
the executed position and the position prevailing on the market before the
evaluated at a benchmark price. For a execution of the trade. The most frequent
given transaction, the monetary value per are based on the following benchmark
share is given by the signed difference prices, depending on whether they
between the average price obtained for the consider the accurate time of execution
trade and the benchmark price. A positive or not.
value indicates an execution that is less Absolute indicators:
favourable than the benchmark price, while • T-1 Close: previous night's closing price
a negative value indicates execution more • T Open: opening price of the day
favourable than the benchmark. Negative
costs, in other words, are savings. The Time-related indicators:
choice of benchmark is primordial here. • Ask: last ask price before execution
For one, the benchmark price should be • Bid: last bid price before execution
easy to obtain or compute. For another, it • Last: last trade price before execution
should be ideal, in the sense that it should • Midpoint bid-ask: last bid-ask average
enable assessment of the price that would before execution


Post-Trade Analysis
(2) Indicators built upon intraday 2. Implementation shortfall

benchmarks This method, widely recognised as the
These indicators are based on benchmark most effective means of measuring
prices that represent a kind of average transaction costs, was initially proposed
market price for the day of the trade by Perold (1988) and its principle is easily
execution. The most frequent are based on understood. It involves assessing the
the following benchmark prices, depending impact of trading on portfolio returns by
on whether they focus on the exact time of computing the difference between the net
execution or not: returns on a paper portfolio and those on
Absolute indicators: a real portfolio.
• Daily VWAP: the volume-weighted
average price of the day (1) Original framework
• Daily LHOC: the average of the lowest, Determining implementation shortfall is a
highest, opening and closing prices of the two-step process. First, at the time of the
day investment decision, the investor needs
• Midpoint H/L: the median of the highest to build and price a paper portfolio. This
and lowest prices of the day portfolio will be an imaginary holding
4 - The VWAP computed over
a certain long period of time
consisting of all the security positions the
(n days) around the trade Time-related indicators: investor decides to take. It is assumed that
execution time also exists.
This multi-day VWAP can be • Available VWAP: the VWAP computed these positions are acquired at the price on
relevant for trades on very from the market opening to the trade the market at the time it was decided to
illiquid securities.
time acquire them. Consequently, the resulting
• Interval VWAP: the VWAP calculated paper portfolio, unlike the corresponding
over a fixed time interval around the trade actual portfolio, does not incur any
execution time4 transaction costs. The difference between
the paper portfolio return and the actual
(3) Indicators built upon post-trade portfolio return is the “implementation
benchmarks shortfall”. Figure 10 gives a summary
This final category contains indicators overview of the method.
referring to benchmark prices that
prevail on the market after the execution
of the trade. These indicators are less
heterogeneous than in the two previous
categories and the most frequently used in
the industry are listed below.
Absolute indicators:
• T Close: closing price of the day
Time-related indicators:
• Next mid bid-ask: next bid-ask average


Post-Trade Analysis
Figure 10: Principle of the implementation shortfall time and the actual trade price, it actually
30% delivers a nearly complete measure of
Potential Return
implicit costs. Indeed, in this formulation,
25% Perold’s market impact includes costs
resulting from spread, market impact,
20% operational delay and market timing
Implementation Shortfall
Return
(Total Transaction Costs)

delay. Therefore, the opportunity cost
15%
component refers only to the missed trade
10%
opportunity cost, i.e., to the cost related to
Realised Return unfilled or partially filled orders.
5%
The mathematical notations for measuring
0.0 the implementation shortfall for a given
-10 -5 -2 Security 1 2 3 5 10 15 20 30
Purchase
Date
trade in a particular security can be
presented as follows.
Days Prior to Days after
Security Purchase Date Security Purchase Date
X = total quantity to execute (total trade

5 - If an investor
simultaneously buys and
size)
Potential Return
sells shares in a security and
Realised Return xi = number of shares executed at price i
recorded buys at the best ask
and sells at the best bid, the Research Period (Manager conducts research and X, xi > 0 for a buy; X, xi < 0 for a sell
develops a sound investment strategy whithin an
resulting paper portfolio will
incur a loss due to the spread. optimal environment) P0 = quotation midpoint at the time of the
investment decision
Source: Plexus Group
PT = quotation midpoint at the end of
trading
Much as with the benchmark comparison, Pi = execution price of ith trade
the key here is the choice of price for EC = all the explicit costs associated with
valuing the paper portfolio. Perold’s the trade
definition requires that all executions
occur at the spread midpoint prevailing on When the trade is completed, the
the market at the time of the investment implementation shortfall is:
decision. This specific valuation is necessary
to avoid charging the paper portfolio with ⎡⎛ N
⎞ ⎤
IS = ⎣⎡ XPT − XP0 ⎦⎤ − ⎢⎜ XPT − ∑ xi Pi ⎟ − EC ⎥
one-half the spread cost on average.5 ⎣⎝ i =1 ⎠ ⎦
N
Calculated with the decision spread IS = ∑ xi Pi − XP0 + EC

midpoint, transaction costs are then divided i =1
into what the author terms execution costs When the trade is not completed, the
and opportunity costs. Execution costs implementation shortfall becomes:
cover all the costs that can be put down
to the actual trade, such as explicit costs ⎡ N N
⎤ ⎛ N
⎞
IS = ⎢ ∑ xi Pi − ∑ xi P0 ⎥ + ⎜ X − ∑ xi ⎟ (PT − P0 ) + E
and market impact. However, as market ⎣ i =1 i =1 ⎦ ⎝ i =1 ⎠
impact is here measured by the difference
⎡ N N
⎤ ⎛ N
⎞
= ⎢ ∑ xi Pi − ∑ xi P0 ⎥ + X − ∑ xi (PT − P0 ) + EC
between the market price at theISdecision
⎜
⎣ i =1 i =1 ⎦ ⎝ i =1 ⎠⎟


Post-Trade Analysis
We can illustrate this primary version of (2) Expanded framework

the implementation shortfall with a very There are several ways to expand Perold’s
simple example. Suppose an investment approach to fine-tune the measurement
decision to buy 500 shares on a given of transaction costs. Some are presented
security traded in € whose implementation in Kissell and Glantz (2003). The expanded
details are the following: version we propose here is adjusted for the
X = 500 classification of implicit cost components
x1=200 and x2=200 provided in section II.
P0 = 20
PT = 22 To measure transaction costs and properly
P1 = 21 and P2 = 21.5 identify where they occur, the investor can
consider four points on the time line: the
Based on the above data, the investment decision time, the order release
implementation shortfall is calculated as time, the time at which the broker begins
follows: to implement the trade and the time at
which the broker stops trading. With the
IS = ⎣⎡(200 × 21 + 200 × 21.5 ) − 400 × 20 ⎦⎤ + (500 − 400 ) (22 − 20 ) + EC
method below, the investor can then
S = ⎣⎡(200 × 21 + 200 × 21.5 ) − 400 × 20 ⎦⎤ + ( 500 − 400 )(22 − 20 ) + EC
identify four reference prices to calculate
the implementation shortfall:
For this investment decision, the missed
trade opportunity cost measure is €200 X = total quantity to execute (total trade
or €0.4 per share, while the other implicit size)
costs (spread, market impact, operational xi = number of shares executed at price i
and market timing opportunity costs) X, xi > 0 for a buy; X, xi < 0 for a sell
amount together to €500 or €1 per P0 = quotation midpoint at the time of the
share. These cost measures can also be investment decision
expressed in basis points (bp). In this case, PR = quotation midpoint at the time the
we obtain 200bp for the missed trade order is released to the broker
opportunity cost and 500bp for the other PS = quotation midpoint at the time the
implicit costs. broker starts working the order
PT = quotation midpoint at the end of
As we have seen, the implementation trading
shortfall is easily understood and Pi = execution price of ith trade
implemented. Although it seems to be an EC = all the explicit costs associated with
effective means of measuring transaction the trade
costs, Perold’s framework delivers
interesting information about how large
transaction costs are but not really about
where they are incurred. To better identify
where they are incurred, the original
version must be expanded.



Post-Trade Analysis
⎡ N N
⎤ ⎡ N N
⎤ ⎡ N N
⎤ ⎛ N
⎞
IS = ⎢ ∑ xi PR − ∑ xi P0 ⎥ + ⎢ ∑ xi PS − ∑ xi PR ⎥ + ⎢ ∑ xi Pi − ∑ xi PS ⎥ + ⎜ X − ∑ xi ⎟ (PT − P0 ) + EC
⎣ i =1 i =1 ⎦ ⎣ i =1 i =1 ⎦ ⎣ i =1 i =1 ⎦ ⎝ i =1 ⎠
}
}
}
}
(1) (2) (3) (4)
In this equation, part (1) measures the 3. Common Pitfalls and

operational opportunity cost, part (2) Shortcomings in Transaction Cost
measures the market timing opportunity Measurement
cost, part (3) measures the cost due to The two approaches that have been
both spread and market impact and part described in depth both require an
(4) measures the missed trade opportunity appropriate benchmark price. Identifying
cost. this price is not always as easy as it
seems. Transaction cost indicators are
To illustrate the degrees of interpretation numerous and their quality often depends
made possible with this expanded on the choice of benchmark. The “ideal”
approach, we may apply this formula to transaction cost indicator should provide
the previous example. For this purpose, an accurate, complete and relevant
the following information must be added: measure of implicit costs. This definition
PR = 20.5 and PS = 21. emphasises several issues that should be
considered when building or choosing
IS = [400 × 20.5 − 400 × 20 ] + [400 × 21 − 400 × 20.5 ] + ⎣⎡(transaction cost ×
200 × 21 + 200 21.5 ) − 400 × 21⎦⎤ + (500 − 400 )
indicators.
0.5 ] + ⎣⎡(200 × 21 + 200 × 21.5 ) − 400 × 21⎦⎤ + (500 − 400 ) (22 − 20 ) + EC (1) Which benchmark for which
purpose?
As observed before, the missed trade In a nutshell, the benchmark on which
opportunity cost measure is €200 or the indicator of transaction costs is built
€0.4 per share, while the other implicit determines what is really measured.
costs amount together to €500 or €1 Accordingly, different benchmarks serve
per share. In addition, we are now able different purposes. As explained above,
to split the global cost of 500 into its benchmark prices can be categorised as
main components: €200 for operational pre-trade, intraday and post-trade. Both
opportunity cost, €200 for market the completeness and the relevance of the
timing opportunity cost and €100 for measures they deliver vary greatly.
both spread and market impact. This
breakdown is relevant as it allows an The pre-trade benchmarks provide the
understanding of exactly where costs are best measures of the implicit costs related
incurred. In our simple example, it seems to the implementation of an investment
obvious that minimising investor-to- decision and should be favoured. They
intermediary transmission delays could include the costs due to spread, market
reduce transaction costs substantially. impact, operational and market timing
delays; only missed trade opportunity
costs are ignored. Although indicators


Post-Trade Analysis
built on pre-trade benchmarks are quite through a measure of price reversion.

complete measures, their accuracy varies However, whatever the specific post-trade
with the gap between the investment benchmark price, this measure refers
decision time and the time at which the only to the temporary impact, because
benchmark price is determined. The most the potential permanent impact is
accurate cost indicator is built on the price incorporated into future prices and is
prevailing at the decision time. When this impossible to isolate. Although their
price is unknown, the prior night’s closing usefulness as measures of cost is limited
price or the opening price is typically when taken alone, post-trade benchmarks
used as a proxy for the decision price, are more relevant when combined with
but the indicators it delivers are then less pre-trade benchmarks to give a measure
accurate. of missed trade opportunity costs.
Such a combination is present in the
The intraday benchmarks do not provide implementation shortfall approach.
good measures of the implicit costs. In fact,
indicators built on intraday benchmarks (2) Critical review of the most popular
do not really give a measure of implicit indicators
costs but rather an indication of the Although we assert that an effective
execution price relative to the average measure of implicit transaction costs
market price of the day or of a given is the difference between the decision
time period. Indicators built on intraday price (its proxy when not available) and
benchmarks are thus neither proper the execution price, we know that this
nor useful cost measures. For example, assertion has not always won unanimous
knowing that a trade has underperformed backing. In this subsection, we offer an
or outperformed the daily VWAP by a in-depth presentation of the transaction
given number of basis points does not cost indicators most often used by the
provide the investor with any insight into industry. We look at both the advantages
how large the implicit costs are and how and disadvantages of these indicators.
they can be better managed. Instead, this
information allows the investor to gauge (a) Spread midpoint benchmark
the quality of the price obtained, if we An indicator based on the spread midpoint
consider that the daily VWAP is a good relies on the signed difference between
indicator of the fair market price. As we the trade price and the average of the bid-
will see in section V, intraday benchmark and-ask at a given time. Spread midpoint
prices are best used to assess the trading indicators are very popular, essentially
performance of intermediaries. because they are very easy to implement
and interpret. The cost of a buy at the
Like the previous category, the post-trade “ask” (or a sell at the “bid”) is one-half of
benchmarks do not really deliver a proper the spread. Varying the time at which the
and complete indicator of implicit costs, spread midpoint is determined delivers
although they are sometimes used for various indicators.
this purpose. Instead, they can provide
an indicator of market impact only


Post-Trade Analysis
The quotation midpoint prevailing at the counterintuitive because market impact

investment decision time is the reference is mainly associated with adverse price
price for computing the implementation movement. The main shortcomings of the
shortfall. The implementation shortfall effective spread are twofold. First, it can
is the most effective of the indicators of provide a measure of market impact but
transaction costs founded on the spread without a distinction between temporary
midpoint. The only disadvantage of and permanent parts. Second, the effective
this indicator is that it requires a large spread may be a poor indicator of market
amount of data, i.e., intraday quotation impact for large orders completed through
data as well as decision time and order multiple trades. When orders are split into
data. Collecting all this data and dealing smaller lots, the indicator should measure
with it can be difficult or quite costly for the total impact of executing the entire
investment managers. In practice, time- order and not simply add up the cost
stamped information on individual trades of each lot. As the first lots impact the
may be incomplete or not gathered in one market, they make the remaining lots
place but held in different systems and more expensive. An indicator based on the
perhaps not in the same firm. Therefore, effective spread will underestimate the
the implementation shortfall indicator total market impact costs if a different
indirectly requires that data acquisition and midpoint is used for each trade.
treatment costs not be a consideration.
The realised spread is based on a post-
The effective spread relies on the quotation trade quotation midpoint. It equals twice
midpoint prevailing at the time of the the signed difference between the trade
trade execution. The effective spread is price and the midpoint observed at a
twice the liquidity premium, which is specified time after the trade. Like the
the signed difference between the trade effective spread, this indicator attempts
price and the time-of-trade midpoint. This to measure market impact. However, this
indicator ignores all the opportunity costs measure misses the permanent impact
but gives a measure of market impact cost, as previously explained. Furthermore,
when compared with the quoted spread. the greater the time gap between trade
When the trade is completed at the quoted execution and midpoint determination,
price (a buy at the ask/a sell at the bid), the the noisier the cost indicator will be.
effective spread equals the quoted spread
and market impact is zero. When the trade (b) VWAP
is completed outside the spread (the large The volume-weighted average price
order is filled at prices outside the best (VWAP) indicator is the signed difference
quote), the effective spread is larger than between the trade price and the market
the quoted spread and market impact is volume-weighted average price computed
positive. When price improvements mean over a given interval. The most frequently
that the trade can be completed inside the used indicator relies on the daily VWAP,
spread, the effective spread is smaller than i.e., the volume-weighted average of
the quoted spread and market impact is all prices in the trading day. Nowadays,
negative. This last case may be a little almost all data vendors as well as some


Post-Trade Analysis
trading venues compute and publish daily signed difference between the trade price
VWAP in real time. However, the window and the average of the lowest, highest,
can be shorter (interval/available VWAP) opening and closing prices of the day.
or longer than one day (multi-day VWAP). This indicator is widely recognised in the
The VWAP-related indicator is widely industry, even though it suffers from the
used in the industry, mainly because it same shortcomings as VWAP indicators.
is easily interpreted: it indicates whether Furthermore, as a simple average of prices
the trader received a higher or lower irrespective of traded quantities, the LHOC
price than did the “average trader” of is less representative of the fair market
the measurement interval. Nevertheless, price since it misses the dimension of
the VWAP benchmark is inappropriate market depth included in the VWAP.
for correct measurement of transaction
costs.
4. Measuring Transaction Costs
(c) Closing price benchmark under MiFID
The indicator based on the closing price Under pressure from MiFID, TCA is set to
is the signed difference between the grow and transaction cost measurement
6 - Fragmentation occurs
when not all orders relative
trade price and the closing price of the in particular is likely to become both a
to a given security interact day. This benchmark is very attractive in widespread and a more than essential
with each other in a
single execution system. the industry because it is easy to get and practice. However, two main issues are
Markets are fragmented
when people can trade the
requires little data processing. All other emerging with the new regulation and
same securities in various popular indicators require intraday market are expected to make transaction cost
trading venues. Markets are
consolidated when all market trade data (VWAP, LHOC) or quotation measurement more complex in the near
participants trade on the
same execution venue. In
data (spread midpoint), while the indicator future. The first has to do with market
practice, we obviously refer to relying on the closing price needs only fragmentation and the second with data
fragmentation as soon as a
given security is cross-listed summary daily market data. Furthermore, availability and processing.
or may be traded on multiple
execution venues.
many investment firms prefer to use
closing prices as benchmarks because (1) Market fragmentation
they already value their portfolios at these The European trading landscape is likely
prices. to become more fragmented6 post-
MiFID since the new regulation clearly
Although attractive, the closing price encourages the proliferation of liquidity
may not provide an effective measure of pools and promotes competition by
implicit costs. It can only deliver indicators putting an end to the well-established
of temporary market impact, because order concentration rule. This rule had
permanent impact is incorporated into hitherto made it possible for most of
future prices. Moreover, such indicators the traditional exchanges to maintain
are noisy when the time gap between the monopolistic positions and be viewed as
execution of the trade and the market official providers of reference prices in
close is long. a great majority of European countries.
By allowing multiple trading venues to
(d) Average of LHOC compete, MiFID is, by design, allowing
The indicator relying on the LHOC is the liquidity pools to fragment and put

