Northern Trust

1
Attribution Analysis
An integral part of the investment management
program
Simon Willcox
Paul DOuville
2
Presentation Overview

• A comprehensive suite for asset servicing
• Plan Sponsors to asset managers
• The attribution and data evolution
• A generational model
• A generational range of issues
• The asset class / market evolution
• The simplicity of Equities versus the realities
  of Fixed Income
• Desire for stock level analysis
• Linking performance attribution and risk
  attribution
• From performance Attribution to risk adjusted
  performance attribution
• Northern Trusts development approach
• The core client requirements of an effective
  attribution toolkit
• Asset Owner versus Asset Manager needs
• Current trends, future needs

3
A Comprehensive Suite for Asset Servicing

• What has been the return on our assets?
• Performance measurement (return, excess return,
  attribution)
• By Manager
• By consolidation (e.g. total plan / all passive
  portfolios / all active portfolios)
• By asset class / region / country / sector /
  stock
• Why have the assets performed that way?
• Attribution Analysis
• Asset Managers
• Has asset owner received expected performance,
  true to label validation
• How has each investment desk performed,
  consistency and validation of investment process,
  where to spend time on process
• Have the desks interacted effectively (currency
  management)
• Asset Owners
• how have their decisions impacted performance,
  asset v liability monitoring

4
A Comprehensive Suite for Asset Servicing
Asset owners and managers have a fiduciary
responsibility to monitor their investments.
What analysis is required to enable them to do
this?

• Was this within our investment guidelines?
• Compliance / Risk Monitoring
• External investment restrictions
  (legal/regulatory).
• Internal investment restrictions placed on
  individual managers to limit overall plan risk
• What risks did we take along the way and what
  type of risks lay ahead?
• Risk Analysis
• Ex-Post Have we taken too much risk, or too
  little
• Ex-Ante Is our current risk exposure in line
  with our investment requirements

5
Attribution Objectives
Each user may have a different purpose and level
of understanding, but the output must be
understandable and useful to all parties.

• Meets its purpose
• Maps to clients objectives and fund managers
  investment process
• Accurate and timely delivery
• Easily interpreted
• By clients with supplemental and meaningful
  commentary from fund managers
• By fund managers / marketing and sales
• By third parties /consultants receiving data from
  multiple sources
• An aid to future investment decisions
• Complementary to in-house risk models and other
  front office systems
• Supports previous ex-ante analysis and asset /
  liability modelling
• Automated
• Scalable
• Flexible

6
Northern Trusts Approach to Supporting Clients
Needs
A full suite of attribution capabilities is a
must have for all of clients from a precise
methodology through to timely and meaningful
reporting

• Methodology and formulae
• Needed returns based approach to compliment
  existing excess return driven reporting
• Preferred industry standard methodology but
  wanted integration into infrastructure
• Algorithms programmed directly into core platform
• Delivery and Frequency
• Daily service essential for asset managers
  quantitative analysis and speedy turn around
• Monthly / quarterly service to compliment daily
  information qualitative analysis with
  commentary
• Solution on line reporting via Northern Trust
  Passport
• Review of data integrity refining the process
• Daily delivery can pinpoint implementation issues
  prior to month end
• Investment goals / processes can be reviewed more
  frequently
• Solution daily performance, available daily
• Integration in client reporting across segments
• Headline numbers and impacts needed to be
  highlighted in client reporting
• Solution design and build client reporting
  blocs, marrying numbers with analysis

