“A Survey on the Four Families of Performance Measures”€¦ · (2011) Reward-to-range Ratio...
Transcript of “A Survey on the Four Families of Performance Measures”€¦ · (2011) Reward-to-range Ratio...
1Department of Economics and Management“Marco Fanno”, University of Padova
Vth Conference on Computational and Financial Econometrics
- University of London, December 2011 -
“A Survey on the Four Families
of Performance Measures”
Massimiliano Caporin1 Grégory M. Jannin2
Francesco Lisi3 Bertrand B. Maillet4
2A.A.Advisors-QCG (ABN AMRO),
Variances and University of Paris-1 (PRISM)
3Department of Statistical Sciences,
University of Padova
The fourth author thanks the Europlace Institute of Finance for financial support. This presentation engages onlyits authors and does not necessarily reflect the opinions of their employers. The usual disclaimers apply.
4A.A.Advisors-QCG (ABN AMRO),
Variances and University of Paris-1 (CES/CNRS and EIF)
Agenda
Motivation
Literature
The Four Families of Performance Measures
Preliminary Conclusions
Extensions
A Survey on theFour Families
of PerformanceMeasures
M. CaporinG. Jannin
F. LisiB. Maillet
2 / 25
Motivation
Since the introduction of the Sharpe ratio (1966), different
measures of portfolio performance are proposed in the literature
The most complete studies so far are: Aftalion and Poncet (2003),
Bacon (2008), Cogneau and Hübner (2009a and 2009b)
The aim of this paper is to present a complete survey of portfolio
performance measures, according to four main families:
Relative performance measures;
Absolute performance measures;
Density-based performance measures;
Utility-related performance measures.
1
5
6
A Survey on theFour Families
of PerformanceMeasures
M. CaporinG. Jannin
F. LisiB. Maillet
3 / 25
Some Surveys about Performance Measurement
Bacon C., (2008), Practical Portfolio Performance, Wiley Finance Series,384 pages.
Cogneau P. and G. Hübner (2009a), “The (More Than) 100 Ways tomeasure Portfolio Performance. Part 1: Standardized Risk-adjustedMeasures”, Journal of Performance Measurement 13(4), 56-71.
Cogneau P. and G. Hübner (2009b), “The (More Than) 100 Ways tomeasure Portfolio Performance. Part 2: Special Measures andComparison”, Journal of Performance Measurement 14(1), 56-69.
6
A Survey on theFour Families
of PerformanceMeasures
M. CaporinG. Jannin
F. LisiB. Maillet
Literature (1/2)
4 / 25
The Four Representative Performance Measures
Sharpe W., (1966), “Mutual Fund Performance”, Journal of Business 39(1),119-138.
Jensen M., (1968), “The Performance of Mutual Funds in the Period 1945-
1964”, Journal of Finance 23(2), 389-419.
Keating C. and W. Shadwick, (2002), “A Universal Performance Measure”,Journal of Performance Measurement 6(3), 59-84.
Morningstar Inc., (2002), “Morningstar: The New Morningstar RatingTM
Methodology”, Morningstar Research Report 22/04/02 (available atdatalab.morningstar.com/Midas/PDFs/Research_StarRating.pdf), 20 pages.
6
A Survey on theFour Families
of PerformanceMeasures
M. CaporinG. Jannin
F. LisiB. Maillet
Literature (2/2)
5 / 25
General Form of Relative Performance Measures(most often expressed in return per unit of risk)
where are the (rescaled) returns, P is a function that
depends upon the observed performance, and R is a
corrected risk measure of the investor’s portfolio under study,
such as R =R .
Objective: comparing the observed (rescaled) performance of
the managed portfolio per unit of risk.
2
6
A Survey on theFour Families
of PerformanceMeasures
M. CaporinG. Jannin
F. LisiB. Maillet ,
1**
p*pp rrPM RP
The Four Families of Performance Measures (1)
6 / 25
*
*
*pr
The Four Families of Performance Measures (1)
The Reward-to-variability ratio (Sharpe, 1966)
where is the expectation operator, and are
respectively the returns and the total risk of the portfolio p,
and is the risk-free asset.
Interpretation: this measure evaluates the compensation
earned by the portfolio manager, as gauged by the expected
excess return per unit of portfolio total risk.