III. Measuring transaction costs with

post-trade analysis
the very existence of an acknowledged given security. It is then necessary to

reference price at risk. At first sight, pay attention to the venue(s) the order
both benchmark comparison and is routed to.
implementation shortfall are made
conceptually difficult in the absence of When an order is routed to and entirely
an acknowledged unique price for the executed on a single trading venue
securities traded. When a security is (for whatever reason10), it seems quite
traded at the same time on various consistent to use as benchmarks prices
execution venues, which of the coexisting available on that venue. But when an
prices is the best reference for the order is split into smaller lots that are
security? Several elements can be put routed to several execution venues to
forward to address this point. sweep the market by taking liquidity from
all sources at once—as may be frequent
Under MiFID, investment firms have for large orders—complications arise and
potential access to a wide range measuring transaction costs becomes
of execution venues in the form of a true brain-teaser. When there are
regulated markets,7 MTFs8 and systematic multiple price sources for an order, the
7 - This regime refers to the
traditional exchanges in the
internalisers.9 However, as part of the ability to keep track of and consolidate
MiFID language. best execution obligation, MiFID requires all these prices is eventually the major
8 - Multilateral trading
facilities are similar to that investment firms determine their determinant of the ability to measure
regulated exchanges in that
they allow clients to enter
own unique set of execution venues and costs properly. Theoretically, each lot
into negotiations without define an execution policy documenting should be considered a single order
taking part in a transaction
as counterparty. They include the selected trading venues as well as for which transaction costs could be
all forms of multilateral
negotiations such as order
the factors that led to this selection. This measured based on benchmarks prevailing
books, block trades, periodic selection process reduces the number of on its execution venue. Next, the total
auctions and any other
mechanism resulting in liquidity pools to consider when executing cost of trading for the initial order should
negotiations between two
counterparties.
orders and creates a “frame of reference” be obtained by adding all costs related
9 - This new regime allows within which each firm must meet its to each lot. This approach requires great
investment firms on an
organised, frequent and best execution obligation. Hence, when data storage and processing capabilities.
systematic basis to deal
on their own account by
it comes to measuring transaction costs,
executing client orders only the list of selected execution venues (2) Data availability and treatment
outside a regulated market
or MTF. will be relevant to determine reference Accurate transaction cost measurement
10 - This can be a constraint
set by the client, an
prices. requires a large amount of time-stamped
automated choice when the data about orders and trades. Besides,
venue quotes the best of the
prices available, etc. When only one trading venue is selected as we have just mentioned, post-MiFID
for a given security, we are back to a pre- market fragmentation is likely to tighten
MiFID situation with a single recognised these requirements substantially. In
reference price; therefore, transaction practice, investment firms can have
costs can be measured as simply as trouble collecting information about their
previously described. However, this case own trades as well as collecting market
is likely to be an exception and, most of data. The reason is twofold:
the time it is likely that more than one • Incompleteness of internal databases
trading venue has been selected for a Time-stamped information on individual


post-trade analysis
trades may be incomplete. For example, whether executed on a regulated market

time-stamped data can become available or not. Such consolidated pan-European
only from the time of the release to the market data is being offered by some data
broker. In this case, no data including vendors and new offers should appear in
decision time is available, with the coming months. All these solutions in the
consequence that some transaction cost routing and transformation of real-time
indicators cannot be calculated with market and trade data that aggregate
accuracy. Another possibility is that time- feeds from multiple contributors and
stamped information is held not in one publish this information to a firm’s market
place but in different systems and perhaps data system in real-time are likely to help
not in the same firm. build relevant databases and tools for
• Access to intraday market data proper transaction cost measurement.
Both benchmark comparison and Nevertheless, aggregated market data is
implementation shortfall require intraday still often restricted to equities or liquid
market data, in addition to information securities. Investment firms can collect
about the trades being analysed. When consolidated intraday market data for
several executive venues have been other securities, but the acquisition cost
selected, more data and greater processing can be higher.
capability are required. Consequently, the
firm’s ability to manage all trading data It seems clear then that MiFID has huge
is key. Indeed, one of the most important operational and IT implications for
aspects is the ability to consolidate prices investment firms. There is no doubt that
from multiple and unconnected systems. these implications are going to strengthen
Clearly, the ability to source prices and the presence of technology providers in
integrate them through a single system is the European execution landscape.
a great advantage.
• Data quality, completeness and integrity
For the implementation of MiFID, the
regulator has left to the undustry the choice
of organising the infrastructure needed
to consolidate market and reported data.
As no central solution has emerged, the
risk of having data reported in numerous
places and incorrectly consolidated is
great, resulting in an absence of quality,
completeness or integrity and making the
implementation of post-trade analysis
impossible.
MiFID indirectly addresses the above

issue with harmonised post-trade public
transparency requirements that facilitate
aggregation for all transactions on equities,


post-trade analysis

IV. Estimating Transaction Costs
with Pre-Trade Analysis

IV. Estimating Transaction Costs with

Pre-Trade Analysis
This section is devoted to the second to understand the characteristics of the

mission of TCA—transaction cost estimation. proposed trade and assess the degree of
Its primary purpose is to forecast both difficulty. The concept of “difficulty” is
transaction costs and risk associated with primordial, as it predicts the amount of work
strategies for a given trade or trade list not that will be required to execute the order
yet executed. This assumes that the implicit and the overall need to adjust the trading
costs of future orders can be estimated strategy to changing market conditions.11
from the implicit costs of past trades, using
trade-specific variables (order size, price All the data gathered from pre-trade analytics
limit, etc.), market-related variables (spread, is usually presented in a pre-trade report
depth, recent volume, recent price change, wherein information may be categorised as
etc) and security-specific variables (average security-specific, market-related and trade-
volume, capitalisation, volatility, etc.). specific. We review each category below and
provide insight into how it can help gauge
Since transaction cost estimation is done the difficulty of the order to execute.
ex ante and is more than the mere provision
of cost estimates, it is often referred to as (1) Security-specific information
11 - Kissell and Glantz (2003) pre-trade TCA or simply pre-trade analysis. All the data delivering relevant information
define trading difficulty as
a function of the amount of It is essential for investment managers who about the security to trade fall into this
work required to execute an
order and the overall need for want to offer competitive portfolio returns category. It includes:
traders to adjust to changing to do pre-trade analysis before acting on • the country in which the security was
market conditions and make
modifications to the specified investment decisions. It helps them assess issued
trading strategy.
the difficulty of trades, compare trading • the economic sector
strategies, and develop appropriate strategies • the market capitalisation
for particular market circumstances. In line • the venues where the security is traded
with these three objectives, we can present • some fundamentals such as price-to-
pre-trade TCA as a three-step process: earnings and price-to-book ratios
• Collection and analysis of pre-trade • the index the security belongs to, if any
data • the recent historical price movement
• Cost and risk estimation • the price momentum
• Optimisation of trading strategies • the historical average daily trading
volume
Each step needs an appropriate approach • the historical price volatility
that requires a large amount of data and • the average trade size
particular tools as well as the development • the intraday volume pattern
of relatively complex models. We provide a • etc.
detailed description of each step below.
This list provides a quick reading of the
risk associated with the security to be
1. Collection and Analysis of traded. It is not exhaustive and any variable
Pre-Trade Data investment managers consider relevant
The first step of pre-trade TCA involves could be added.
collecting and analysing pre-trade data


Pre-Trade Analysis
(2) Market-related information With a pre-trade report summarising

This category contains all the data delivering security-specific, market-related and trade-
relevant information regarding the current specific data, investment managers are able
market conditions for the security to trade. to gain a thorough understanding of the
The purpose is here to assess the market trade to be completed. They can then assess
risk of the proposed order. If several trading its difficulty and know at a glance whether
venues are available, comparisons must the trade will be subject to substantial
be made on the conditions prevailing at market impact (for example, the order is
each venue. The variables most relevant large in comparison with the quoted depth
to a description of market conditions are and the average trade) and/or opportunity
commonly: costs (for example, the liquidity to execute
• the last trade prices the order immediately is lacking). Difficult
• the best quotes and the quotation trades will require permanent adjustments
midpoint to the trading strategy.
• the quoted spread
• the quoted depth
• the recent traded volume 2. Cost and Risk Estimation
• etc. The second step of pre-trade TCA is cost
decomposition and estimation. Only
This list should provide the best possible implicit transaction costs are considered
representation of the current market prices here because they depend on the trading
and liquidity available for the security strategy and may thus be controlled during
to be traded. The list is exhaustive if the implementation. The same is not true for
security is traded in transparent liquidity explicit costs, which are known in advance
pools; otherwise, it is not. Indeed, in and do not really vary with the trading
transparent markets, the prevailing liquidity strategy, other than with the choice of
conditions can be easily assessed through venue and intermediary.
the displayed limit order book or market
makers’ quotations. In dark liquidity pools, The proper framework for estimating
by contrast, no information about trading transaction costs attempts to provide a
conditions is disclosed. probabilistic distribution of both an expected
cost and a risk parameter for a trade to be
(3) Trade-specific information completed. Two essential principles emerge
This category covers order size, trade here and it is essential to understand
direction (buy/sell) and other characteristics them clearly. First, cost estimates, unlike
of the order to be executed. When time or transaction cost measures, are not single
price constraints are attached to the order, values. Computing cost estimates involves
they must be added since they can make providing a complete picture of potential
implementation harder. For example, if the costs with their respective risk, where both
client wants the trade to be completed parameters vary with the implementation
before a certain time, this deadline should strategy. Second, estimating a distribution
appear in the list. of cost-risk couples makes it possible to
assess alternative strategies and does


Pre-Trade Analysis
the groundwork for implementation that market impact (Kt). For the latter, we simplify
meets the investor’s goals and respects his reality since there is no distinction between
preferences. permanent impact and temporary impact
that dissipates over time.
Various sophisticated tools and models are
on offer today to this end. Our objective is From the above cost structure, estimating
not to review all of them; it is impossible transaction costs results in making the sum
because there are too many and there is little of the forecasted value of each driver: price
public information about how they work, appreciation, market impact and timing
essentially for commercial reasons. We intend risk. To get such values, we need estimation
instead to document the construction of a models that are able to predict cost values
complete framework for proper forecasting when applied to a specific trading strategy.
of transaction costs. Accordingly, we have We review below each cost driver and use
chosen to refer hereafter to the analytical Kissell and Glantz’s material to show what
approach proposed by Kissell and Glantz such models look like and how they deliver
(2003). These authors put in the public estimates.12
domain a considerable amount of technical
12 - We present here a
summarised version of the
and very detailed material dealing with the (2) Price appreciation
models developed by the means of modeling and forecasting implicit As explained above, price appreciation is
authors. For more detail, see
Kissell and Glantz (2003). transaction costs. Their work allows us to the natural price movement of a security.
illustrate the approach without promoting It is sometimes referred to as price trend.
a particular commercial tool. We will see It is a truism to say that transaction costs
also that this approach makes it possible to depend on price appreciation: momentum
determine the cost profile of any trading traders tend to incur large transaction costs
strategy through simultaneous estimation because they buy when prices have risen
of cost and risk, which is the basis for and sell when they have fallen. By contrast,
developing optimisation techniques to contrarian traders buy when prices have
derive an efficient trading frontier and fallen and sell when they have risen, so their
determine the optimal strategy most closely transaction costs tend to be lower. Taking
aligned with the investor’s final objective. price appreciation into account thus adds
substantial value to the implementation
(1) Cost structure process.
Kissell and Glantz (2003) build their
analytical framework for cost estimation on To predict the cost of price appreciation, we
the following price trajectory formulation: need first to develop price forecasts over
Pt = Pt −1 + Ut + Kt + Et , where Pt is the the trading horizon and then determine
price of trade t, Ut is the natural price cost estimates for the specified execution
appreciation from time t-1 to t, Kt is the strategy.
market impact of trade t and Et is the
price volatility with Et~N(0,σ²). In this (a) Price forecast models
price trajectory, we can identify price Both fundamental analysis and technical
appreciation (Ut) and timing risk (Et) as analysis provide price forecasts in the form
drivers for opportunity costs as well as of expected short-term or long-term price


Pre-Trade Analysis
movement. We can derive from it target changes by a constant percentage of the

prices for all the periods in the trading current price. Applying this model to our
horizon. Three models can be used for this example, we get the exponential price
purpose. To present them and facilitate change (g) and the expected price (Pt) as
understanding, we refer to the following follows:
numerical example. 1 ⎛P ⎞ 1 ⎛ 125 ⎞
g = ln ⎜ T ⎟ = ln ⎜ = 0.0026%
n ⎝ P0 ⎠ 8500 ⎝100 ⎠⎟
Suppose a security that is currently traded
at €100 (P0) and is expected to increase Pt = P0 e ( m × g ) =100 × e ( 68 × 0.000026 ) =100.18
to €125 (PT) after one year. If we consider
fifteen-minute periods, we have 8,500 Return and growth models are
trading intervals (n) over the forecasted interchangeable since they deliver
horizon.13 We can first determine the similar results for any short-term trading
expected price change over each trading horizon.
interval and then calculate the expected
price at the end of two days (sixty-eight (b) Cost estimates
trading intervals (m)). Consistent with the implementation
13 - We consider that there
are thirty-four intervals of
shortfall approach and if we apply the
fifteen minutes in a usual Linear model linear price change model, the transaction
trading day (9:00AM to
5:30PM) and 250 trading This model assumes constant price changes cost component due to price appreciation
days in a year.
14 - The method is similar
regardless of current prices. Applying this is estimated as follows.14
when we consider return and model to our example, we get the constant
growth models.
price change (ΔP) and the expected price P0 = security price at the beginning of
(Pt) as follows: trading
1 1 ΔP = linear price change forecast
ΔP =
n
( PT − P0 ) =
8500
(125 −100 ) = 0.0029 Pj = forecasted price in period j;
X = order size; X>0 for buys and X<0 for
Pt = P0 + mΔP =100 + 68 × 0.0029 =100.20 sells
xj = number of shares traded in period j;
n
Return model
This model assumes that the price changes
∑x j
=X
i =1
by a constant percentage of the current U(X) = cost of price appreciation for order X
price. Applying this model to our example, n
we get the percentage price change (r) and U( X ) = ∑ x j Pj − XP0

the expected price (Pt) as follows: j =1
n n
1 1 U( X ) = ∑ x j P0 + ∑ x j j ΔP − XP0
⎛P ⎞ n
⎛ 125 ⎞ 8500
j =1 j =1
r = ⎜ T ⎟ −1 = ⎜ −1 = 0.0026%
⎝ P0 ⎠ ⎝100 ⎠⎟ n
U( X ) = XP0 + ∑ x j j ΔP − XP0
j =1
Pt = P0 (1 + r ) =100 × (1 + 0.000026 ) =100.18
m 68
n
U( X ) = ∑ x j j ΔP
Growth model j =1
This model also assumes that the price


Pre-Trade Analysis
U(X) estimates the cost expressed in than price appreciation and depends on
monetary units for the total order size. To elements such as imbalances, volatility,
get the cost per share or in basis points, we trading style and liquidity, the approach
simply divide U(X) by the order size (X) or is more difficult and technical. We provide
the traded value at the beginning of trading hereafter a summarised version of Kissell
(XP0). Because U(X) is a cumulative function, and Glantz’s market impact model and
predicting the cost of price appreciation for illustrate how it can be used to derive cost
a specific trade list that contains several estimates for given trading strategies.
orders and securities results in making the
sum of cost estimates for each order and (a) Market impact model
listed security. As explained just above, market impact
depends on several factors:
We can illustrate the method above with a • Order size and imbalances: market impact
small example. Consider an order of 170,000 is positively related to order size and the
shares to be executed over three trading liquidity supply-demand imbalances it
periods. The current security price is €50 causes;
and this price is expected to increase by • Volatility: higher volatility means greater
15 - Liquidity demanders
consume the available
€0.10 per period. If the execution strategy price elasticity and higher transaction
liquidity by trading at market involves trading 60,000, 60,000 and 50,000 costs;
prices and contribute to the
supply-demand imbalance. By shares in periods 1, 2 and 3; we estimate the • Trading style: aggressive trading strategies
contrast, liquidity providers
supply liquidity to obtain
cost of price appreciation as follows: lead to higher market impact than do passive
better prices for trades that strategies;
they would like to complete.
Hence, only liquidity U( X ) = 60000 × 0.10 + 60000 × 2 × 0.10 + 50000 × 3 ×conditions:
• Market 0.10 = 33000market impact is
demanders pay temporary
U( impact
X ) =costs.60000 × 0.10 + 60000 × 2 × 0.10 + 50000 × 3 × 0.10 = 33000 negatively related to the available liquidity
16 - Permanent impact is due over the trading horizon.
to the net effect of all market
participants and all market This cost of €33,000 is equivalent to €0.19
participants incur permanent
impact costs. This assumption
or 39bp per share. Based on the previous factors, Kissell and
is valid if (i) all trades convey Glantz (2003) model the total market impact
some information to the
market and the security (3) Market impact cost as: MI = αI + (1 − α )I , where α is the
current fundamental value
is continuously assessed; (ii)
Market impact is the price shift caused by percentage of temporary impact; (1-α) is
the information content is a particular trade. It can be broken down the percentage of permanent impact and I
proportional to the trade size
and similar for buy orders into short-lived impact that dissipates over is the total market impact cost. The latter
and sell orders.
time and long-term disturbances in the is expressed in monetary units and depends
price trajectory. on both imbalance and volatility.
As with price appreciation, we need first a Assuming that only liquidity demanders
model that relates the market impact cost incur temporary impact costs15 and that
of past trades to a number of trade-specific, permanent impact is a function of the
market-related, and security-specific net imbalance,16 the market impact cost
variables. We then use the calibrated model expressed in monetary units per share is
to obtain cost estimates for the specified rewritten as:
execution strategy. However, because market αI (1 − α )I
impact is a more complex cost component MI = + where Vside is the
Vside Q


Pre-Trade Analysis
traded volume from liquidity demanders As previously explained, I is a function of

and Q is the net imbalance. both imbalance and volatility. By defining
V
Q , the authors obtain the following
η= side
In this top-down approach, the market market impact formulations:

impact cost for an order of X shares is then MI( Q ) = I( Q,σ ).( αη −1 + (1 − α )) = I( Q,σ ).d( η)
given by:
MIbp = Ibp ( Z ,σ ).d( η)
⎡ αI (1 − α )I ⎤
MI( X ) = X ⎢ + ⎥
⎣ Vside Q ⎦ . The first equation specifies the total market
impact cost in monetary units, the second
As we may observe, if the order size is in basis points, with imbalance stated as
equal to the imbalance and accounts for a percentage of average daily volume (Z).
all liquidity demand (X= Q= Vside), the Both equations now present market impact
total market impact cost incurred by the as a product of two functions where d(η)
investor is I. Hence, the market impact is the dissipation function and I(Z,σ) is the
cost for a specific trading strategy MI(xk) instantaneous market impact function.
can be derived according to the following The variables in each function have to
principles. First, the percentage of temporary be determined and the parameters have
17 - The results reported by
impact in any trading period k is equal to the to be jointly estimated with advanced
Kissell and Glantz (2003) are percentage of the imbalance in that period. regression techniques. We report below
based on numerous analyses
on various sample sizes, The total temporary impact cost can then be what the authors propose given their own
volatility, participation rates
and mixed data samples.
allocated to each trading period based on analyses.17
the percentage of imbalance in each period
q
( k where qk is the net imbalance in period Variables
Q
k). Next, by applying an average permanent The instantaneous market impact function
impact cost across all trades, allocation of is built on two variables, Z and σ. The first
total permanent impact across periods is not represents the imbalance expressed as a
necessary. The cost of market impact for the percentage of average daily trading volume
trading strategy xk is therefore calculated and is computed as follows:
as follows: Q
Z= ×100
n
⎡ q αI (1 − α )I ⎤ ADV ,
MI( xk ) = ∑ xk ⎢ k + ⎥
⎣ Q V Q ⎦
k =1 side
where n
Q = ∑ sign( vi )
(b) Parameter estimation i =1
To calibrate their model, Kissell and Glantz with sign(vi) equal to the signed trade size
(2003) start with the total cost of market and
impact over an entire trading day when
1 T
X=Q: ADV = ∑v
T t =1 t with T usually equal to
⎡ αI (1 − α )I ⎤ αIQ thirty trading days. The second variable
MI( X = Q ) = Q ⎢ + ⎥= + (1 − α )I
⎣ Vside Q ⎦ Vside σ refers to the price volatility factor and
corresponds to the close-to-close measure
of the volatility of logarithmic price change