7
The Data Evolution
The development of the servicing of program
management has been an evolutionary process and
can be represented by five key generations.
Delivery implications within a Generational Model
of performance delivery
Performance Measurement High Level Performance Measurement Detailed Attribution High Level Attribution Detailed Risk Other Ex Ante
Generation 1 Figures Production Quarterly ? Quarterly ? ? ?
Generation 2 Added Value Monthly ? Quarterly ? Quarterly Quarterly
Generation 3 Global Support GIPS requirements included Monthly Monthly Monthly Monthly with more detail Quarterly with more detail
Generation 4 Focused Cost Centre More detail needed Weekly / daily Monthly Weekly / daily with more detail Weekly / daily with more proof of controls Monthly with more detail
Generation 5 Integrated Feedback More interpretation, and link with attribution / risk More interpretation, and link with attribution / risk Weekly / daily -more interpretation than production Weekly / daily -more interpretation than production Weekly / daily -more interpretation than production Weekly / daily -more interpretation than production
More detail More frequency Data intensity for
all models Cost of delivery Performance and
Risk
1990 - 94
1994 - 98
1998 - 2001
2001 - 03
2003 - now
Source Investit Intelligence Outsourced
Performance Measurement for Investment Managers
8
Demands on Attribution Models Through
Diversification
Portfolio holdings and prices
Contribution
D A T A I N T E G R I T Y
Portfolio, benchmark, broad assets, universes
Policy / Balanced
Portfolio, benchmark, broad assets to stock level
Equities
Portfolio, benchmark, sensitivity to
characteristics
Fixed Income
9
The Asset Class and Market Evolution

• Profile of markets are / have changed
• Peer groups versus custom benchmarks move to
  custom allows for more frequent attribution
  reporting (UK move from 75 peer group to less
  than 10 in last 5 years)
• Peer group comparison more relevant in the U.S.
  with large markets and similar mandates
• Equities v Fixed Income v Alternatives
• Liability matching pressures have seen a move
  towards fixed income
• Use of alternative assets is growing
  significantly
• Increased use of OTCs independent models
  required to derive characteristics
• Regulatory Requirements
• Global Investment Standards are limited with
  respect to attribution
• Difficult to point to any one correct approach
  although many are universally standard (eg stock
  level equity attribution)
• Risk awareness
• Attribution is another useful tool for
  identifying risk within a policy / mandate

10
Understanding the Inputs to Attribution Analysis

• Increased analysis gt Increased need for data
  integrity gt Increased focus

Start weights
Average weights
Many different flavours in FI (eg maturity bands)
Increased complexity with frequency
11
The Process Needs to be Seemless
A true end to end mechanism is essential such
that transactional and accounting data flow
through to attribution results with no need for
human intervention
7.Applyformulae and deliver results
1. Accounting Data
6. Index data Sources and mapping
2.Portfolio holdings and transactions
Data flow, Understanding dependencies
5. Application ofClassification Schemes to
produce returns
3.Positional and transaction code mapping
4. Performance calculation Engine
12
Performance Attribution Sample Output (Total
Fund)
Total Fund attribution is a tool to quantify the
impact of strategic investment decisions and
implementation decisions.
13
Performance Attribution Sample Output (Sector
Equity)
Sector level attribution is useful for all types
of funds, even tracker funds as per this example,
to highlight to the manager where extra return is
being generated. Was it deliberate?
14
Performance AttributionSample Output (Stock
Level Equity)
Stock level attribution is useful to pinpoint
where good stocks were chosen and just as
importantly which poor performing stocks were
avoided.
15
What Is Fixed Income Attribution Analysis?

• Returns-based attribution model
• Flexibility to calculate attribution results
  based on client specific mandate types (eg Govs.
  Vs, Corporates)

16
Equity and Balanced vs. Fixed Income Attribution
Should different attribution methodologies exist
between equity vs. fixed income strategies?

• Similarities
• Both are returns based (portfolio vs. benchmark)
  and look to decompose the excess return into the
  conscious decisions of the investment process
• Top Down approachBoth look at the impact of
  investing in specific markets or assets and
  choice of stocks within the market / category
• Differences
• Fixed Income portfolios tend to have a greater
  degree of currency management within the fund
  which needs to be measured independently (passive
  versus active)
• Fixed Income models are more risk orientated with
  the use of duration to measure interest rate
  sensitivity
• Excess returns tend to be smaller, so results in
  fixed income are much more sensitive to price and
  characteristic differences

17
Understanding the Effects
If interest rates fall, then the returns are
positive so overweight duration in rising markets
is good and underweight duration is bad

• DurationPositive / negative impact on excess
  return from a parallel shift in the yield curve
• Yield Curve PositioningPositive / negative
  impact on excess return from a change in shape in
  the yield curve
• Sector / CountryYield spread movements between
  Gov. and Non Gov. Bonds or currencies of issue
• Bond SpecificDid we pick good performing bonds
  along the yield curve? Was credit part of the
  benchmark?
• CurrencyDid we pick good or poor performing
  currencies ?