1,p p f pS E r r
2
1
A Survey on theFour Families
of PerformanceMeasures
M. CaporinG. Jannin
F. LisiB. Maillet
7 / 25
The Main Sharpe-based Performance Measures
2
1
3
6
4
A Survey on theFour Families
of PerformanceMeasures
M. CaporinG. Jannin
F. LisiB. Maillet
Author Name RescaledPerformance
RiskMeasure
CorrectionCoefficient
Morey and Vinod(2001)
DoubleSharpe Ratio
ܧ ݎ െ ݎStandardDeviation
ௌߪଵ
Zakamouline andKoekebakker(2009)
Adjusted forSkewness SharpeRatio
ܧ ݎ െ ݎStandardDeviation
൛ͳ ଷǡ × ݏ ݎ × 3ଵ
× ܧ ݎ െ ݎ ൈ ߪଵ ൟ
ଵଶ
Israëlsen(2005)
Reward-to-absolute-excess return Ratio
ܧ ݎ െ ݎStandardDeviation
ݏ^1 ܧ ݎ െ ݎ
Sharpe(1994)
InformationRatio
ܧ ݎ െ ݎTracking
Error-
Treynor(1965)
Reward-to-volatilityRatio
ܧ ݎ െ ݎ Beta -
The Four Families of Performance Measures (1)
8 / 25
Performance Measures based on Other Risk Measures
2
1
3
6
4
A Survey on theFour Families
of PerformanceMeasures
M. CaporinG. Jannin
F. LisiB. Maillet
Author Name RescaledPerformance
CorrectedRisk Measure
Yitzhaki(1982)
Gini Ratio ܧ ݎ െ ݎ Gini coefficient
Darolles et al.(2009)
L-performanceTL-moment of
order 1TL-moment of order 2
Konno and Yamazaki(1991)
Mean AbsoluteDeviation Ratio
ܧ ݎ െ ݎ Mean Absolute Deviation
Caporin and Lisi(2011)
Reward-to-range Ratio ܧ ݎ െ ݎ Range
Young(1998)
Minimax Ratio ܧ ݎ െ ݎ Minimax Portfolio
Dowd(2000)
Reward-to Value-at-Risk Ratio
ܧ ݎ െ ݎ Value−at−Risk
Sortino and Satchell(2001)
Reward-to k-LowerPartial Moments Ratio
ܧ ݎ െ ݎLower Partial Moment
of order k
Martin and McCann(1989)
Ulcer PerformanceIndex
ܧ ݎ − ݎAverage Squared
Weekly Drawdowns
The Four Families of Performance Measures (1)
9 / 25
The Main Limits of Relative Performance Measures
Rankings are consistent only if portfolio returns are ellipticallydistributed and/or the representative agent has a quadraticutility function (except for Zakamouline and Koekebakker,2009)
Use of derivative instruments (fat-tailed and skewed returndistributions) or/and strategies with (time-varying) leverageeffects yield to misleading conclusions (see Kao, 2002; Aminand Kat, 2003a and 2003b; Gregoriou and Gueyie, 2003)
A Survey on theFour Families
of PerformanceMeasures
M. CaporinG. Jannin
F. LisiB. Maillet
The Four Families of Performance Measures (1)
10 / 25
2
1
3
5
6
4
General Form of Absolute Performance Measures(expressed in return)
where is a transformation function (generally linear),
are the (rescaled) returns, is a function that depends upon
the observed performance and is a function that is related
to the theoretical performance of the investor’s portfolio under
study, conditionally to a set of information denoted .
Objective: comparing the observed (rescaled) performance of the
managed portfolio to its theoretical performance, considering a
model.
,, ** pth
pp rrPM PPΓ
A Survey on theFour Families
of PerformanceMeasures
M. CaporinG. Jannin
F. LisiB. Maillet
The Four Families of Performance Measures (2)
11 / 25
P
thP
2
1
3
5
6
4
The Alpha (Jensen, 1968)
where is the expectation operator, are the returns of
the portfolio p, are the returns of the market portfolio m,
is the risk-free rate, is the sensitivity of investor’s
portfolio returns with respect to market portfolio returns.
Interpretation: this measure assesses the extra-performance,
realized by the portfolio manager, given its sensitivity to
systematic risk.