Pre-Trade Analysis
⎛P ⎞
for the previous T trading days (most of ϕQ = sign( Q ) ⎜ Q −1⎟ ×10 4 .
the time T=30). Mathematically, we have
⎝ P0 ⎠
1 T
σ= ∑
T −1 t =1
( gt − g) 2 , where Parameter estimates
For the parameter α in the dissipation
Pt function, the authors found in almost all
gt = ln their regressions that it is about 95%; this
Pt −1 .
estimate is not dependent on time and size.
The dissipation function relies on the variable We can then rewrite the dissipation function
η that represents the participation number. as: d( η) = 0.95η −1 + 0.05 . Since α refers
It is used as a proxy of trading style and to the temporary impact, Kissell and Glantz’s
is calculated as the number of equal size findings suggest that temporary impact cost
orders executed in the market over the same is by far the largest part of total market
trading period as the investor’s order. impact cost.
side
V
Specifically, we have η = . Kissell and Glantz (2003) investigate three
Q
possible structures for the instantaneous
18 - According to Kissell
and Glantz (2003), the
Taking the absolute value of Q is required to market impact function: the linear function,
three functions yield fairly ensure that the participation rate is positive. the non-linear function and the power
consistent results across
size and volatility. However, The lower the participation number is, the function. Each requires i parameters termed
the non-linear and power
functions exhibit smaller
shorter the trading time horizon is, and the ai. We present below each possible function
regression errors, suggesting more aggressive the strategy is. with both the general formulation and
that the true relationship
between cost and size is the formulation including the estimated
non-linear.
Finally, estimation of the various parameters parameters.18
of the market impact model requires an
additional variable that represents the • Linear function:
trading cost associated with the imbalance Ibp ( Z ,σ ) = a1 Z + a2 σ + a3
Q. This cost expressed in monetary units
Ibp ( Z ,σ ) = 8Z + 0.3σ + 90
is computed according to the following
equation: • Non-linear function:
a2
Ibp ( Z ,σ ) = a1 Z + a 3 σ + a4
ϕQ = ∑ x j pj − XP0 = XPQ − XP0 = X( PQ − P0 )
Ibp ( Z ,σ ) = 35 Z 0.65 + 0.3σ +15
where P0 is the market price at the beginning • Power function:
of trading and PQ is the volume-weighted Ibp ( Z ,σ ) = a1 Z 2 σ
a a3
average price for shares on the same side

Ibp ( Z ,σ ) = 25 Z 0.38σ 0.28
as the imbalance, that is:
PQ =
∑v p
i i
for i ∈ Q. (c) Forecasting market impact cost
∑vi
The forecast of market impact cost for an
The cost of imbalance expressed in value per order K(X) and a particular trading strategy
share and in basis points is given by K(xk) is based on the following equations:
ϕQ = sign( Q ) (PQ − P0 ) and


Pre-Trade Analysis
K( X ) = I( 0.95η −1 + 0.05 ) We finally obtain that the market impact

cost (in monetary units) for an order of X
n
0.95Iqk 0.05I
K( xk ) = ∑ xk + shares implemented over n-trading periods
k =1 Qv k ,side Q following strategy xk is:
To apply these equations, estimated values n
⎡ 0.95Ixk 0.05I ⎤
are needed for imbalance Q, liquidity K( xk ) = ∑ xk ⎢ + ⎥
demander volume by period vk,side and
k =1 ⎣ X( xk + 0.5vk ) X ⎦
instantaneous cost I. Kissell and Glantz
(2003) discuss two cases that require specific Since market impact cost is a cumulative
forecasting techniques. function, the market impact cost estimated
for a list of m-securities traded over
First case: the imbalance is equal to the n-periods is simply the sum of costs for all
order size orders. That is:
This situation corresponds to the common
m n ⎡ 0.95Ii xi ,k 0.05Ii
assumption that Q=X and each xk=qk, since K( xk ) = ∑ ∑ xi ,k ⎢ + ⎥
intentions of other market participants i =1 k =1 ⎢⎣ X i ( xi ,k + 0.5vi ,k ) X i ⎦⎥
are hard to assess before trading begins.
19 - Patterns of daily trading
volumes are used to exhibit a Under this assumption, instantaneous
m n
cost⎡ 0.95Ii xi ,k 0.05Ii ⎤
day-of-week effect. A factor K( x
is determined as follows: k ) = ∑ ∑ x i ,k ⎢
X ( x + 0.5v )
+
X
⎥
specifying the day’s historical
percentage of the average
i =1 k =1
⎣⎢ i i ,k i ,k i ⎦⎥
daily volume can improve the I = Ibp ( Z ,σ )10 P0 Q where
−4
daily volume forecast.

20 - This case is possible
X Second case: there is an incremental
when investors can formulate
realistic expectations
Z= 100 imbalance due to other market
about buying and selling ADV
pressure from other market
participants
participants. It often As for liquidity demander volume, This situation corresponds to the assumption
happens at times of public
announcements or index the expected volume on the day that Q=X+Y, where Y is the expected net
reconstitution.
V(t) is first computed as: where imbalance of other market participants.20
DOW(t) is a day-of-week adjustment Appropriate adjustments in the previous
factor accounting for the weekly volume developments are necessary to incorporate
effect.19 The percentage of volume in this piece of information. Accordingly,
each trading period vk is then obtained the instantaneous market impact cost
by applying the security volume profile uk, becomes
which represents the percentage of the day’s
total volume traded in each period: I = Ibp ( Z ,σ )10 −4 P0 X + Y , where
vk = V (t )uk X +Y
. Z= 100 with X or Y >0 for buys
Since the total imbalance is assumed equal to ADV
the order size, vk consists of an equal amount and X or Y<0 for sells. Next, the liquidity
of buy and sell volume. Consequently, the demander volume in each trading period
liquidity demander volume vk,side will be is now:
equal to the imbalance in each period (xk)
vk ,side = xk + y k + 0.5vk
plus half of the market volume. We have
thus v k ,side = xk + 0.5v k .


Pre-Trade Analysis
It is worth noting that the absolute value assume that the parameter values obtained
function is necessary to account for the by the authors are suitable for our examples.
cases when the investor’s order and the net However, pre-trade analysis assumes that
imbalance of other market participants are the market impact of future orders can
on opposite sides. Furthermore, whenever be predicted from the market impact of
the net imbalance of other market past orders. This assumption is valid when
participants is larger and on the opposite liquidity conditions do not change too much.
side, the investor incurs a savings rather So, we recommend to those interested in
than a cost. To address this potential savings, using this market impact model to determine
the calculation must provide the investor their own calibration with their own recent
with the correct sign for market impact data sample, since the data used by the
cost. This is made through the following authors may be not relevant for estimation
cost function: on other markets, securities and periods.
sign( k ) = sign( X ).sign( X + Y ) Example 1: Forecasting market impact for

an order when the imbalance is equal to
From all the previous adjustments, the order size
market impact cost (in monetary units) An investor has a buy order for 100,000
for an order of X shares implemented over shares (X) on a particular security whose
n-trading periods following strategy xk is: current market price is €30 (P0), daily
n
⎡ 0.95I volatility
xk + y k is 200bp/day 0.05I ⎤(σ) and average
K( xk ) = sign( k ).∑ xk ⎢ +
daily volume is 1,000,000 ⎥ shares (ADV).
k =1 ⎣⎢ X + Y ( xk +The y k + 0.5vk ) X + Y ⎦⎥
investor wants to execute the order
n
⎡ 0.95I xk + y k 0.05I ⎤ over an entire day whose day-of-week
K( xk ) = sign( k ).∑ xk ⎢ + ⎥
k =1 ⎢⎣ X + Y ( xk + y k + 0.5vk ) X + Y ⎦⎥ adjustment factor is about 0.95 (DOW).
Given the data at hand, how can he estimate
the market impact cost for his order (in
For a list of m-securities traded over monetary units)?
n-periods, the market impact cost
forecasting equation becomes: The following five steps lead to an
m n ⎡ answer:
0.95I xi ,k + y i ,k 0.05I ⎤
K( xk ) = ∑ ∑ sign( ki ). xi ,k ⎢
i
1 express the imbalance + as a i percentage
⎥
i =1 k =1 xi ,k + y i ,k + 0.5vi ,k ) X i + Yi ⎦⎥
⎣⎢ X i + Yi ( of ADV:
100000
m n ⎡ 0.95Ii xi ,k + y i ,k 0.05Ii ⎤ Z= ×100 =10
K( xk ) = ∑ ∑ sign( ki ). xi ,k ⎢ + ⎥ 1000000
i =1 k =1
⎣⎢ X i + Yi ( xi ,k + y i ,k + 0.5vi ,k ) X i + Yi ⎦⎥
2. determine the market impact instanta-
(d) Illustration neous cost using the power function (for
We now present a set of examples to example):
illustrate how Kissell and Glantz’s (2003)
market impact model can, 0.38 once it is I = ( 25 ×10 0.38 × 200 0.28 ) ×10 −4 × 30 ×100000 =
calibrated, be usedIfor
= (easy
25 ×10 × 200 0.28 ) ×10 −4 × 30 ×100000 = 79314
cost estimates.
For convenience, we change nothing and


Pre-Trade Analysis
3. compute the expected volume on the 4. calculate the participation number:

day:
V = 0.95 ×1000000 = 950000 100000 +100000 + 0.5 × 950000
η= = 3.38
100000 +100000
4. calculate the participation number: 5. determine the sign:
100000 + 0.5 × 950000 sign(100000 ) × sign(100000 +100000 ) = + × + =
η= = 5.75
100000
sign(100000 ) × sign(100000 +100000 ) = + × + = +
5. determine the expected market impact 6. compute the expected market impact
cost: cost:
K( X ) = 79314 × ( 0.95 × 5.75 + 0.05 ) =17070
−1 100000
K( X ) = +206429 × × ( 0.95 × 3
K( X ) = 79314 × ( 0.95 × 5.75 + 0.05 ) =17070
−1
100000 +100000
100000
K( X ) = +206429 × × ( 0.95 × 3.38 −1 + 0.05 ) = 34214
100000 +100000
Example 2: Forecasting market impact for
an order when there is an incremental
imbalance Example 3: Forecasting market impact for
Suppose a situation similar to the previous a given strategy when the imbalance is
example, except that the investor believes equal to order size
now that there will be an additional An investor has a buy order for 200,000
imbalance of 100,000 shares (Y) on the same shares (X) and wants to execute 40,000
side as his order. Given the data at hand, shares (xi) in each of the coming five periods.
how can he estimate the market impact Forecasts for expected market volume in
cost for his order? each period are 250,000 (v1), 200,000 (v2),
100,000 (v3), 200,000 (v4) and 250,000 (v5).
Now the following six steps lead to an If the instantaneous market impact cost
answer: is estimated at €150,000 (I), how can he
1. express the imbalance as a percentage compute the cost related to his strategy?
of ADV: ⎡ 0.95 xk
I n ⎤ 150000
Z=
100000 +100000
×100 = 20
K( xk ) = ∑ xk ⎢
X k =1 ⎣ xk + 0.5vk
+ 0.05 ⎥ = ∑
1000000 ⎦ 200000 k
I n
⎡ 0.95 x ⎤ 150000 5
⎡ 0.95 x 2 ⎤
K( xkthe
2. determine ∑ ximpact
) =market ⎢
k
+ 0.05 ⎥ =
instanta-
X k =1 k ⎣ xk + 0.5vk 200000
∑ ⎢ x + 0.5vk + 0.05 xk ⎥ = 50271
neous cost using the power function (for ⎦ k =1 ⎣ k k ⎦
I n example):
⎡ 0.95 xk ⎤ 150000 5 ⎡ 0.95 xk2 ⎤
K( xk ) = k ⎢
X
∑x + 0.05 ⎥ = ∑ ⎢ x + 0.5v + 0.05 xk ⎥ = 50271
I =⎣ x( k25+ ×
k =1 0.5v
20k0.38 × 200⎦ 0.28200000
) ×10 −4 k×=130
⎣ ×k 200000
k = 206429 ⎦
I = ( 25 × 20 × 200 ) ×10 × 30 × 200000 = 206429
0.38 0.28 −4
Example 4: Forecasting market impact
for a given strategy when there is an
3. compute the expected volume on the day: incremental imbalance
V = 0.95 ×1000000 = 950000 An investor has a buy order for 200,000
shares (X) and wants to execute 40 000
shares (xi) in each of the coming five periods.
Forecasts for expected market volume in


Pre-Trade Analysis
each period are 250,000 (v1), 200,000 (v2), cost for the security, Δp is the expected
100,000 (v3), 200,000 (v4) and 250,000 (v5). price trend per period for the security, X is
Furthermore, the investor believes that there the order size, xk is the number of shares
will be an incremental market imbalance of to trade in period k and vk the expected
100,000 shares (Y). Incremental imbalances volume for the security in period k. Hence,
allocated to each period are 35,000 (y1), the uncertainty surrounding the transaction
30,000 (y2), 20,000 (y3), 10,000 (y4) and cost estimate can be computed as the
5,000 (y5) respectively. If the instantaneous standard deviation of the previous equation,
market impact cost is estimated at €300,000 that is:
(I), how can he determine the cost related
ℜ( φ) = E( φ 2 ) − [E( φ) ]
2
to his strategy?
I n
⎡ 0.95 xk + y k ⎤
K( xk ) = sign( k ) ∑ x ⎢⎢ x
k
+ 0.05 ⎥ Assuming that volume and price movement
X +Y k =1 ⎣ k + y k + 0.5vk ⎦⎥ are independent21 simplifies the calculation
300000 5
⎡ 0.95 xk xk + y k ⎤ and allows the following equation:
K( xk ) = +
200000 +100000
∑⎢⎢ x + y k + 0.5vk
+ 0.05 xk ⎥ = 83350
k =1 ⎣ k ⎦⎥ ℜ( φ) = σ 2 [U( xk ) ] + σ 2 [K( xk ) ]
where timing risk is obtained through
(4) Timing risk a simple combination of price risk and
Timing risk is related to the uncertainty liquidity risk.
associated with price movements and cost
estimates. It can be broken down into price (a) Price risk
risk, which consists of price volatility, and The most common measure of price risk
liquidity risk, the result of fluctuations in is the standard deviation of price returns.
21 - There is a negligible market conditions. Price risk affects price For trading purposes, price return volatility
degree of correlation between
price changes and market
appreciation estimates (U(xk)) while liquidity must be converted into monetary units per
volume, but the absolute risk affects market impact estimates share as follows:
value of price changes and
that of volume are correlated. (K(xk)). Here again, we will refer to Kissell 1. assuming that p is not much different
This means that the presence
of high volume does not
and Glantz’s material to document how from each pt, we can write:22
indicate the direction of price timing risk components can be modelled
movement. pt − pt −1 pt − pt −1 1
22 - Natural log of price and forecasted. rt = ≅ = Δpt
returns can also be used. pt −1 p p
Both calculations deliver very
similar results for normal Consistent with what we have seen earlier, 2. since p is a constant:
market conditions.
we can write the following transaction cost 1
σ 2( r ) = σ 2( Δpt )
estimate equation: p
2
⎡ ⎤ ⎡ 0.95 x I 2
⎤
φ( xk ) = U( xk ) + K( xk ) = ⎢ ∑ xk k Δp ⎥ + ⎢ ∑3. for short time+periods
k
0.05I ⎥ (≤ 5 days) and
⎣k X( xk + 0.5vk )
⎦ ⎣ k with E( r ) ≤ 50% and⎦ σ( r ) ≤ 75% on
a yearly basis, the current security price p0
⎡ ⎤ ⎡ 0.95 xk2I ⎤
φ( xk ) = U( xk ) + K( xk ) = ⎢ ∑ xk k Δp ⎥ + ⎢ ∑ + 0.05I ⎥ may be substituted for p with little loss of
⎣k ⎦ ⎣ k X( xk + 0.5vk ) ⎦ , accuracy:
1 2
σ 2( r ) ≅ σ ( Δpt )
where I is the instantaneous market impact p02


Pre-Trade Analysis
4. hence, variance and volatility expressed Price risk of a list of m-securities

in monetary units per share are easily If we consider a list of m-securities traded
obtained as follows: σ ( Δpt ) ≅ p0 σ ( r )
2 2 2
over n-periods, the approach described
and σ( Δpt ) ≅ p0 σ( r ) . above must be extended to a changing
portfolio whose risk depends on both
5. therefore, the total variance and individual security volatility and the
volatility in monetary units for an covariance of price movement across all
order is: σ 2( X ) ≅ X 2 p02σ 2( r ) and pairs of securities. To address this issue, we
σ( X ) ≅ Xp0 σ( r ) . need first to understand how to convert
the covariance matrix into monetary units
per share and how to compute the total
Price risk of a specific trading strategy in a portfolio variance and risk.
single security
Based on the variance of the security Using the same approximation methodology
expressed in monetary units, computing price as for individual security volatility, we
risk for a specific trading strategy results can convert the covariance of returns of
in computing the risk of a one-security securities i and j as follows:
position that fluctuates per period. If xk is 1 1
the number of shares traded in period k, Cov( ri ,rj ) = E( ri rj ) − E( ri )E( rj ) ≅ ⎡E( Δp
pi pj ⎣
rk is the number of unexecuted shares at
1 1
Cov( ri ,rj )k =and
E( rσ²
r ) − E( ri )E( rj ) ≅ 1 j ) 1− E( Δpi )E( Δpj )1⎤ 1
the beginning of period i j is the per Cov( r ⎣⎡,rE(j )Δp Δp ⎦
≅ E( ri prj ) −cov( )E(i r, jΔp
E( ri Δp ) ≅j )
i
p p
Cov( j ri ,rj ) =p
i ⎡E( Δpi
period variance for the security in monetary pi pj ⎣
i
0 ,i 0,j
units/share, the total price risk 1 of1the trade