18
Understanding the Data Requirements

• Portfolio and index (as per the classification
  scheme)
• Market Value (income accrued versus received)
• Effective duration (allows for options on bonds)
• Effective maturity bands for classification
  purposes (callable bonds)
• Currency of Issue rather than country of risk
• Credit rating official versus implied
• Sector classification (Government / Corporate etc)

19
Considerations for fixed income attribution -
General

• Different interpretation of added value
• Yield spread versus total return
• Buy hold versus transaction based
• Arithmetic versus geometric
• Data integrity and consistency
• Security characteristics
• Greater variety of benchmark schemas/definitions
• Front office systems to back office analysis and
  reporting
• Index data requirements cost, distribution and
  formatting
• Greater complexity with derivatives
• Greater portfolio turnover
• Greater complexity transitioning historical
  information

20
Performance Attribution Sample Output (Fixed
Income)
Fixed Income Attribution highlights the impacts
of effective positions (duration adjusted weights)
21
Risk Attribution / Risk Adjusted Performance
Attribution

• Decomposition of ex post and ex ante measures
  into meaningful factors
• Ex ante risk attribution normally built from
  multi factor model approach
• Ex post risk adjusted performance attribution can
  be built from statistics such as tracking error /
  information ratio decomposition
• Ex Post Risk Attribution
• Based on standard deviation of active returns
• Attribution built from asset volatility and
  correlation to tracking error (but using
  traditional performance definitions of selection
  and allocation)
• Selection impact built from asset weight times
  active return volatility in asset
• Allocation impact built from size of asset active
  weight times active return volatility in asset
  relative to overall benchmark.
• Ex Post Risk Adjusted Performance Attribution
• Information Ratio useful (active return /
  tracking error)
• Use of risk weights rather than investment
  weights how much of risk budget has been spent
  on asset

22
Risk Attribution / Risk Adjusted Performance
Attribution
Sample UK Equity 1 Year Attribution Results
23
Strategic Approach to Attribution Developments

• Full integration within infrastructure remains
  crucial
• Analysis is embedded within Northern Trust
  infrastructure
• Important to ensure that data updates at source
  flow efficiently through to end analysis and
  attribution
• Industry leading professionals researching new
  methodologies and approaches
• Active participation in external performance
  conferences
• Close ties with performance professionals in the
  industry
• Continued investment in capital expenditure
• Senior management continue to comit resources to
  analysis and decision support capabilities
• Displined approach to development, testing and
  implementation
• Staying on current with new investment approaches
• Working with clients and asset managers to ensure
  new investment strategies are captured within
  performance
• Strategic alliances to deliver industry leading
  capabilities

24
Recent Developments Around Attribution

• Daily relative attribution capability through
  Passport (Fundamentals)
• Equity (regional, country, sector and stock
  level)
• Popular with investment management community
• Daily indexes information from all major index
  vendors
• Specialist benchmark team negotiating and
  managing index vendor relationships
• Daily benchmark building functionality
• Benchmarks feeding attribution can be built daily
  and rebalanced monthly
• Strategic alliances with third party index and
  characteristic providers

25
Current Trends, Future Needs

• Absolute / Hedge Funds
• Some strategies do not lend themselves to
  relative attribution decomposition
• Contribution analysis is the start
• but do need in-depth knowledge of investment
  decisions (e.g. pairing)
• Derivatives greater complexity and evolution of
  purpose
• Attribution methodologies have existed for
  significant periods of time
• Treatment of ETD v OTCs
• But again, need in-depth knowledge of investment
  decisions to ensure attribution reflective of
  investment process
• Active vs. passive currency management
• Need to strip out forward contracts between
  passive and active decisions
• Need to allocate cost of hedging to appropriate
  investment desk
• Separate measurement of currency overlay programs
• LDI
• A valuation process decomposition of asset
  value versus liability value
• Non published benchmarks derived from asset
  exposure
• Characteristics enhancements and custom indexes
  development

26
Thank You
Simon Willcox
Paul DOuville