A Survey on theFour Families
of PerformanceMeasures
M. CaporinG. Jannin
F. LisiB. Maillet
,,mpfmfpJp rrErrE
The Four Families of Performance Measures (2)
12 / 25
The Main Jensen-type Performance Measures
2
1
3
6
4
A Survey on theFour Families
of PerformanceMeasures
M. CaporinG. Jannin
F. LisiB. Maillet
Author Name RescaledPerformance
TheoreticalPerformance
Fama(1972)
1. Selectivity Measure2. Risk Compensation Measure
ܧ ݎݎ ǡߚ
ݎ ǡߚݎ
Black(1972)
Zero-betaCAPM
ܧ ݎ െ ܧ ௭ݎ ܧ ݎ െ ܧ ௭ݎ ൈ ǡ௭ߚ
Treynor andMazuy (1966)
Market TimingModel
ܧ ݎ െ ݎܧ ݎ െ ݎ ൈ ଵǡǡߚ
+ ܧ ݎ െ ݎଶൈ ଶǡǡߚ
Henriksson andMerton (1981)
ParametricMarket Timing Model
ܧ ݎ െ ݎܧ ݎ െ ݎ ൈ ଵǡǡߚ
� ݔ ݎ െ ܧ ݎ ǡͲ ൈ ଶǡǡߚ
Connor andKorajczyk (1986)
Multi-factorCAPM
ܧ ݎ െ ݎ ܧ ܨ ൈ ǡிೖߚ
ୀଵ
Ferson andSchadt (1996)
Multi-factor ConditionalCAPM
ܧ ݎ െ ݎ ܧ ܨ ൈ ǡிೖߚ ݐ
ୀଵ
The Four Families of Performance Measures (2)
13 / 25
Other Miscellaneous Absolute Performance Measures
2
1
3
6
4
A Survey on theFour Families
of PerformanceMeasures
M. CaporinG. Jannin
F. LisiB. Maillet
Author Name Performance Measure
Henriksson and Merton(1981)
Non-parametricMarket Timing Model ଵ ݐ ଶ ݐ െ ͳ
Moses et al.(1987)
Diversification-adjusted AlphaMeasure
ߙ× ߣ
ଵ
Grinblatt and Titman(1989)
Positive Period WeightingMeasure
×௧ݓ ǡ௧െݎ ݎ
௧ୀଵ
Modigliani and Modigliani(1997)
Risk-Adjusted PerformanceMeasure
ܧ ݎ × ͳ ǡߣ െ ݎ ൈ ǡߣ
Cantaluppi and Hug(2000)
Efficiency Ratio ܧ ݎ െ ݎ × ܧ ݎ െ ݎଵ
Muralidhar(2001)
Correlation-AdjustedPerformance
ܧ ݎ ൈ ݓ × ܧ ݎ ൈ ݓ+ ݎ × ͳെ െݓ ݓ
Muralidhar(2002)
Skill, History And Risk-Adjusted Measure
{ ܧ ݎ ൈ ݓ × ܧ ݎ ൈ ݓ+ ݎ × 1 − −ݓ ݓ ൟ ×
The Four Families of Performance Measures (2)
14 / 25
The Main Limits of Absolute Performance Measures
Performance computed with these measures are stronglyinfluenced by the reference portfolio (market portfolio orproxy)
Almost all of them assumes stability of the systematic risksensitivities of the investor’s portfolio over time (except forFerson and Schadt, 1996)
Disregards skewness and kurtosis of the studied portfolioreturns which may alter rankings when using investmentstrategies based on derivative instruments (except for Hwangand Satchell, 1998)
A Survey on theFour Families
of PerformanceMeasures
M. CaporinG. Jannin
F. LisiB. Maillet
The Four Families of Performance Measures (2)
15 / 25
General Form of Density-based Performance Measures(expressed in scalar terms – no unit)
where are the (rescaled) returns, is a function that
depends upon the observed performance and is a
measure focusing on a specific (left) part of the support of the
density of returns.
Objective: comparing the observed (rescaled) performance of
the managed portfolio to an expression depending on its
losses.
A Survey on theFour Families
of PerformanceMeasures
M. CaporinG. Jannin
F. LisiB. Maillet
,1
**
ppp rrPM PP
The Four Families of Performance Measures (3)
16 / 25
P
P
The Omega measure (Keating and Shadwick, 2002)
where is the cumulative distribution function, are
the returns of the portfolio p and is a threshold.
Interpretation: it compares the potential gains of the
managed portfolio over its potential losses, both defined
according to a threshold.