Cov( ri ,rj ) ≅ cov( Δpi , Δpj ) 1 1
p0 , j
schedule over n-periods pis0 ,icalculated as Cov( ri ,rj ) ≅ cov( Δpi , Δpj )
follows: n
p 0 ,i
p0 , j
σ 2( xk ) = ∑ rk 2σ 2
k =1
Hence, the covariance of security prices
n
expressed in monetary units/share is:
σ( xk ) = ∑r σ k
2 2
k =1
Cov( Δpi , Δpj ) ≅ p0 ,i p0 , j cov( ri ,rj ) ≅ p0 ,i p0 , j ρi , j σ
We can illustrate thisCov(
approach with
Δpi , Δp j
a small
)≅ p0 ,i p0 , j cov( ri ,rj ) ≅ p0 ,i p0 , j ρi , j σ( ri )σ( rj ) ,
example. Suppose a trader has an order for
12,000 shares (X) and wants to execute where ρi,j is the correlation coefficient, and
4,000 shares (xi) in each of the coming three the covariance in monetary units for two
periods. If the per period volatility of the orders Xi and Xj is then:
security is estimated at €0.04/share, the
Cov( X i Δpi , X j Δpj ) ≅ X i p0 ,i X j p0 , j cov( ri ,rj )
price risk of the given strategy is computed
as follows: Expanding to a list of m-securities, the
covariance matrix C in monetary units/
σ 2( xk ) =12000 2 × 0.04 2 + 8000 2 × 0.04 2 + 4000 2 × 0.04 2 = 358400
share is computed from the covariance
2
( xk ) =12000 2 × 0.04 2 + 8000 2 σ( xk ) 2=+ 4000
× 0.04 358400 2 = 599
0.04 2
=2 358400 matrix of return Cr as follows: C = DCr D ,
σ 2( xk ) ×=12000 × 0.04 2 + 8000 2 × where 0.04 2 D+ is4000 2
× 0.04 2
= 358400
the diagonal matrix of the current
xk ) = 358400 = 599 σ( xk ) = 358400 = 599 price of the m-securities:


Pre-Trade Analysis
⎛ p1 0 ... 0 ⎞
⎜ 0 p ... 0 ⎟ I x j2
D=⎜ 2
⎟
K( x ) =
X
∑x + 0.5v j
j j
⎜ ... ... ... ... ⎟
⎜ ⎟ and then make two assumptions. First, that
⎝ 0 0 ... pm ⎠
successive volumes are independent from
From the above approximations for security period to period. Second, that the estimation
volatility and covariance matrix, we can error of the instantaneous market impact
easily compute the total variance and risk function I may be ignored. Hence, they
for a static portfolio of m-securities. If X obtain that liquidity risk is estimated as
represents a vector of security positions follows:
where xi is the number of shares held in ⎡I x j2 ⎤ ⎛ I ⎞2
the ith security and C is the covariance σ 2( K ( x )) = σ 2 ⎢ ∑ ⎥=⎜ ⎟ ∑
matrix expressed in monetary units per ⎢
⎣ X j ( x j
+ 0.5v )
j ⎦⎥ ⎝X⎠ j
share, we obtain the total risk of the 2
2
( X ) ⎡
2 XI' CX and
xj ⎤ ⎛ I ⎞2 ⎛ x j2 ⎞
portfolio as follows: σ =
σ ( K ( x )) = σ ⎢ ∑ ⎥ = ⎜ ⎟ ∑σ ⎜
2 2
⎟
σ( X ) = X ' CX . ⎣⎢ X j ( x j + 0.5v j ) ⎦⎥ ⎝ X ⎠ j ⎝ x j + 0.5v j ⎠
23 - For more detail, see

Kissell and Glantz (2003), pp.
If we consider now a list of m-securities Since the only random variable in this
126-127. traded over n-periods, its price risk has to expression is vj, the authors solve it with
be derived from the changing portfolio the theorem below, which relies on the
approach. If xk is a vector where xi,k is the assumption that the function is at least
number of shares of security i to be traded twice differentiable.23
in period k, rk is a vector where ri,k is the
number of unexecuted shares of security i THEOREM
in period k and C the per-period covariance Let v be a random variable with E(v)=μ and
matrix expressed in monetary units per σ²(v)= σ². If Y=H(v) then,
share, we compute the total risk in monetary H''( μ) 2
units for the specified trading strategy as E(Y ) ≅ H( μ) + σ
2
follows:
σ 2(Y ) ≅ [H'( μ) ] σ 2
2
n n
σ 2( xk ) = ∑ rk 'Crk and σ( xk ) = ∑ r Cr k
'
k If the expected value and variance of
k =1 k =1
each period j is E(vj)=vj and σ²(vj)= σ²vj
(b) Liquidity risk respectively, liquidity risk is computed as
Liquidity risk is the risk of unexpected follows:
movements in market conditions. It must x 2
H( v ) =
be assessed to account for the total risk ( x + 0.5v )
involving price impact estimates. −x 2
H'( v ) =
2( x + 0.5v ) 2
x 4σ 2 ( v )
To formulate liquidity risk, Kissell and Glantz σ 2(Y ) =
4( x + 0.5v ) 4
(2003) start from the following total market
impact expression: Hence, the liquidity risk of the execution
schedule Kx is given by:


Pre-Trade Analysis
⎛I⎞
2
⎛ x j2 ⎞ ⎛ Iand
⎞ timingx j risk
2 4
σ 2( vprovide
) very interesting
σ ( K( x )) = ⎜ ⎟
2
⎝X⎠ ∑ ⎝⎜ x + 0.5v ⎠⎟ ⎝⎜ Xmaterial

σ 2
=
⎠⎟ ∑ that, when we
j
put all the pieces
j 4( x j + 0.5v j )
4
j j j
together, makes it possible to determine
2 σ(2 K( x )) = σ 22( K( x )) 4 2 the cost profile of any trading strategy
I
⎛ ⎞ ⎛ x ⎞ I
⎛ ⎞ xj σ (vj )
σ ( K( x )) = ⎜ ⎟ ∑ σ ⎜ ⎟ =⎜ ⎟ ∑
j
2 2
through simultaneous estimation of cost
⎝X⎠ j ⎝ x j + 0.5v j ⎠ ⎝ X2 ⎠ j 4( x⎛j +I 0.5v
2
⎞ j 2 ) 4
⎛ x 2
and⎞ risk. 2
⎛ I This x j4σ 2( v j )of material is
⎞ combination
σ ( K( x )) = ⎜ ⎟ ∑ σ ⎜
⎝⎜ X ⎠⎟ ∑
j
⎝X⎠ j ⎟ =foundation
σ( K( x )) = σ 2( K( x )) ⎝ x j + 0.5v the
j ⎠ j 4(for 0.5v
x j +the )4
development
j
of
techniques to determine optimal trading
σ( K( x )) = σ 2( K( x )) strategies and to derive an efficient trading
frontier. We address these points in the next
The liquidity risk for a trade list of several subsection.
securities can be determined as above but
by making an additional assumption of
independence based on zero correlation 3. Optimisation and Efficient
of excess volumes24 across securities and Trading Frontier
across periods for the same securities. So, The cost profile of a specific trading strategy
assuming independence, the liquidity risk xk summarises the cost and risk estimates
24 - Volume quantities above/
below the mean in each
for a trade list is: 2
associated with the strategy. It is expressed
trading interval. ⎛I ⎞ xij4σ 2( vij ) as θk = ( φ( xk ),ℜ( xk )) , where θk is the
25 - There is a negligible σ 2( K( x )) = ∑ ⎜ i ⎟ ∑4( x
degree of correlation between i ⎝ Xi ⎠ j ij
+ 0.5vij ) 4 cost profile, φ(xk) is the expected implicit
price changes and market
volume but the absolute
transaction ( φ( and
θk =cost xk ),ℜ((xxkk) ))is the forecasted
value of price changes and σ( K( x )) = σ 2( K( x )) timing risk for the strategy xk. Based on cost
that of volume are indeed
correlated. This means profiles, comparisons of several strategies
that the presence of high
volume does not indicate the
Now that we know how to compute both are easy to make.
direction of price movement. price and liquidity risk, and assuming
that volume and price movement are Consider a numerical example for an order
independent,25 we forecast the total timing to be executed and two possible strategies.
risk for a given trade schedule through a We show below how we determine the cost
simple combination of them, that is: profile of each strategy and how we can
compare the findings.
2
⎛ Ii ⎞ xij4σ 2( vij )
ℜ( xk ) = ∑ rj Crj + ∑ ∑ ⎜ ⎟
'
Let us first suppose a VWAP strategy
j ⎝ X i ⎠ 4( xij + 0.5v ij )
4
j i
for an order of 120,000 shares of XYZ
stock. The market price is currently
It must be said that, in practice, the equal to €80 and is expected to reach
estimation and incorporation of liquidity €82 at the end of the trading day. Daily
risk into the timing risk term is complicated. volatility is estimated at 120bp. The
It is for that reason that the timing risk of execution strategy and expected market
a strategy is often referred to only as the conditions are summarised in table 3 and
price risk of the strategy. the instantaneous market impact cost for
the stock is estimated at €150,000.
The estimation models and techniques
for price appreciation, market impact


Pre-Trade Analysis
Table 3: Expected market conditions over the trading horizon Table 5: Expected market conditions over the trading horizon
Expected volume VWAP strategy 10% participation
Trading Expected volume
Trading strategy
period Executed Residual
Buy Sell period
shares shares Executed Residual
Buy Sell
shares shares
1 200,000 200,000 24,000 120,000
1 200,000 200,000 40,000 120,000
2 150,000 150,000 18,000 96,000
2 150,000 150,000 30,000 80,000
3 100,000 100,000 12,000 78,000
3 100,000 100,000 20,000 50,000
4 50,000 50,000 6,000 66,000
4 50,000 50,000 10,000, 30,000
5 50,000 50,000 6,000 60,000
5 50,000 50,000 10,000 20,000
6 100,000 100,000 12,000 54,000
6 100,000 100,000 10,000 10,000
7 150,000 150,000 18,000 42,000
7 150,000 150,000 0 0
8 200,000 200,000 24,000 24,000
8 200,000 200,000 0 0
To determine the cost profile for this VWAP Table 6: Cost profile of the 10% participation strategy
strategy, we must compute the expected Implicit costs Risk
price appreciation cost, market impact cost PA* MI** Costs***
and timing risk. For the latter, we focus on linear growth
price risk alone. The final results are presented Total
77.500 76.972 30.350 107.850 53.343
in table 4 and some helpful details for each euros (€)
Euros/
calculation are noted below. share (€/ 0.65 0.64 0.25 0.90 0.44
share)
Table 4: Cost profile of the VWAP strategy
Basis
Implicit costs Risk points 81 80 32 112 56
PA* MI** Costs*** (bp/share)
linear growth * price appreciation,

Total ** market impact,
135,000 134,607 22,768 157,768 70,310 *** total costs with the linear model
euros (€)
Euros/
share (€/ 1.13 1.12 0.19 1.31 0.59 Comparing tables 4 and 6, we observe
share)
now that θbp ( VWAP ) = (164.73) and
Basis
points 141 140 24 164 73
θbp (10%) = (112.56 ) . So, the second strategy
(bp/share) is clearly less costly and risky than the first. In
* price appreciation,
fact, given expected market conditions, the
** market impact, 10% participation strategy offers the lowest
*** total costs with the linear model
risk and cost because it completes trading
Note: As for price appreciation, the linear model delivers a constant at the sixth period and does not expose the
price change of 0.25 per trading period while the growth model
provides an exponential price change of 0.0031. Concerning timing
order to adverse price movements over the
risk, the stock variance per trading period (1,800bp) is estimated entire trading day.
from daily volatility. To express it in monetary units, we multiply
it by the square root of the current stock price.
(1) Optimisation formulation
With the same method, we are also able to In uncertain market conditions, the
determine the cost profile of an alternative preservation of asset value is essential
strategy such as a 10% participation strategy for most investors and we can define the
(see tables 5 and 6). ultimate goal of implementation as asset
value preservation in the presence of risk.


Pre-Trade Analysis
This definition means that we need to This constraint ensures that optimisation
consider cost and risk jointly. Accordingly, will provide a strategy that executes orders
optimisation must be viewed as a means over the specified trading horizon.
of determining the trading strategy that
resolves the trader’s dilemma, which will • Shrinking portfolio: ri , j ≤ ri , j −1
find the suitable trade-off between cost and This constraint ensures that the portfolio
risk. Indeed, market impact and timing risk is continuously decreasing in shares and
are conflicting terms; an optimal strategy prevents optimisation from making the
results therefore in finding the proper trade- position either longer or shorter than the
off. initial position.
x i ,k
Broadly, optimisation in the context of • Participation rate: ≤α
x i ,k + v i ,k
trading can be formulated in three ways:
• Minimise cost subject to a specified level This constraint places an upper bound on
of risk, that is: participation volume in each period.
Min φ( xk ) s.t. ℜ( xk ) ≤ ℜ * ; ℜ * is
the maximum risk exposure and φ( xk ) • Cash balancing
the expected transaction cost. ⎡ 0.95 xi ,k
• Balance trade-off between cost and risk, −Dm in
≤ ∑ ∑ x i ,k ⎢ pi + k .Δpi +
X i ( xi ,k + 0.
that is:
i k
⎣⎢
Min φ( xk ) + λ.ℜ( xk ) ; λ is the level⎡of risk 0.95 xi ,k Ii 0.05Ii ⎤
aversion or the marginal−D ≤ ∑ ∑ x
m in rate of substitution
i ,k ⎢ p i
+ k .Δp i
+
X i ( xi ,k + 0.5vi ,k )
+
X i ⎦⎥
⎥ ≤ Dm a x
i k ⎢
⎣
between cost and risk.
• Maximise probability of price improvement, This constraint is usually used in the
that is: implementation of a trade list including buy
Max Pr[ φ( xk ) ≤ L*] ; L * is the highest and sell orders to ensure that the net cash
acceptable cost. position in any period is within a specified
range. In the above approach, the objective
Specifically, we can illustrate the second function calculates the total cost by adding
formulation with the following proposal, price appreciation and market impact plus
which incorporates some real-world λ units of timing risk. λ=1 indicates that
constraints, from Kissell and Glantz investors are equally concerned with cost
(2003): and risk. λ >1 (<1) refers to investors who
Objective function are more (less) concerned with risk than cost
m n m n ⎡ 0.95I andx who thus0.05I prefer⎤ aggressive n (passive)
Min ∑ ∑ k .xi ,k Δpi + ∑ ∑ xi ,k ⎢ ∑
i i ,k
strategies. So, + for ai ⎥given
+ λ. risk raversion,
k
'
Crk the
i =1 k =1 i =1 k =1
⎣⎢ X i ( xi ,k optimisation
+ 0.5vi ,k ) X i ⎦⎥ k =1
output includes the number
of shares of each security to be traded in
m n m n ⎡ 0.95Ii xi ,k 0.05Ii ⎤ n
∑ ∑ k .xi ,k Δpi + ∑ ∑ xi ,k ⎢ +
X i ⎦⎥
⎥ + λ. ∑ rk 'Crk each period (trading schedule) as well as
i =1 k =1 i =1 k =1
⎣⎢ X i ( xi ,k + 0.5vi ,k ) k =1
cost estimates for each component (cost
profile).
Constraints n (2) Efficient trading frontier
• Completion: ∑ x i ,k
= X i The approach described above is in
k =1


Pre-Trade Analysis
the same vein as Markowitz’s portfolio period. However, we can observe situations
optimisation framework, a very popular in which the incremental price appreciation
tool from classic finance theory. It shows cost begins dominating the market impact
the trade-off between risk and return in cost reduction, resulting in an increase of
an investment portfolio: to achieve higher total transaction costs. It is for this reason
returns, investors must take on more risk. that optimisation with negative risk aversion
The same is true for the question at hand: is not really relevant for investors.
to incur lower transaction costs, the trader
Figure 11: Trading frontier
must implement a riskier strategy. As in
portfolio theory, this takes us to the notion
of an efficient frontier.
The notion of a trading frontier is explored

in several academic and professional works
that deal with execution optimisation.26
It can be defined as the set of strategies
that contain the lowest cost for the
26 - See, for example,
Almgren and Chriss (2000)
specified level of risk, in which strategies
and Nevmyvaka et al. (2005). are determined through optimisation for Like the efficient frontier in portfolio
all possible values of λ. Figure 11 shows a theory, the efficient trading frontier (EFT)
trading frontier. The cost curve is a convex is the set of all the dominating trading
function in which each points corresponds strategies, that is, the set of strategies that
to a unique degree of risk aversion (λ). contain the lowest cost for the specified
This function decreases over the range risk and the least risk for a specified cost.
0 ≤ ℜ ≤ t and increases when 0 ≤ℜ≤ >tt. It is quite easy to see in figure 11 that only
Strategy C (θ(z,t)) corresponds to the strategies on the curve to the left of C
minimum cost of the frontier. Based on the are optimal. Although A and B exhibit the
above optimisation equation, this minimum same cost level (y), only A is a true optimal
cost will occur at frontier. Based on the strategy because it contains less risk for
above optimisation equation, this minimum the cost (s<u). Hence, only strategies lying
cost will occur at λ=0 since risk is excluded. on the ETF are rational trading decisions.
For positive values of λ, we obtain points Any other strategy is irrational because at
with higher costs and lower risk than at the least one alternative strategy can always
minimum cost point. For example, strategy be found with less risk for the same cost,
A (θ(y,s)) is on the curve to the left of C lower cost for the same risk, or with both
since s<t and y>z. By contrast, points on lower cost and risk. Consequently, the ETF
the curve on the right of C are associated is a valuable means of identifying the cost
with negative values of λ. It is the case for profile of optimal trading strategies and of
strategy B whose cost profile is θ(y,u). An choosing the most appropriate execution
important element here is that when λ<0 (in view of the investor’s preferences or
the optimisation is minimised at the value final objectives).
where the quantity of risk is maximised.
This corresponds to a strategy wherein
trading is spread over as long as possible a
V. Trading Performance
Measurement

V. Trading Performance Measurement
The last task assigned to TCA is the performance assessment. As we have seen,
assessment of trading performance. Like transaction cost measurement attempts
transaction costs, trading performance to determine the source and magnitude
must be measured after the execution of of trading costs. Trading performance
the trade. Here, it is crucial to understand assessment, by contrast, looks into the
clearly how evaluation of trading justifications for these costs and seeks
performance differs from transaction cost to identify any costs caused by poor
measurement, although both are done ex implementation decisions. In other words,
post. performance evaluation attempts to
assess the performance of intermediaries
when they execute trades. The ultimate
1. Transaction Costs vs. Trading aim is to determine the most effective
Performance intermediaries by market, instrument and
Even today there is widespread confusion trading strategy. With this information
of transaction cost and trading at hand, investment managers can
performance measures; this confusion considerably reduce the time required
leads to the frequent industry adoption to select the best intermediary for the
of inappropriate practices. Although the execution of a specific trade.
main reason for this ongoing confusion
is undoubtedly the elusive and sometimes Given the above objective, a good trading
controversial notion of best execution performance measure should help investors
(we will come back to this important determine if the actual transaction costs
notion), it must be said that some in the incurred were reasonable under the
industry also contribute to its currency. market circumstances that prevailed at
Some intermediaries, for example, the time of the trade. Proper performance
have incentives to argue that trading evaluation must therefore assess the real
performance evaluation is extremely capability of intermediaries and, to a
complex and that it should be founded on larger extent, determine if they add value.
market impact that cannot be observed or Distinguishing between skill and luck can
measured. They can thus consistently come make it possible to determine if execution
up with arguments, some more relevant can indeed be termed best.
than others, for declaring benchmarks
or peer groups unfit for assessing the
quality of their trades. In this way, they 2. Current State of the Industry
avoid being subjected to performance and Practices
assessment. In addition, some consultants The most common performance
have a vested interest in this confusion, measurement practice in the industry is the
since it justifies their lucrative business. benchmark comparison described above.
In any case, effective transaction cost So, here we address its shortcomings as a
management would require clarification. gauge of trading performance. Next, we
will introduce the relative performance
Post-trade analysis consists of two parts: measure developed by Kissell and Glantz
transaction cost measurement and trading (2003). Although this measure is not

widely used, it appears to be an improved given time period. The belief behind these
metric that is more consistent and stable benchmarks is that you did a good job when
because it allows comparisons across you did better than the “average trader” of
securities and across days. Furthermore, the day or period. In this sense, intraday
this approach shares the principle of peer benchmarks are the most consistent with
group review with the innovative and a performance measure, assuming that
systematic framework for assessing the the benchmark is a good indicator of
quality of execution that we propose and the fair market price. However, although
will develop further in the document. intraday benchmarks account for market
conditions and trading activity, they do
(1) Benchmark comparison not provide a meaningful metric that is
This approach involves comparing the consistent across days and across stocks.
monetary value difference between We will look into this key problem later.
the executed position and the position
evaluated at some benchmark price. For The post-trade benchmarks (T close, next
a given transaction, the monetary value mid bid-ask) attempt to measure the
per share is given by the signed difference market impact of trades. When used in
between the average price obtained for the trading performance analysis, they serve
trade and the benchmark price. A positive more as a measure of investor skill than as
value indicates an execution that is less a measure of trader performance. For funds
favourable than the benchmark price, that track a benchmark index, post-trade
while a negative value indicates execution benchmarks indicate the contribution of
more favourable than the benchmark. As execution to total tracking error.
we have mentioned above, the benchmarks
usually used include pre-trade, intraday (b) Major shortcomings
and post-trade reference prices. As a whole, when used to assess trading
performance, benchmark comparison has
(a) Critical review of benchmarks major shortcomings that make it inefficient
The pre-trade benchmarks (T-1 close, T and/or misleading. These shortcomings are
open, last trade price, last bid, last ask) listed and described below.
provide transaction cost measures, not
performance measures. They deliver good Benchmark comparison may be biased
proxies for the costs of spread, market The benchmark comparison may provide
impact, operational and market timing biased performance measures that
delays, but they do not provide any insight essentially depend on the reason for the
into actual execution performance since trade. According to Harris (2003), biases
they do not help gauge whether the costs may arise when trading decisions depend
incurred are acceptable given market on past price changes or when investors
conditions at the time of trading. are well informed about future price
changes. Some benchmark prices can
The intraday benchmarks (VWAP, LHOC) deliver performance measures that will be
compare the execution price and a kind of systematically good or bad depending on
average market price of the day or of a whether the investor uses a momentum

or contrarian strategy. For example, liquidity and never take it. Hence, they
the opening price delivers performance always buy at the “bid” or sell at the “ask”.
measures that are easily biased. Momentum This behaviour makes their performance
investors will push down the trading look great but comes at the expense of
performance measure because they buy increased timing risk.
(sell) when prices have risen (fallen), so the
opening price is low (high). By contrast, Benchmark comparison needs a unique
contrarian investors buy (sell) when prices and relevant reference price
have fallen (risen), so the opening price Benchmark comparison assumes that the
is high (low). This pushes up the trading benchmark is an appropriate price for
performance measure. the value of the security. The only “true”
value of a security is the price at which
Benchmark comparison may be gamed an actual trade was made. Sometimes,
Gaming problems arise when the benchmark is not such a price. For
intermediaries know the benchmark that example, the spread midpoint, the VWAP
will be used to measure their execution or the LHOC may be prices at which no
performance and consequently time their trade has taken place. Furthermore,
trades based on their evolving benchmark market fragmentation and proliferation of
score. liquidity pools make the determination of
the right benchmark more difficult. Indeed,
The consideration of trade-timing is the benchmark comparison becomes
important for investors who give their conceptually difficult in the absence of a
intermediaries discretion over the timing recognised unique price for the securities
of their trades. Proper trading performance traded. When a security is traded at the
measures should help investors ensure that same time on various execution venues,
their brokers make appropriate use of this which of the coexisting prices is the
discretion. Trade-timing effects are best benchmark price for the security? With
measured when the benchmark does not MiFID around the corner, this issue is of
depend on the time of the trade. When the great importance. By allowing multiple
benchmark relates to the execution time, trading venues to compete, this new
intermediaries can accelerate/decelerate piece of European regulation is, by design,
trades based on their evolving benchmark allowing liquidity pools to fragment and
score. This behaviour is questionable putting the very existence of recognised
because it can result in higher total benchmarks in danger.
transaction costs for the end investor.
Benchmark comparison is not
Harris (2003) describes several situations in standardised
which brokers can game their evaluations. Finally, for at least two reasons, benchmark
For example, brokers who have discretion comparison does not offer a unified
over how aggressively they fill orders can framework to enable easy assessment of
easily game a performance measure based execution performance across a series of
on the spread midpoint prevailing at the trades at any aggregate level.
time of the trade. They will simply supply

First, benchmarks provide absolute its average execution price. Modelled after
measures of costs, which are often the percentile ranking, the RPM attempts
expressed in basis points. Determining to provide a descriptive and meaningful
whether the trading performance is bad, measure of trading performance that is
normal or good is not straightforward and consistent across days and across stocks.
benchmarks do not make comparison easy Not unlike VWAP-based measures, in that
because they deliver inconsistent metrics. it properly accounts for contemporaneous
For example, a measure of 10bp could market conditions and trading activity,
mean good performance in one security the RPM is more robust because it allows
but bad performance for another. Similarly, multiple comparisons.
it could also mean good performance
today but bad tomorrow. Next, existing (a) Calculation and interpretation
measures are built on benchmark prices The RPM is built on a combination of both
that often depend on trade characteristics. market volume and number of trades,
Consequently, using only one benchmark which makes it possible to account for
price to analyse the execution quality of a the potential skew of large block trades
universe of trades can be criticised, unless at extreme prices. Accordingly, if P* is the
27 - The domination of
the sell-side in the TCA
all the trades take place at the same time average execution price of the order and
services provision is also and are of a similar size and difficulty. Pi is the price of the ith trade, the RPM is
problematical. TCA tools and
services have been at the calculated as follows for a buy:28
heart of the development of
The absence of a standardised framework
the sell-side industry over 1
the last five years. With the has led to the absence of consensus about RPM = ⎡RPMtra des + RPMvolum e ⎦⎤ , with
multiplication of electronic
a universal indicator and accounts for the
2⎣
trading capabilities (direct
market access or algorithmic
availability in the industry of a great many *
trading), brokers have
endeavoured to provide indicators. As a result, investors are often RPMtra des =
∑ trades Pi > P
and
substantial tools for their
clients to help them measure confused.27 Furthermore, from the absence ∑ trades
transaction costs and, to a
of consensus emerges the difficulty *
larger extent, assess trading
performance. Created as of comparing trading performance RPMvolum e =
∑ volume Pi > P
independent boutiques,
most TCA providers were and execution quality across different ∑ volume
quickly acquired by the major
brokerage firms so that they
intermediaries. The heterogeneity of trades
could become a significant and diversity of measures complicate any Graphically, we present the RPM by plotting
and visible element of the
sell-side value proposition. attempt at comparative analysis of trading the percentage of activity traded at a
That the sell-side has totally
taken over the market for
performance. price less favourable than or equal to the
TCA confirms the strategic average execution price. For this purpose,
importance of this function
for the industry. (2) Relative performance measure activity needs to be sorted from lowest to
28 - The RPM calculation is
symmetric for a sell order
The relative performance measure (RPM) highest prices for sells and, by contrast,
(Pi<P*). is a metric proposed by Kissell and Glantz from highest to lowest prices for buys.
(2003). Founded on peer group review, it Figure 12 illustrates the RPM calculation
compares the average execution price of for a sell order with an average execution
the trade to all market activity over the price of €31. The figure shows an RMP
trading period. For a given trade, the RPM is of 80% for the order, meaning that 80%
the percentage of all activity in the market of all market activity was traded at prices
that traded at a price less favourable than lower than €31. The interpretation is that

the trader did a good job since 80% of who exhibit little deviation in performance
market activity traded at less favourable measures are consistent, while those who
prices than he did. If the order had been often have extreme scores are gambling or
a buy, all market activity higher than €31 taking risks.
would deliver a RPM of 20%, indicating
poor performance. (b) Weighted RPM
When intermediaries have discretion over
Figure 12: Relative performance measure
how they execute orders within a specific
horizon, the RPM based on all market
activity traded in that horizon reflects
their ability to work orders and seize
the best available trading opportunities.
However, when the trader is given a
predefined execution strategy, using the
RPM over the entire trading horizon may
provide noisy performance measures
when the instructions or constraints
29 - We have a similar
phenomenon on the sell
The RPM is superior to any other set by the investor are not neutral. This
side. An aggressive strategy benchmark-based performance measure is the case when they force traders to
in a falling (rising) market
will deliver a high (low) RPM because it provides a meaningful metric execute the majority of shares at times
for sells, suggesting good
(poor) performance. A passive
that is consistent across days and across of the least or most favourable prices.
strategy will result in the stocks. Suppose, for example, that a For example, an aggressive strategy in a
opposite.
trader’s performance measure built on a falling (rising) market will refer to a low
benchmark is 15bp in stock A and 40bp (high) RPM for buys, suggesting poor
in stock B. As we mentioned earlier, this (good) performance.29
information cannot help determine if this
was good or bad performance or make any To avoid noise and really distinguish
comparisons across stocks. By contrast, if between trading performance and investor
the trader has an RPM of 20% in stock constraints, the RPM must be adjusted
A and 75% in stock B, we can easily upon the specified strategy as follows:
determine that the trader did badly in xj
stock A but very well in stock B. This means RPM * = ∑ RPM j ,
j X
that the trading performance was better
in stock B than in stock A, even though where RPMj is the RPM computed over
the trader has a better benchmark score in the trading period j, xj is the quantity to
stock A. These conclusions are meaningful execute in period j and X the total order
and easy to reach with the RPM since it is size. This weighted RPM attempts to
consistent over time and across stocks. provide insight into the quality of prices
obtained by the trader during the times he
Furthermore, using a sufficient number is requested to trade.
of observations and analysing the
distribution of RPM facilitates a reading (c) Measure of value-added
of the consistency of intermediaries. Those When intermediaries deviate from

instructions or constraints set by the It is easy to see that, when traders follow
investor (on purpose or not), the final result the prescribed strategy, the value-added is
may be higher or lower total transaction zero. When they deviate, a positive value-
costs. In such situations, what is important added indicates good market timing ability
beyond the RPM is to determine whether (the trader achieves better prices and lower
the trader’s decision has added value costs) and, by contrast, a negative value-
(lower costs) or hurt overall performance added means poor market timing ability
(higher costs). Let us take an example to (the trader achieves worse prices and higher
illustrate this concern. costs). The VA metric thus assesses the real
capability of intermediaries and contributes
Suppose a trader who is requested to to distinguishing between skill and luck.
execute a given order before midday
but decides (for whatever reason) to A step further is to assess the overall
work it through the entire day. Although contribution to cost attributable to the
he minimises market impact cost, the trader’s market timing. This could be
investor incurs higher total transaction done by plotting together on a chart the
costs because of adverse price movement. normalised difference between the actual
30 - For periods where
xj=0, the RMPj is computed
It is possible for the trader to get a and the expected transaction costs and
assuming that the trader favourable RPM but it will not give any the VA of the trader.
achieves the average price in
the period. insight into the effect of his decision on
final performance. (d) RPM vs. VWAP
The RPM is somewhat similar to VWAP-
To address this issue and quantify the based measures since both properly
value-added by the trader’s own market account for the prevailing market conditions
timing, Kissell and Glantz (2003) define by comparing the average execution price
a new metric (VA) to calculate the and all other contemporaneous market
percentage of the total RPM attributable activity. In that sense, both measures
to the deviation decision, that is: suffer from the same shortcoming. When
RPM( x * ) − RPM( x )
∑ ( x j −anx j )RPM
*
order j accounts for the main part of
VA = =
j
market activity in the actual trading period,
RPM( x * ) RPM(itsx * average
) execution price converges to
the VWAP computed over that period and
RPM( x ) − RPM( x )
* ∑ ( x *j − x j )RPM j
the RPM converges to 50%, suggesting
j
VA = = ,
RPM( x * ) RPM( x * ) average performance. In this case,
interpreting these results becomes quite
where RPM(x) is the RPM that the trader hard since the execution being analysed is
would have got had he followed investor its own reference.
instructions/constraints and RPM(x*) is
his actual RPM.30 So, by comparing the The main advantage of the RPM over
deviation in strategy, we can assess the any VWAP-based indicators is, as already
value that the market timing of the trader mentioned, that the RPM is more robust
adds to the implementation. because it allows multiple comparisons,
i.e., across securities and in the same

security across days or periods. This is (1) Best execution, this elusive
due to the consistency of the RPM, which concept
makes comparison easy. A performance Best execution means many things for
of 75% means exactly the same thing many people, but everybody agrees that
whatever the day, the security or even it includes at least trading performance
the trader. Furthermore, unlike the assessment. In theory, best execution
RPM, the VWAP exhibits an asymmetric is often defined as a measure of how
distribution since most market activity well investors’ trades are executed. This
is likely to be on one side of the VWAP. definition is vague as it encompasses
This means that 50% of prices and volume components such as the trade price, the
are not higher or lower than the VWAP. execution speed, the opportunity for
We can thus observe situations in which price improvement, the probability of full
intermediaries underperformed the VWAP execution, the respect of anonymity and
but still outperformed the majority of the level of explicit costs. Although the
market activity. In those cases, the VWAP total charges paid by the investor are still
scores may lend themselves to misleading the main dimension, the others are not to
interpretations. be ignored.
Everyone admits that the lowest

3. Regulatory Pressure: MiFID and transaction cost does not necessarily imply
its Best Execution Obligation best execution, but the consideration of
The absence of a standardised framework all components intensifies debate when it
for easy assessment of trading performance comes to checking whether best execution
becomes problematic with the arrival of has been achieved, essentially because
the MiFID. Indeed, with this new piece of some of these components are not easy
European regulation, investment firms are to measure and/or combine. Either the
going to enter an environment where they information is not directly available
will have to demonstrate that they have (execution speed, price improvement) or
executed at the best possible conditions the component itself is not measurable
for their clients while taking into (anonymity is respected).
consideration potential multiple liquidity
pools. To address this issue, we will first According to Kissell and Glantz (2003), best
focus on the concept of best execution in execution can be categorised as price, time,
general. We will see that nowadays this and size factors whose importance varies
concept is very fashionable in the industry, with the goals and objectives of investors.
although it is not always well understood For example, value and passive investors
and does mean not the same thing to are concerned mainly with price factors,
everyone. Then, we will move on to the growth and momentum investors focus
MiFID best execution obligation and show on time factors, while large investors and
that the regulator has provided neither a mutual funds pay special attention to size
clear definition nor a measurable objective factors. In this respect, Kissell and Glantz
to make up for the current absence of (2003) define best execution as “the whole
consensus in the industry. process of managing transaction costs

throughout all phases of the investment (2) MiFID best execution obligation
cycle to ensure that the portfolio realizes This new rule is a crucial element of investor
its highest returns possible”. protection and must be viewed as the
natural counterpart of a full liberalisation
Even if the above definition is not yet of the marketplace. This obligation is
widely recognised in the industry, we found in the now famous Article 21, which
strongly believe that it is the right has been at the heart of the arguments
way to clear up most of the confusion put forward by many opponents of
surrounding best execution. Consistent MiFID. After having summarised the main
with this definition, Kissell and Glantz details of the provision, we will show
(2003) claim that checking whether best that although MiFID puts the spotlight
execution is achieved results in analysing on best execution, it has not provided a
the entire transaction cost management clear definition or a measurable objective
process. This thorough check thus requires that would make it possible to determine
a complex post-trade procedure that whether trades have been executed in the
should consist of the following steps: best possible fashion.
1. Evaluate the implementation decision:
31 - Article 21 states,
however, that whenever there
Is the trading strategy optimal? Does it lie In the 2004 Directive, the best execution
is a specific instruction from on the ETF? obligation has been defined as an obligation
the client the investment
firm shall execute the 2. Measure transaction costs actually of means whereby investment firms are
order following the specific
instruction.
incurred: Where and why did they occur? required to have taken all reasonable
3. Estimate transaction costs ex post: What steps to obtain the best possible result
are cost estimates with known market for the client. This obligation is therefore
conditions? structured around three major principles:
4. Compare actual transaction costs with (i) an obligation of means to achieving the
ex-post cost estimates: Is there over/ best possible result for the client, involving
underperformance? factors that determine whether or not this
5. Measure execution performance: Has best possible result has been achieved;31
the intermediary executed at fair market (ii) documentation of an execution policy
prices? Did he add value? that includes the selected execution venues
6. Compare the cost difference with the and documentation of the parameters that
value-added: Is over/underperformance justify this selection;
related to superior/poor execution or to (iii) an obligation for investment firms to
favourable/adverse price movement? demonstrate, at the demand of the client,
7. Record performance of intermediaries that execution has been carried out in
accordance with the agreed execution
Although this procedure looks attractive policy and that the execution policy allows
and quite complete from a theoretical achievement of the best possible result on
standpoint, it is rarely applied, as debate a consistent basis.
on the definition of best execution and/
or trading performance measurement is MiFID considers best execution in the
ongoing. context of client categorisation and in a
tiered manner. Accordingly, the assessment

of best execution for professional clients in what the best execution obligation is
must consider factors that include not only able to achieve with regard to the fair
fees but a wide range of elements that can protection of all parties.33
be used to qualify the quality of a trade,
namely “price, costs, speed, likelihood of
execution and settlement, size, nature
or any other consideration relevant to
the execution of the order”. Next, retail
clients are said to have achieved best
execution when they have received the
best price net of expenses. This means
that the net price should usually be the
most important element; the other factors
play a limited role. Finally, trades executed
between firms categorised as “eligible
counterparties”32 fall outside the scope of
the best execution obligation.
32 - This category refers to
the most experienced firms
operating in the marketplace, Begun as an obligation of result in a
such as professional trading
firms and asset managers principle-based regulatory approach,
executing client orders.
33 - For more detail on
the best execution obligation has been
MiFID and its best execution actively fought by industry representatives
obligation, see D’Hondt and
Giraud (2007). and has slowly turned into a more modest
obligation of means that remains complex
and ambiguous. Furthermore, as a direct
result of the underlying complexity of
defining best execution, the regulator has
shifted the focus to the means rather than
simply to the result. By providing only
partial guidelines on how investment firms
should assess the quality of execution
through a double system of criteria and
factors to be taken into consideration,
but without determining how they will
actually work, the regulator has frustrated
all industry participants, with those
seeking a principle-based approach left
only with an overly prescriptive rule and
those seeking the definition of precise
criteria for complying with the obligation
left with too much flexibility in the
application of Article 21. This situation has
led to significant resistance and mistrust

VI. A New Framework:
the EBEX Indicators

the EBEX Indicators
In the post-MiFID environment, the role trade-timing consideration. EBEX allows the
of TCA is set to take on even greater investor to determine easily whether or not
importance. As we have seen, however, the trading performance of his intermediary,
the current absence of a standardised and trader or even algorithm is consistently at
widely recognised framework for assessing the top of the class.
the quality of the entire trading process is
problematic. In this section, we try to fill this EBEX is being actively discussed by
gap first by offering a unified framework for professionals and academics, and we are
measuring ex post the quality of execution confident that positive developments will
and then by showing how this framework be proposed in the very near future to allow
can be incorporated into the analysis of it to cope with specific situations that the
the entire transaction cost management first version did not make allowances for,
process. such as investor constraints or the cost
for the investor associated with the broker
Because benchmark comparison suffers from performance.
several shortcomings, we have developed a
standardised framework in the same vein as After having exposed the objectives and
34 - We became aware of the
existence of the RPM after
the RPM34 and have opted for an absolute principles behind this new methodology, we
having defined our own tools. measure of the quality of the price obtained: will describe in detail how our indicators are
a score between 0 (bad performance) and built, as well as how they can be interpreted.
1 (good performance). Then, to illustrate both the framework and
the level of interpretation made possible,
Our approach, the EBEX framework (EDHEC we will present the results of an empirical
Best Execution), is founded on indicators that study conducted on a sample of orders for
facilitate comparisons of a large universe of Euronext blue chips. Next, we will compare
trades and provide insightful information EBEX and other approaches to highlight the
not only about final performance (the advantages of our framework.
absolute EBEX indicator) but also about
the possible justification of the performance
(the directional EBEX indicator); a measure 1. General presentation
of the quality of the market timing is thereby Our approach aims to provide a simple
provided. answer to the following question:
EBEX has several advantages over current Given a transaction handed over to a broker,
industry practices. It is very simple, provides trader or algorithm and executed for a given
a standardised framework for assessing the price at times that are recorded under given
quality of execution across a series of trades time constraints, to what extent have other
aggregated at any level, and complies with brokers, traders or algorithms executed
the MiFID requirement to demonstrate that comparable volumes to this transaction,
the target has been reached. It also delivers either before or after this transaction, at
absolute and meaningful measures of a better price?
execution quality, allowing straightforward
and objective interpretation, and includes


the EBEX Indicators
The answer to this question can be split This approach is built on two indicators.
into four important elements: The Absolute EBEX indicator measures the
• The time at which the order is handed over quality of execution in a peer group review.
(release time) to an intermediary (a broker, The Directional EBEX indicator identifies
a trader or an algorithm) is the first point whether the broker, trader or algorithm has
of reference, while the time at which the implemented the execution too slowly or
order is entirely filled (execution time of the too aggressively. In other words, the first
last lot referring to the order in the event of indicator assesses the quality of execution
splitting) is the second point of reference. itself while the second indicator brings
The third and final point of reference is information about why the quality of
the next market close if the intermediary execution is as observed.
has discretion over the day to execute the
trade or, in the event of specific instructions At this stage, we can identify similarities
from the investor, the time at which the as well as differences between the EBEX
intermediary must stop trading. framework and the RPM. Both involve
• The size of “competing trades” is not peer group review and compare the
important as such; the relevant measure average execution price of the trade to all
is how many times a volume comparable contemporaneous market activity. Although
to the order has been executed at a better they have the same philosophy, the EBEX
price, which is a first measure of the quality framework makes it possible to gain quick
of the price obtained. The price must be insight not only into the final trading
compared to small trades executed at better performance (the absolute EBEX indicator)
prices (the broker, trader, or algorithm could but also into the possible justification of the
have split the order better) as well as with performance (the directional EBEX indicator),
larger trades (the order could have been thereby providing a direct measure of the
grouped with a larger flow of orders to be quality of the market timing. This point is
executed in block if such trading capability very relevant, as we know that pre-trade
is offered). This important point makes our analysis should help intermediaries predict
approach different from the RPM, which is likely market conditions and define an
built on a combination of both the number appropriate trading strategy when they
of trades and volume. have discretion over how they fill the order.
• Volumes traded before at a better price Getting the best price in the market may not
allow one to assess whether the broker, always be a realistic objective, but execution
trader or algorithm has been too patient. at fair market prices certainly is. What is
• Volumes traded after at a better price allow relevant for an investor then is to know
one to assess whether the broker, trader or where, with respect to the industry average,
algorithm has been too aggressive. the orders executed by his intermediary
fall.
EBEX relies on the elements above to measure
the quality of execution as part of a peer
group review and to identify whether the
broker, trader or algorithm has implemented
the execution too aggressively or too slowly.


the EBEX Indicators
2. Detailed Presentation of the The absolute EBEX indicator for order i is

Indicators calculated (first for a buy, then for a sale)
Our two fundamental indicators rely on as follows:
the same method and are easy to compute
and interpret. We describe in depth how
they are built and highlight their multiple
advantages over the benchmark comparison
approach. Next, we present another related
indicator whose purpose is to associate the
intermediary performance score with a cost
for the end investor. We finish by describing
how our indicators can be adjusted to
account for specific trading instructions
and constraints.
(1) Absolute EBEX

The absolute indicator of EDHEC Best
35 - In the event of specific
instructions from the investor, Execution for an order is the difference
the measurement window between one and a ratio between the In both equations, each element is defined
ends at the specified time
the intermediary must stop aggregate volumes traded at prices better as follows:
trading.
36- Si can be simplified but is than the average trade price obtained for • EBEXabs,i is the absolute best execution
shown for clarity. the order divided by the order size and the indicator for order i during the day
aggregate volumes without consideration • day is the interval between the time the
of price divided by the size of the order. As broker receives order i and the next market
illustrated in figure 13, the measurement close
window for the ratio corresponds to the • Si is the size of order i 36
period over which the intermediary has • APi is the average trade price obtained
discretion to execute the trade, that is, for order i
from the time the broker receives the order • N is the number of trades at a price better
(release time) to the next market close.35 than APi during the time interval
•  is the size of trade n at a price
Figure 13: Time window for the Absolute EBEX higher (lower) than APi during the interval
day
Trade Order Order Market
decision release execution close • M is the total number of trades during
the time interval day; M ≥ N
Absolute EBEX •  is the size of trade m during the
time interval day
Given the way it is built, the absolute EBEX

indicator can only take values between
zero and one. Interpretation is thus greatly
facilitated, as shown in figure 14 below.


the EBEX Indicators
Figure 14: Interpretation of the Absolute EBEX indicator still depend on market conditions as well
The closer to 0 EBEXabs is, the worse the execution! as price volatility. If the price range was
Among all the "similar" trades of the interval, most got a price better less than 40bp yesterday, a value of 30bp
than the trade price of the broker. A terrible performance!
indicates that the order got one of the least
favourable prices of the day. Furthermore,
0 1
if the price range today is much larger than
it was yesterday, a value of 40bp implies
better trading performance, even if the
The closer to 1 EBEXabs is, the better the execution! deviation from the VWAP is higher. So,
Among all the "similar" trades of the interval, few got a price better benchmarks do not make comparison easy
than the trade price of the broker. An excellent performance!
at all. By contrast, if the absolute EBEX is
0.65 today and was 0.45 yesterday, we can
The absolute EBEX indicator is superior to conclude that the quality of execution is
any other performance measure based on better today.
benchmark comparison because it provides
a meaningful metric that is consistent over The second advantage of the absolute
time and across securities. Two simple EBEX is that it makes possible evaluation
examples illustrate this advantage. of intermediary trading performance
consistence. This evaluation can be done
Example 1: Suppose two orders for which easily by using a sufficiently large sample
the benchmark comparison to the VWAP is of trades executed by the intermediary
15bp in stock A and 40bp in stock B. Both and analysing the distribution of absolute
executions are less favourable than the EBEX. Traders who exhibit little deviation
VWAP but one could be tempted to believe in performance measures are consistent
that the execution quality is higher in stock while traders who often get extreme scores
A, since 15bp is lower than 40bp. In fact, are gambling or taking risks. Moreover,
this way of thinking is misleading, as the comparative analysis across several
difference may be caused by differences in intermediaries may easily be conducted
market conditions, trading patterns, or stock based on absolute EBEX distributions wherein
prices. So, even if order size is comparable, the score of each intermediary is defined
the values themselves are not. By contrast, if as a risk-adjusted trading performance
we observe an absolute EBEX of 0.2 in stock computed as the mean value of EBEX divided
A and 0.75 in stock B, we can conclude that by its standard deviation. We illustrate this
the trader did a bad job in stock A but an approach with the following example.
excellent job in stock B. This means that the
trading performance was better in stock B Suppose two brokers who executed
than in stock A, although the trader has a approximately the same number of orders on
better benchmark score in stock A. similar stocks. According to their respective
absolute EBEX distribution, broker A exhibits
Example 2: Suppose a single stock for an average absolute EBEX of 0.65 for a
which the benchmark comparison to the standard deviation of 0.45 while broker B
VWAP is 40bp today but was 30bp yesterday. provides an average EBEX score of 0.55 for
Although the stock is identical, the values a standard deviation of 0.33. These figures


the EBEX Indicators
suggest that, on average, broker A performs and after that the trade was executed. The
better than broker B (0.65>0.55) but that directional EBEX indicator for order i is
he is also less consistent (0.45>0.33). computed as follows:
EBEX dir ,i = NBBEX i , j − NABEX i ,t

As a result of what we have just seen above,
the absolute EBEX indicator may be readily
used to set a clear performance objective NBBEXi,j and NABEXi,t are thus the
for any intermediary. Accordingly, an active components of the directional EBEX
trader will be expected to attempt to provide indicator. Their definition and computation
an average EBEX as high as possible with are very similar; only the measurement
the lowest standard deviation. For example, window of reference differs, as shown in
any investor could request his intermediary figure 15.
to reach a defined average absolute EBEX to
Figure 15: Time window for the Directional EBEX
justify the use of an active market timing
strategy. This performance objective would
Trade Order Order Market
be very relevant because the absolute EBEX decision release execution close
indicator cannot be gamed, as it involves NBBEX NABEX

37 - Or any other
intermediary.
post-trade peer group analysis. Directional EBEX = NBBEX - NABEX
Given these advantages, the absolute EBEX NBBEXi,j

method offers tools that comply with MiFID NBBEXi,j stands for Number of Before-Better
requirements. Indeed, under MiFID, traders Executions for order i over the time interval
must demonstrate that they have executed j. This component can be defined as a ratio
at the best possible price for their customers between the aggregate volumes traded at a
while taking into consideration potential price better than the average trade price of
multiple liquidity pools. For this purpose, it order i divided by the size of order i and the
is necessary to compare each trade with the aggregate volumes without consideration
universe of all the trades referring to the of price divided by the size of order i. This
same security and executed on all available ratio is computed over the interval j, which
trading venues. As our approach involves goes from the time the broker receives
peer group analysis and theoretically order i (release time) to the time order i is
includes as many trading venues as required, completely filled (execution time).
it is a suitable and relevant tool that should
play an important role in the post-MiFID The mathematical notations referring to
environment. NBBEXi,j are given below (first for sell orders,
then for buy orders):
(2) Directional EBEX N
Directional EDHEC Best Execution for an ∑V P > APi

n,j
n =1
order indicates how the broker37 could have (Si )
traded over time to provide better execution. NBBEX i,j = M
This indicator results from the combination ∑V m ,j

of two sub-indicators that measure the m =1
volumes traded at better prices both before (Si )


the EBEX Indicators
The mathematical notations referring to

N
NABEXi,t are given here below, for sell orders
∑V P < APi
n,j
and buy orders respectively.
n =1
(Si ) N
NBBEX i,j = M ∑V P > APi
n ,t
∑V m ,j
n =1
( Si )
m =1
NABEX i ,t =
(Si ) M
∑V m ,t
In both equations, each element is defined m =1
as follows: ( Si )
• NBBEXi,j is the number of better executions
for order i during the interval j N
• j is the interval between the time the ∑V P < APi

n ,t
n =1
broker receives order i and the time order ( Si )
i is completely filled NABEX i ,t = M
• Si is the size of order i 38 ∑V m ,t

• APi is the average trade price obtained m =1
38 - Si can be simplified but

for order i ( Si )
is displayed for the sake of
clarity. • N is the number of trades at a price better
39 - In the event of specific
instructions from the investor, than APi during time interval j In both equations, each element is defined
the measurement window
ends at the specified time
•  Vn P, j> ( < ) APi is the size of trade n at a price as follows:
the intermediary must stop higher (lower) than APi during the interval • NABEXi,t is the number of better executions
trading.
40 - Si can be simplified but j for order i during the interval t
is displayed for the sake of
clarity.
• M is the total number of trades during • t is the interval between the time order
the time interval j; M ≥ N i is completely filled and the next market
•  Vm , j is the size of trade m during time close
interval j • Si 40 is the size of order i
• APi is the average trade price obtained
NABEXi,t for order i
NABEXi,t stands for Number of After- • N is the number of trades at a price better
Better Executions for order i over the than APi during time interval t
time interval t. This component is a ratio •  Vn P,t > ( < ) APi is the size of trade n at a price
between the aggregate volumes traded at a higher (lower) than APi during the interval t
price better than the average trade price of • M is the total number of trades during
order i divided by the size of order i and the the time interval t; M ≥ N
aggregate volumes without consideration • Vm ,t is the size of trade m during time
of price divided by the size of order i. This interval t
ratio is computed over the interval t, which
starts at the time order i is completely filled Now that both components of the directional
(execution time) and ends at the market EBEX indicator have been presented, we
close of the day.39 can focus on how they can be interpreted
to characterise the timing of the trade.
This interpretation is very easy because


the EBEX Indicators
both range from 0 to 1, given the way had more opportunities to trade at a better
they are built. Figure 16 contributes to an price after the order was executed.
understanding of the interpretation.
A direct comparison of both indicators yields
Figure 16: Interpretation of the Directional EBEX components
our directional EBEX indicator, interpretation
Broker should
be more aggressive!
of which is even easier. The goal of this
1 indicator is to give information about how
the intermediary could have traded over
To be improved Poor
time to provide better execution. Given its
construction, the simple difference between
NBBEXi
Broker should
be more patient! NBBEX and NABEX, the directional EBEX
Best To be improved
indicator can range from -1 to +1. Figure
17 (below) summarises the interpretation
0 1 of the directional EBEX indicator.
NABEXi
NBBEX is close to zero when few traders Figure 17: Interpretation of the Directional EBEX indicator
execute at more favourable prices before
order execution. In this case, the quality
of execution may be said to be high. The
intermediary did a good job because few
traders obtained better prices. By contrast,
NBBEX is close to one when many traders
got more favourable prices before the
order execution time. Hence, the quality
of execution is low and the intermediary
should have been more aggressive. Indeed,
the intermediary would have had more The directional EBEX indicator exhibits a
opportunities to trade at better prices before negative value when NBBEX is lower than
the order was executed. NABEX. In this situation, we can say that
the intermediary should have been more
The way we can interpret NABEX is similar, patient because more opportunities would
except that we focus on what happens after have arisen to enable trading at a better
order execution. NABEX is close to zero price after the order execution. By contrast,
when few traders obtained more favourable the directional EBEX indicator is positive
prices after the execution of the order. The when NBBEX is larger than NABEX. In this
quality of execution is then high because case, the intermediary should have been
few traders execute at better prices than more aggressive because more opportunities
the intermediary did. NABEX is close to one would have arisen to trade at a better price
when many traders obtained better prices, before the order execution.
indicating a poor trading performance. In
this case, the intermediary did a bad job in The specific situation in which NBBEX is
the sense that more patience should have equal to NABEX corresponds to a directional
been shown. The intermediary would have EBEX of zero. This means that there were


the EBEX Indicators
as many better executions before as after intermediary to assess the consistency his
the execution of the order. In this case, market timing. Again, comparative analysis
the intermediary could have traded at any across several intermediaries may be done
other time to provide better execution. based on directional EBEX distributions
This specific case can also refer to the wherein the score of each intermediary
outstanding situation where both NBBEX is a risk-adjusted performance computed
and NABEX tend to zero. This should mean as the mean value of EBEX divided by its
that the intermediary chose exactly the standard deviation. An active trader will be
right moment to trade and that his market expected to provide an average directional
timing was perfect. In this last specific EBEX as close to zero as possible with the
situation, the zero directional EBEX is lowest standard deviation.
accompanied by a very high absolute EBEX
score. (3) $EBEX
In addition to the two fundamental
When investors give their intermediaries indicators that measure the quality of
discretion over how they implement execution and the quality of market timing,
their trades, they like to assess whether we have developed a third indicator, known
their trades are well timed. As we have as $EBEX, whose purpose is to indicate the
just seen, the directional EBEX offers a cost to the investor of the intermediary’s
straightforward and easy-to-interpret trading performance. In short, this indicator
measure of the quality of the market likens the absolute EBEX scores reached
timing that varies from –1 (too aggressive) by intermediaries to an opportunity cost
to 1 (too slow) and makes it possible for their clients. The $EBEX is computed
to gauge the intermediary’s use of the trade-by-trade for a given absolute EBEX
discretion he has been granted. This point that is defined arbitrarily or, when possible,
is relevant as pre-trade analysis should help corresponds to the performance objective
intermediaries predict likely market set by the investor.
conditions and define an appropriate
trading strategy. However, to be really Take, for example, an EBEX target of 0.75
relevant, the directional EBEX assumes specified by the investor. To determine
that the intermediary has full discretion the total cost implied by the quality of
about how he fills the order within a execution achieved by his intermediary, we
given time window and that he does not focus only on the trades with an EBEX lower
systematically execute the order at the end than 0.75 and we compute trade-by-trade
of the specified period. The first assumption the signed difference between the actual
is realistic since many brokers are often execution price and the price that would
given the market close as a deadline. The have yielded an EBEX of at least 0.75. We
second assumption depends more on then multiply each price difference by its
the intermediary’s commitment to best corresponding trade size and add up all the
execution. weighted price differences to get the $EBEX
indicator. To facilitate interpretation and
Much as with absolute EBEX, we may analyse allow meaningful comparisons, we express
the distribution of directional EBEX of an the $EBEX in % of the total monetary


the EBEX Indicators
volume executed by the intermediary. The Constraints are generally dependent on

$EBEX can be viewed as an opportunity either price or volume. Most correspond to
cost since it represents the loss of failing to predetermined core trading strategies that
achieve particular trading performance. are applied manually or, more and more,
automatically, as a result of algorithmic
Suppose, for example, a broker who has trading that has become the must-have
executed 100 orders for a single investor in the last few years and delegates the
with a total traded volume of €49,827,671. scheduling and execution of an order to a
His distribution of absolute EBEX exhibits computer programme.
a median score of 0.50 while he was set a
target of 0.75. About eighty of the 100 trades We can categorise price constraints as
he completed missed the target. Based on explicit and implicit strategies. Explicit
these trades, the $EBEX is equal to €104,586 strategies promote trading at a given price
and corresponds to about 0.21% of the total or better. The most frequent targets are
volume traded. Now suppose another broker the closing price of the day (market-on-
who has also executed about 100 orders close orders) or the bid-ask midpoint at the
for the same investor with a slightly lower release time (arrival price orders41) but the
41 - Some algorithm
providers refer to them as
total volume of €43,405,588. He too has a target may also be a price directly specified
implementation shortfall median score of 0.50 for the absolute EBEX by the investor (€25, for example). Implicit
orders.
but his $EBEX amounts to €182,546, about strategies are mainly VWAP strategies and
0.42% of the total volume he executed. customised strategies. A VWAP strategy
Using $EBEX to compare the two brokers schedules the trade according to the historical
helps the investor to identify that, though average volume profile of the security: it
the quality of their execution is similar, the breaks up the order into smaller lots and
second broker’s underperformance is more trades them at every x minutes depending
expensive. on the volume that has traded historically
in that interval. A customised strategy gives
(4) Potential adjustments a trader who is not benchmarked the ability
When intermediaries have discretion over to stop or cancel the execution of an order
how they execute orders within a specified during the day or the trading horizon. For
horizon, EBEX indicators based on all market example, trading in a particular security
activity traded in that horizon reflect their may be stopped once a particular price or
ability to work orders and seize the best percentage of traded volume is reached.
available trading opportunities. However,
when intermediaries are given limited Volume constraints correspond to
discretion, the EBEX indicators computed participation strategies that schedule the
over the entire trading horizon may lead trade by market volumes. These strategies
to noisy performance measures, especially result in executing the order as a specified
when the constraints set by the investor are constant percentage of the actual market
not neutral. This is true especially when they volume, regardless of price. The key feature
force the intermediary to execute at times here is that the duration of the trade is
of the least or most favourable prices. determined not by a timeframe, but by the
market conditions of the security.


the EBEX Indicators
When intermediaries are given explicit price The example shown in figure 18, a VWAP
constraints, they have a specific target to strategy for an order of 120,000 shares,
beat but they still have discretion over illustrates this approach. If eight intervals
how they will work the order to reach are considered over the trading horizon,
their objective. Since they do not have any the schedule of execution is predetermined
restrictive timing constraints within the to participate proportionally in the total
trading horizon, EBEX indicators based on all volume over the horizon. The order is
market activity traded in that horizon reflect broken up into eight lots that depend on
their ability to time orders and seize the best expected market conditions. Once the
available trading opportunities. By contrast, entire order is executed, the absolute EBEX
when intermediaries are given implicit price indicator for each lot can be computed
constraints or volume constraints, they have by comparing its average execution price
limited discretion over how they can work and the actual market activity within the
the order to reach the investor’s objective. corresponding trading period. Next, the
In fact, both implicit price constraints and quality of execution for the entire trade
volume constraints impose indirect timing is obtained by computing the weighted
constraints on intermediaries, who are then average of all the absolute EBEX scores.
forced to schedule in a predetermined way
Figure 18: Illustration of the weighted absolute EBEX for
the trade by the historical average volume a VWAP strategy
profile of the security. In such situations, Trading
Expected volume VWAP strategy
we need to avoid noise and distinguish the period
trading performance of the intermediary Buy Sell
Executed EBEX
shares score
and the instructions given by the investor.
1 200,000 200,000 24,000 0.52
For this purpose, EBEX indicators may be
2 150,000 150,000 18,000 0.63
adjusted on both the specified strategy
3 100,000 100,000 12,000 0.71
and the expected market conditions over
4 50,000 50,000 6,000 0.44
the trading horizon as follows: 5 50,000 50,000 6,000 0.54
T
xi , j 6 100,000 100,000 12,000 0.52
EBEX a bs ,i = ∑ EBEX a bs ,ij
j =1 Xi 7 150,000 150,000 18,000 0.72
8 200,000 200,000 24,000 0.27
where j is the time interval, T is the total Weighted EBEX 0.53
number of intervals within the entire trading
horizon, xi,j is the quantity to execute in
interval j according to the specified strategy 3. Illustration of the Framework
for order i, Xi is the total size of order i and To illustrate both the use and the
EBEXabs,ij is the absolute EBEX indicator of the interpretation of our framework, we
lot xi,j computed over interval j. EBEX a bs ,i is present hereafter the findings that we
then the weighted absolute EBEX for order i obtained for a relevant sample of orders
that attempts to measure the global quality provided by a European investment firm.
of execution achieved for this order while Before presenting and interpreting the
accounting for the times that trades are results, we will describe the data that
requested of the intermediary. we used as well as the assumptions that
we made to calculate our indicators.


the EBEX Indicators
A brief description of the order sample is start working it. Similarly, we assume that
also included. the time the response is sent by the broker
is the execution time of the order.
(1) Data and assumptions 2. No consideration for split executions:
Our sample consists of 4,542 Euronext we assume that the order is fully executed
stock orders over the three-month period at the time the broker answers, without
from January to March 2005. These orders considering the execution time of multiple
were fully executed, so each order may be trades, if any, to entirely fill the order.
viewed as one trade. We should point out We assume that the time-stamp response
that we have no information regarding corresponds to the time the broker
the potential constraints attached to these stopped timing the market and working
orders. This information was unavailable the order.
when we collected these order data. 3. No specific instruction set by the investor:
we assume that the broker has full discretion
The order database contains the following over the timing of the trade to complete. As
information for each order/trade: a result, we always compute our indicators
• ISIN code for identifying the stock using as a last point of reference the market
• trade direction—buy or sell close of the day of execution.
• traded volume and its average execution 4. The Euronext order book is the market
price of reference: as explained above, we use
• time-stamp for the time the order is sent only Euronext public intraday trade data
to the broker—release time to compute our indicators. Our peer group
• time-stamp for the time of broker response analysis does not take into consideration
with execution details—response time the entire universe of trades since we lacked
• identification code for the broker who information about trades executed off the
completed the trade—broker ID Euronext central order book (block trades
executed on the upstairs market and OTC
To apply our method to this order sample, trades).
we used the Euronext public trade data
over the same period. This data set contains (2) Analysis of the order sample
the following information for each trade The sample contains orders that exhibit
executed in the central order book: various characteristics in terms of direction,
• Euronext internal code for the stock size, release time and execution time.
• time-stamp for the time the trade is
executed—execution time Figure 19 shows the distribution of buy and
• price and the number of shares traded sell orders over the three-month period.
January, with about 39% of orders, is the
Given the data at hand, we had to busiest month. By contrast, only about 27%
formulate four assumptions to compute of orders are released in March. 51.47%
our indicators. of sample orders are for sells, 48.54% for
1. No transmission delay: we assume that buys—a good balance. Although it does
the time the order is sent to the broker is the not appear here, we can add that a great
time the broker receives the order and can majority of the orders are executed the


the EBEX Indicators
day of their release. Only about 14% are Figure 20: Distribution of orders by release time
executed later. 1200
1000
Figure 19: Distribution of buy (B) and sell (S) orders by month
Number of orders
1800 800
1750
600
15.35% 19.06%
1500 17.06%
Number of orders
400
1150
900 200
23.54%
750
0
500 14.60%
7:00
8:00
9:00
10:00
11:00
12:00
13:00
14:00
15:00
16:00
17:00
18:00
10.40%
250
1-h interval
0
Jan Feb March
Month
Figure 21: Distribution of orders by execution time
Order direction S B 2500
To document order release and execution 2000

times, we divide each trading session
Number of orders
into one-hour intervals. We show the 1500

distribution of orders by release time in
figure 20 and by execution time in figure 1000
21. As the figures show, most orders are
released in the morning with a peak 500
between 10:00 and 12:00. The bulk of the

orders are executed in the 17:00 interval 0
7:00
8:00
9:00
10:00
11:00
12:00
13:00
14:00
15:00
16:00
17:00
18:00
that includes the market close. At the

time (2005), the market closed at 17:40
1-h interval
at the latest on Euronext and it could be
surprising that some orders are executed The distribution of orders by size is shown in
after the close. This might be the result of figure 22. For a meaningful comparison of
off-market trades but it is more likely that stocks, we categorise the size of each order
they are the result of our assumption that with respect to the daily average volume
the broker response time is the execution (DAV) of the stock. We define DAV as the
time. total traded volume in EUR in a day divided
by the total number of trades completed on
that day. We use the DAV calculated for the
stock and the day to identify the following
five size categories: ]0; 0.5] for orders whose
size does not exceed half the DAV; ]0.5; 1]
for orders whose size exceeds half the DAV
but is still inferior to the DAV; ]1;5] for


the EBEX Indicators
orders whose size lies between one and (a) Overall performance
five DAV; ]5;10] for orders whose size lies Figure 23 shows the empirical distribution
between five and ten DAV; and finally ]10; ] of the absolute EBEX indicators computed
for orders whose size is larger than ten DAV. for all the orders in our sample as well as
Figure 22 shows that the sample contains some of the usual descriptive statistics. As
orders of every size, with a predominance the figure shows, the entire sample exhibits
of large and medium-sized orders. an average absolute EBEX of about 0.50,
with a standard deviation of about 0.31.
Figure 22: Distribution of orders by size
The median is about 0.52 and suggests
1500
relatively average overall execution since
31.01% 50% of the orders have an absolute EBEX
1200 28.11%
lower than 0.52.
Number of orders
900 Figure 23: Distribution of absolute EBEX for the entire order
19.06% sample
600 12
13.28% Statistics
Mean 0.501317
Median 0.519203
8.55% 10
300 Lower Quartile 0.244213
Upper Quartile 0.752985
Std Deviation 0.314073
8
0
Percent
]0; 0.5] ]1; 5] ]10; ]

]0.5; 1] ]5; 10] 6
Order size (in DAV)
4
(3) Results and interpretation 2

Now that we have described the sample of
0
orders, we can present the findings we obtain
0 0.15 0.3 0.45 0.6 0.75 0.9
for our EBEX indicators. The objective here is
Absolute EBEX
to document both the transparency of the
framework and the level of interpretation. Differences between buy (B) and sell (S)
We begin with an analysis of the overall orders are shown in figure 24. As we may
quality of execution. In this case, orders can observe, the quality of execution looks
be aggregated at any level (direction, size, slightly better for buy orders. Both the
trend, etc.) and the interpretation of the mean (0.52) and the median (0.54) scores
absolute EBEX indicator is the most relevant. are higher for these orders while standard
We then show the analysis done broker deviation (0.31 instead of 0.32) is lower.
by broker. Here, interpreting the absolute Consistent with this result, the peak close
EBEX with the directional EBEX as well as to zero is also lower for buy orders: less than
the $EBEX is relevant. 10% against almost 13% for sell orders.


the EBEX Indicators
Figure 24: Distribution of absolute EBEX by trade direction Figure 25: Distribution of absolute EBEX by execution date
10 Statistics Statistics
Mean 0.519459
15
Mean 0.597607
Median 0.538892 Median 0.590401
Std Deviation 0.306763 Std Deviation 0.239332
8 12
6 9
Percent
Percent
B
0
4 6
2 3
Release date = execution date

0 0
Order direction
15 Statistics
12 Statistics
Mean 0.484207 Mean 0.487830
Median 0.503968 Median 0.498637
Std Deviation 0.319941 10 Std Deviation 0.320891
12
Percent
Percent 8
9
6
1
S
6
4
3
2
0 0
0 0.15 0.3 0.45 0.6 0.75 0.9 0 0.15 0.3 0.45 0.6 0.75 0.9
Absolute EBEX Absolute EBEX
Figure 25 shows differences between the Figure 26 exhibits the empirical distribution
orders that are fully executed the day of of absolute EBEX by the order size categories
their release (1) and those that are filled described earlier. No large differences appear
later (0). As the figure shows, the quality of at first sight but, when focusing on statistics
execution appears overall to be better for (not shown in figure 26), both extreme order
the orders that are not entirely executed the size categories exhibit the lowest average
day of their release. Both the mean and the values for the absolute EBEX (0.47 and 0.48
median (0.59) scores are higher for these respectively).
orders and standard deviation is lower.


the EBEX Indicators
Figure 26: Distribution of absolute EBEX by order size Figure 27: Distribution of orders by broker
15 900
Percent
1 10 800 13.88%
5
0 700 5.81%
15 600
Number of orders
3.96%
Percent
3.96%
10
500
2
5 12.83%
7.77%
0 400 6.22%
Order size category
15 300 3.86%
Percent
3.31%
10 14.29% 4.16%
3
5 200 4.91%
0
100 7.22% 6.47%
15
0
Percent
10
Broker
4
5
0
Size category 1 4
15
2 5
Percent
10 3
4
5
0
0 0.2 0.4 0.6 0.8 0.9
Figure 28 presents the empirical distribution
Absolute EBEX of absolute EBEX by broker. Based on the
median EBEX, we can directly conclude
(b) Broker performance that broker 1 offers a higher quality of
In our sample, more than ten different execution than broker 2 or broker 3. Broker
brokers are identified but three of them 1 also appears to be more consistent than
deal with more than 50% of the orders. the two others (standard deviation of 0.31
For convenience, we report here the results instead of 0.35 or 0.33).
for these three brokers only. In order to
Figure 28: Distribution of absolute EBEX by broker
preserve anonymity, we will name them
Statistics
broker 1, broker 2 and broker 3. Figure 27 15
Mean 0.519
Median 0.550
exhibits for each of them the number of 12 Std Deviation 0.312
Percent
9
1
6
trades completed as well as information 3
0
about their size.
Statistics
15 Mean 0.461
12 Median 0.442
Percent
9
Broker
Std Deviation 0.358

2
6
3
0
Statistics
Mean 0.508
Median 0.496
15 Std Deviation 0.331
12
Percent
9
3
6
3
0
0 0.2 0.4 0.6 0.8 1
Absolute EBEX


the EBEX Indicators
In addition to absolute EBEX, we may look we divided this difference by the daily VWAP
at the distribution of directional EBEX in and multiplied the result by 104 to express
figure 29 to analyse the brokers’ market the score in basis points. According to the
timing. For example, broker 1 exhibits a benchmark comparison approach, a positive
negative median directional EBEX meaning (negative) value indicates an execution that
that he tends to be too aggressive. So, this is less (more) favourable than the daily
broker could have reached a higher trading VWAP. Figure 30 is very interesting since it
performance if he were more patient when reveals quite dramatically the inconsistency
executing orders. of VWAP scores.
Figure 29: Distribution of directional EBEX by broker Figure 30: Absolute EBEX vs. VWAP score
15 Statistics
12 Mean -.197
Percent
9 Median -.308
1

3
0
15 Statistics
12 Mean -.023
Percent
9 Median -.021
Broker
2

3
0
42 - Almost all data vendors
as well as some trading
15 Statistics
venues compute and report 12 Mean 0.058
Percent
in real time the daily VWAP. 9 Median 0.004

3

43 - It indicates whether
3
the trader received a higher 0
or lower price than did the -1 -0.7 -0.4 -0.1 0.2 0,5 0,8
“average trader” of the day.
44- This stock was selected Directional EBEX
because it presents the Ignoring extreme performances (bad or
largest number of trades to
examine over the sample
excellent), VWAP scores are globally centred
period (309 trades exactly).
(c) EBEX vs. VWAP on zero and take either positive or negative
The common practice in the industry values whatever the corresponding absolute
has so far been to use the daily VWAP to EBEX. For example, the absolute EBEX is
assess the trading performance, mainly close to 0.7 when only about 30% of all
because of its availability42 and easy market activity was traded at better prices
interpretation.43 However, the absolute EBEX than the execution price of the trade. In
indicator is superior to any VWAP score such a case, the interpretation is clearly
since it provides a meaningful metric that an above-average quality of execution. In
is consistent over time and across securities. figure 30, for this level of EBEX, we observe
We can illustrate this valuable advantage VWAP scores varying from -50bp to +40bp.
with some examples extracted from our When focusing on extreme performances,
empirical analysis for a given stock.44 we identify the same phenomenon. For
very bad performances, the absolute
In figure 30, we plot both the absolute EBEX EBEX equals zero, while VWAP scores vary
and the VWAP scores that we computed from about +95bp to -20bp. For excellent
for all the trades executed on the selected performances, the absolute EBEX reaches 1
stock. The VWAP score was first obtained by while VWAP scores exhibit values ranging
finding the signed difference between the from about -90bp to +20bp. Depending on
execution price and the daily VWAP. Then the prevailing market conditions as well as


the EBEX Indicators
price volatility, VWAP scores do not make but do not deliver any information about
comparison and proper assessment easy at all other market activity dispersion. This
all. By contrast, the absolute EBEX is shown prevents direct assesments of the quality
to be a consistent metric that takes into of execution.
account contemporaneous market activity
Figure 32: Absolute EBEX vs. TVWAP score
and volatility, allowing a direct assessment
of execution quality in the prevailing
circumstances.
I n a d d i t i o n t o t h e V WA P s c o re
inconsistency, figure 30 indirectly reveals
that the daily VWAP should not be viewed
as a “fair” price for any order. By nature, the
daily VWAP is the average of all executions
in a day, including both the opening and
closing prices that represent a large portion
of market activity. As shown by figure 31,
45 - This window goes from
the order release time to the
for the selected stock, the daily VWAP is
market close of the day. therefore often very close to those prices,
which are not always appropriate references
for trades executed during the continuous
session.
Figure 31: Daily VWAP vs. closing/opening price
One way of cutting the bias mentioned

above is to compute a VWAP over a shorter
interval than the day. In figure 32, we plot
both the absolute EBEX and the TVWAP
scores, where the TVWAP is calculated over
the same measurement window as the
EBEX.45 Although slightly reduced, TVWAP
scores still exhibit inconsistency because
they indicate the deviation from an average

Conclusion
Conclusion
In this guide to transaction cost analysis, cost attribution is provided as a service, as

we have attempted to cover the most portfolio performance already is. Current
meaningful tools and techniques available market conditions and the need to extract
to investors, managers and intermediaries every single basis point of outperformance
to properly assess the quality of execution will contribute to this evolution. The
and possibly provide an answer to the EDHEC Risk and Asset Management
difficult challenges posed by article 21 of Research Centre will remain involved in
the MiFID. this evolution in the coming years and
support the publication of research results
Our Transaction Cost Analysis A-Z should obtained not only in the academic world
be viewed only as a starting point and but also by practitioners.
readers should actively refer to the
publications and the authors mentioned
in the pages of our guide. The literature on
transaction cost analysis is not voluminous,
but there is some, and it covers most aspects
of the question. This guide therefore only
provides an overview while attempting to
clarify the various concepts involved.
The industry has begun embracing the

tools and techniques described here and
we are confident that current market
developments and industry changes will
fuel the need for more advanced techniques.
In particular, we strongly believe the
better formalisation of both execution
performances and risk will open the door to
an unlimited area for developing systematic
trading algorithms whose aim is to achieve
optimal execution plans; innovative value
propositions offered by brokers, exchanges
and obviously technology vendors cannot
be far off.
Finally, we strongly believe that the role

of back-office providers in the MiFID Best
Execution chain is key, as they remain in
a unique position to capture and process
transaction data independently from
venues and intermediaries on behalf of
the investors or their representatives. It is
only a matter of time before transaction

References
References
• Almgren, R., and N. Chriss. 2000. Optimal execution of portfolio transactions. Journal
of Risk 3(2): 5–39.
• Boussema, M., A. Bueno and P. Sequier. 2001. Transaction costs and trading strategies:
an empirical analysis on global equity markets. Working paper.
•—. 2002. Transaction costs and trading strategies. In Best execution: executing transactions
in security markets on behalf of investors. European Asset Management Association.
• Copeland, T., and D. Galai. 1983. Information effects and the bid-ask spread. Journal of
Finance 38: 1457-69.
• Demsetz, H. 1968. The cost of trading. Quarterly Journal of Economics 82: 33-53.
• D’Hondt, C., and J.-R. Giraud. 2007. MiFID: the (in)famous European directive? EDHEC
Position Paper (March).
• Domowitz, I., J. Glen and A. Madhavan. 2001. Global equity trading costs. Research
paper, Investment Technology Group.
• Giraud, J.-R. 2004. Best execution for buy-side firms: a challenging Issue, a promising
debate, a regulatory challenge. European Survey on Investment Managers’ Practices
(June).
• Harris, L. 2003. Trading & exchanges: market microstructure for practitioners. New York:
Oxford University Press.
• Ho, T., and H. Stoll. 1981. Optimal dealer pricing under transactions and return uncertainty.
Journal of Financial Economics 9: 47-73.
• Karceski, J., M. Livingston, and E. O'Neal. 2004. Portfolio Transaction Costs at U.S. Equity
Mutual Funds. ZAG comissionned study.
• Kissell, R., and M. Glantz (2003), Optimal trading strategies." New York, AMACOM.
• Munck, N. H. 2005. When transactions went high-tech: a cross-sectional study of equity
trading costs in the light of more sophisticated trading systems. SSRN working paper.
• Nevmyvaka, Y., M. Kearns, A. Papandreou and K. Sycara. 2005. Electronic trading in
order-driven markets: efficient execution. Working paper.
• Perold, A. F. 1988. The implementation shortfall: paper versus reality. Journal of Portfolio
Management 14: 4-9.

References

About the EDHEC Risk and Asset

Management Research Centre
EDHEC is one of the top five The choice of asset allocation Research Centre. All research work must be
business schools in France. The EDHEC Risk and Asset Management part of a research programme, the relevance
Its reputation is built on the and goals of which have been validated from
Research Centre structures all of its research
high quality of its faculty (110
professors and researchers work around asset allocation. This issue both an academic and a business viewpoint
from France and abroad) and corresponds to a genuine expectation from by the Centre's advisory board. This board is
the privileged relationship with the market. On the one hand, the prevailing made up of both internationally recognised
professionals that the school stock market situation in recent years has researchers and the Centre's business partners.
has been developing since its
shown the limitations of active management The management of the research programmes
establishment in 1906. EDHEC
Business School has decided based solely on stock picking as a source of respects a rigorous validation process, which
to draw on its extensive performance. guarantees the scientific quality and the
knowledge of the professional operational usefulness of the programmes.
environment and has therefore Percentage of variation between funds
focused its research on themes
that satisfy the needs of
3.5% Fees
To date, the Centre has implemented six
professionals. EDHEC is 11% Stock Picking 40% Strategic research programmes:
Asset Allocation
also one of the few Asset Allocation and Alternative
business schools in Europe Diversification
to have received the triple
Sponsored by SG Asset Management and
international accreditation:
AACSB (US-Global), Equis Newedge
(Europe-Global) and 45.5 Tactical
Asset Allocation
The research carried out focuses on the
Association of MBAs benefits, risks and integration methods of
(UK-Global). Source: EDHEC (2002) and Ibbotson, Kaplan (2000)
the alternative class in asset allocation.
EDHEC pursues an active
On the other, the appearance of new From that perspective, EDHEC is making
research policy in the field of
finance. The EDHEC Risk and asset classes (hedge funds, private equity), a significant contribution to the research
Asset Management Research with risk profiles that are very different conducted in the area of multi-style/multi-
Centre carries out numerous
from those of the traditional investment class portfolio construction.
research programmes in the
areas of asset allocation and universe, constitutes a new opportunity
in both conceptual and operational terms. Performance and Style Analysis
risk management in both the
traditional and alternative This strategic choice is applied to all of the Part of a business partnership with EuroPerformance
investment universes. Centre's research programmes, whether they The scientific goal of the research is to adapt
involve proposing new methods of strategic the portfolio performance and style analysis
allocation, which integrate the alternative models and methods to tactical allocation.
class; measuring the performance of funds The results of the research carried out by
while taking the tactical allocation dimension EDHEC thereby allow portfolio alpha to be
of the alpha into account; taking extreme risks measured not only for stock picking but also
into account in the allocation; or studying for style timing.
the usefulness of derivatives in constructing
the portfolio. Indices and Benchmarking
Sponsored by Af2i, Barclays Global Investors,
BNP Paribas Investment Partners, NYSE Euronext,
An applied research approach Lyxor Asset Management, and UBS Global Asset
In an attempt to ensure that the research Management
it carries out is truly applicable, EDHEC has This research programme has given rise to
implemented a dual validation system for the extensive research on the subject of indices
work of the EDHEC Risk and Asset Management and benchmarks in both the hedge fund
universe and more traditional investment


classes. Its main focus is on analysing impact and opportunity costs on listed
the quality of indices and the criteria for derivatives order books, the impact of
choosing indices for institutional investors. explicit and implicit transaction costs on
EDHEC also proposes an original proprietary portfolio performances, and the impact of
style index construction methodology market fragmentation resulting from MiFID
for both the traditional and alternative on the quality of execution in European
universes. These indices are intended to listed securities markets. This programme
be a response to the critiques relating to includes the “MiFID and Best Execution”
the lack of representativeness of the style research chair, sponsored by CACEIS, NYSE
indices that are available on the market. Euronext, and SunGard.
In 2003, EDHEC launched the first composite
hedge fund strategy indices. ALM and Asset Management
Sponsored by BNP Paribas Investment Partners,
Asset Allocation and Derivatives AXA Investment Managers and ORTEC Finance
Sponsored by Eurex, SGCIB and the French This research programme concentrates on
Banking Federation the application of recent research in the
This research programme focuses on area of asset-liability management for
the usefulness of employing derivative pension plans and insurance companies. The
instruments in the area of portfolio research centre is working on the idea that
construction, whether it involves improving asset management techniques
implementing active portfolio allocation and particularly strategic allocation
or replicating indices. “Passive” replication techniques has a positive impact on the
of “active” hedge fund indices through performance of asset-liability management
portfolios of derivative instruments is a key programmes. The programme includes
area in the research carried out by EDHEC. research on the benefits of alternative
This programme includes the “Structured investments, such as hedge funds, in long-
Products and Derivatives Instruments” term portfolio management. Particular
research chair sponsored by the French attention is given to the institutional
Banking Federation. context of ALM and notably the integration
of the impact of the IFRS standards and the
Best Execution and Operational Solvency II directive project. It also aims to
Performance develop an ALM approach addressing the
Sponsored by CACEIS, NYSE Euronext, and SunGard particular needs, constraints, and objectives
This research programme deals with two of the private banking clientele. This
topics: best execution and, more generally, programme includes the “Regulation and
the issue of operational risk. The goal of the Institutional Investment” research chair,
research programme is to develop a complete sponsored by AXA Investment Managers,
framework for measuring transaction costs: the “Asset-Liability Management and
EBEX (“Estimated Best Execution”) but Institutional Investment Management”
also to develop the existing framework research chair, sponsored by BNP Paribas
for specific situations (constrained orders, Investment Partners and the "Private Asset-
listed derivatives, etc.). Research also focuses Liability Management" research chair, in
on risk-adjusted performance measurement partnership with ORTEC Finance.
of execution strategies, analysis of market


Seven Research Chairs have been Financial Engineering and Global

endowed: Alternative Portfolios for Institutional
Investors
Regulation and Institutional Investment Sponsored by Morgan Stanley Investment
In partnership with AXA Investment Managers Management
The chair investigates the interaction The chair adapts risk budgeting and risk
between regulation and institutional management concepts and techniques to
investment management on a European the specificities of alternative investments,
scale and highlights the challenges of both in the context of asset management
regulatory developments for institutional and asset-liability management.
investment managers.
Private Asset-Liability Management
Asset-Liability Management and In partnership with ORTEC Finance
Institutional Investment Management The chair will focus on the benefits of the
In partnership with BNP Paribas Investment asset-liability management approach to
Partners private wealth management, with particular
The chair examines advanced asset-liability attention being given to the life cycle asset
management topics such as dynamic allocation topic.
allocation strategies, rational pricing of
liability schemes, and formulation of Dynamic Allocation Models and New
an ALM model integrating the financial Forms of Target Funds for Private and
circumstances of pension plan sponsors. Institutional Clients
In partnership with Groupe UFG
MiFID and Best Execution The chair consists of academic research
In partnership with NYSE Euronext, SunGard, and that will be devoted to the analysis and
CACEIS Investor Services improvement of dynamic allocation
The chair looks at two crucial issues linked models and new forms of target funds.
to the Markets in Financial Instruments
Directive: building a complete framework for
transaction cost analysis and analysing the The EDHEC PhD in Finance
consequences of market fragmentation. The PhD in Finance at EDHEC Business
School is designed for professionals who
Structured Products and Derivative aspire to higher intellectual levels and aim
Instruments to redefine the investment banking and
Sponsored by the French Banking Federation (FBF) asset management industries.
The chair investigates the optimal design
of structured products in an ALM context It is offered in two tracks: a residential track
and studies structured products and for high-potential graduate students who will
derivatives on relatively illiquid underlying hold part-time positions at EDHEC Business
instruments. School, and an executive track for practitioners
who will keep their full-time jobs.
Drawing its faculty from the world’s best

universities and enjoying the support of


the research centre with the most impact conducts regular industry surveys
on the European financial industry, and consultations, and organises
the EDHEC PhD in Finance creates an annual conferences for the benefit of
extraordinary platform for professional institutional investors and asset managers.
development and industry innovation.
The Centre’s activities have also given rise
to the business offshoots EDHEC Investment
Research for Business Research and EDHEC Asset Management
To optimise exchanges between the Education.
academic and business worlds, the EDHEC
Risk and Asset Management Research EDHEC Asset Management Education helps
Centre maintains a website devoted to asset investment professionals to upgrade their
management research for the industry: skills with advanced risk and asset
(www.edhec-risk.com), circulates a monthly management training across traditional
newsletter to over 200,000 practitioners, and alternative classes.
Industry surveys: comparing research advances with industry best practices

EDHEC regularly conducts surveys on the state of the European asset management industry. They look at the
application of recent research advances within investment management companies and at best practices in
the industry. Survey results receive considerable attention from professionals and are extensively reported by
the international financial media.
Recent industry surveys conducted by the EDHEC Risk
and Asset Management Research Centre
The EDHEC European ETF Survey 2008 sponsored by iShares

The EDHEC European Investment Practices Survey 2008 sponsored by Newedge
EDHEC European Real Estate Investment and Risk Management Survey 2007
sponsored by Aberdeen Property Investors and Groupe UFG
EuroPerformance-EDHEC Style Ratings and Alpha League Table

The business partnership between France’s leading fund rating agency and the EDHEC Risk and Asset
Management Research Centre led to the 2004 launch of the EuroPerformance-EDHEC Style Ratings, a free
rating service for funds distributed in Europe which addresses market demand by delivering a true picture
of alpha, accounting for potential extreme loss, and measuring performance persistence. The risk-adjusted
performance of individual funds is used to build the Alpha League Table, the first ranking of European
asset management companies based on their ability to deliver value on their equity management.
www.stylerating.com
EDHEC-Risk website
The EDHEC Risk and Asset Management Research Centre’s website makes EDHEC’s analyses and expertise in
the field of asset management and ALM available to professionals. The site examines the latest academic
research from a business perspective, and provides a critical look at the most recent industry news.
www.edhec-risk.com

About CACEIS
CACEIS is a banking group dedicated CACEIS

to asset servicing. Equally owned by 1 place Valhubert
the groups Crédit Agricole and Caisse 75013 Paris - France
d'Epargne/Banque Populaire, CACEIS Tel.: +33 (0)1 57 78 00 00
services institutional clients and asset
managers worldwide. With €2.3 trillion www.caceis.com
under custody and €1.1 trillion under
administration, we are France's premier
investor services provider and a leading
global player.
Through a network of offices across

Europe, North America and soon Asia,
CACEIS delivers high quality services
covering depositary/trustee - custody,
fund administration, transfer agency and
corporate trust.
Considerable expertise in Alternative

Investment servicing makes CACEIS
one of the leading service providers for
these sophisticated funds. We also offer
a range of specialist services including
cross-border fund distribution support,
designed to assist clients in achieving
their international business development
goals.

About NYSE Euronext
NYSE Euronext (NYX) operates the world’s NYSE Euronext

leading and most liquid exchange group, Cannon Bridge House - 1 Cousin Lane
and seeks to provide the highest levels of London EC4R 3XX - United Kingdom
quality, customer choice and innovation. Tel: +44 (0) 207 623 0444
Its family of exchanges, located in six
countries, includes the New York Stock 39 rue Cambon
Exchange, the world’s largest cash equities 75039 Paris Cedex 01 - France
market; Euronext, the Eurozone’s largest Tel.: +33 (0)1 49 27 10 00
cash equities market; Liffe, Europe’s leading
derivatives exchange by value of trading; www.nyx.com
and NYSE Arca Options, one of the fastest
growing U.S. options trading platforms.
NYSE Euronext offers a diverse array of

financial products and services for issuers,
investors and financial institutions in cash
equities, options and derivatives, ETFs, bonds,
market data, and commercial technology
solutions. NYSE Euronext’s nearly 4,000
listed companies represent a combined
$28.5 / €18.4 trillion (as of May 31, 2008),
in total global market capitalization, more
than four times that of any other exchange
group. Liffe offers an unrivalled range of
global futures and options products, which
are made available to customers worldwide
on its state-of-the-art trading platform,
LIFFE CONNECT®.

About SunGard
With annual revenue of $5 billion, For more information visit:

SunGard is a global leader in software www.sungard.com
and processing solutions for financial
services, higher education and the public Contact:
sector. SunGard also helps information- Elias Nechachby
dependent enterprises of all types to elias.nechachby@sungard.com
ensure the continuity of their business.
SunGard serves more than 25,000 Anders Nilsson
customers in more than 50 countries, anders.nilsson@sungard.com
including the world's 50 largest financial
services companies.
Asset Arena is SunGard’s global suite

of solutions for front-to-back-office
investment support, serving asset
managers, third-party administrators,
insurers and pension fund managers.
Available as an integrated solution set or as
individual offerings, Asset Arena covers the
entire investment management life-cycle,
including trading, compliance, investment
accounting, portfolio management
and performance measurement and
attribution.
An integrated, front-to-back market

data and historical database solution for
storing and managing real-time and high-
volume time series data, FAME is used
by leading institutions in the financial,
energy and public sectors, as well as
third-party content aggregators, software
vendors and individual investors. FAME
provides investment professionals and
intermediaries with a variety of real-time
market data feeds, a Web-based desktop
solution, application hosting, data delivery
components, and tools for performing
analytic modeling.

EDHEC Risk and Asset Management
Research Centre
393-400 promenade des Anglais
BP 3116
06202 Nice Cedex 3 - France
Tel.: +33 (0)4 93 18 78 24
Fax: +33 (0)4 93 18 78 41
E-mail: research@edhec-risk.com
Web: www.edhec-risk.com

D'Hondt C. 2008. Transaction Cost Analysis A-Z. EDHEC. France

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

D'Hondt C. 2008. Transaction Cost Analysis A-Z. EDHEC. France

Uploaded by

Copyright:

Available Formats

An EDHEC Risk and Asset Management Research Centre Publication

Transaction Cost Analysis A-Z

I. Transaction Cost Analysis as Part of the Investment Process................. 9

II. Transaction Cost Components and Drivers...............................................13

III. Measuring Transaction Costs with Post-Trade Analysis.......................27

IV. Estimating Transaction Costs with Pre-Trade Analysis.........................39

V. Trading Performance Measurement............................................................57

VI. A New Framework: the EBEX Indicators...................................................67

About the EDHEC Risk and Asset Management Research Centre............90

About CACEIS . .....................................................................................................94

About NYSE Euronext..........................................................................................96

Printed in France, November 2008. Copyright EDHEC 2008.

An EDHEC Risk and Asset Management Research Centre Publication 3

About the authors

Catherine D'Hondt is an Associate Professor of Finance at the Louvain School of

4 An EDHEC Risk and Asset Management Research Centre Publication

From retail to more professional investors investor protection in a marketplace that

6 An EDHEC Risk and Asset Management Research Centre Publication

An EDHEC Risk and Asset Management Research Centre Publication 7

Section V is dedicated to trading

8 An EDHEC Risk and Asset Management Research Centre Publication

An EDHEC Risk and Asset Management Research Centre Publication 9

I. Transaction Cost Analysis

Four distinctive phases are usually must manage transaction costs

10 An EDHEC Risk and Asset Management Research Centre Publication

I. Transaction Cost Analysis

By nature, all the tasks assigned to TCA must

An EDHEC Risk and Asset Management Research Centre Publication 11

I. Transaction Cost Analysis

12 An EDHEC Risk and Asset Management Research Centre Publication

II. Transaction Cost Components

Understanding exactly what transaction Brokerage commissions are paid to

14 An EDHEC Risk and Asset Management Research Centre Publication

II. Transaction Cost Components

An EDHEC Risk and Asset Management Research Centre Publication 15

II. Transaction Cost Components

16 An EDHEC Risk and Asset Management Research Centre Publication

II. Transaction Cost Components

financial institutions may not benefit as well as by back-office intermediaries

An EDHEC Risk and Asset Management Research Centre Publication 17

II. Transaction Cost Components

18 An EDHEC Risk and Asset Management Research Centre Publication

II. Transaction Cost Components

delivered by back-office providers on the analysis of transaction costs is—because it

The firms that attempt to provide services

An EDHEC Risk and Asset Management Research Centre Publication 19

II. Transaction Cost Components

Figure 2: Weighted average bid-ask spreads in %

0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.40

Source: Various public sources, EDHEC-Risk Advisory

20 An EDHEC Risk and Asset Management Research Centre Publication

II. Transaction Cost Components

Market impact is one of the most costly 0.10

II. Transaction Cost Components

Figure 4: Average costs by turnover At each stage between the investment

22 An EDHEC Risk and Asset Management Research Centre Publication

II. Transaction Cost Components

opportunity to make an investment in the This fall is reported by Domowitz et al.

An EDHEC Risk and Asset Management Research Centre Publication 23

II. Transaction Cost Components

Figure 8: Total transaction costs vs. market system turnover

Source: Munck (2005)

Figure 7: Implicit transaction costs