1
1 ,p p p p pO F r dr F r dr
A Survey on theFour Families
of PerformanceMeasures
M. CaporinG. Jannin
F. LisiB. Maillet
The Four Families of Performance Measures (3)
17 / 25
The Main Density-based Performance Measures
2
1
3
6
4
A Survey on theFour Families
of PerformanceMeasures
M. CaporinG. Jannin
F. LisiB. Maillet
Author Name RescaledPerformance
DownsidePerformance
Farinelli andTibiletti(2008)
One-sidedRisk Measure
ܧ ݔ ݎ െ ǡͲݎ ܧ ݔ ݎ െ ,ݎ 0
Sortino et al.(1999)
Upside-potential ratio
భܯܪ െݎ ݎ ܮ మܯ െݎ ݎ
Biglova et al.(2004)
Rachev ratio ܧ ݎหݎ െ భǡ ܧ ݎหݎ െ మǡ
Biglova et al.(2004)
GeneralizedRachev ratio
ܧ ݔ െݎ, 0భหݎ െ భǡ ܧ ݔ െݎ, 0
మหݎ െ మǡ
Kazemi et al.(2004)
Sharpe-Omega ratio
ܧ ݎ െ ݎ ܧ ݔ െݎ ,ݎ 0
The Four Families of Performance Measures (3)
18 / 25
The Main Limits of Density-based PerformanceMeasures
Performance measures are highly related to the threshold (riskfree rate, Minimum Acceptable Return or Value-at-Risk)
The link with the utility function of the studied agent is notstraightforward
Density are subject to model risk
A Survey on theFour Families
of PerformanceMeasures
M. CaporinG. Jannin
F. LisiB. Maillet
The Four Families of Performance Measures (3)
19 / 25
General Form of Utility-based Performance Measures
(expressed in return per unit of util)
where are the (rescaled) returns, is the expectation
operator, is a value (or utility) function and is a
specific function that depends upon the performance of the
investor’s portfolio.
Objective: incorporating investor's preferences and risk
profiles, through value (or utility) functions, when assessing
the portfolio performance.
2
1
3
5
4
,*pp rEPM VG
A Survey on theFour Families
of PerformanceMeasures
M. CaporinG. Jannin
F. LisiB. Maillet
The Four Families of Performance Measures (4)
20 / 25
G V
The Morningstar Risk-Adjusted Return (Morningstar
Inc., 2002)
where is the expectation operator, are the returns of
a portfolio p, (with ) is the risk aversion coefficient of
the studied investor.
Interpretation: incorporating the behavior of the agent,
through a Power Utility Function, for assessing the portfolio
performance, given a risk aversion coefficient.
2
1
5
12
1 ,p pMRAR E r
A Survey on theFour Families
of PerformanceMeasures
M. CaporinG. Jannin
F. LisiB. Maillet
The Four Families of Performance Measures (4)
21 / 25
The Main Utility-based Performance Measures
2
1
3
6
4
A Survey on theFour Families
of PerformanceMeasures
M. CaporinG. Jannin
F. LisiB. Maillet
Author Name TransformedPerformance
Value (or utility)Function
Stutzer (2000)PerformanceIndex ఏఢԹష
∗
Kaplan(2005)
Lambdameasure ఏఢԹష
∗
Gemmill et al.(2006)
Loss-AversePerformanceMeasures
ଵ
ଶଵ
ଵ భ
ଶ మ
Ingersoll et al.(2007)
Manipulation-ProofPerformanceMeasure
ଵ
ଵ ଵఒ
The Four Families of Performance Measures (4)
22 / 25
The Main Limits of Utility-based PerformanceMeasures
Strongly dependent on the utility function (exponential, poweror logarithmic) characterizing the behavior of the studiedinvestor
Highly sensitive to the investor’s attitude towards risk throughthe risk aversion coefficient (which can be time-varying)
Rankings obtained with these measures are directly related tothe benchmark (risk free rate or proxy)
A Survey on theFour Families
of PerformanceMeasures
M. CaporinG. Jannin
F. LisiB. Maillet
The Four Families of Performance Measures (4)
23 / 25
2
1
3
5
6
4
Preliminary Conclusions
We propose a Survey of performance measures in a uniform and
comprehensive framework through four main families clearly
identified, namely relative, absolute, density-based and utility-
related performance measures
We define these main families according to two essential criteria:
the unit in which it is expressed, and the way the measure is built
We formulate a general expression for each of the four categories
of performance measures, in which most of the performance
measures fits
A Survey on theFour Families
of PerformanceMeasures
M. CaporinG. Jannin
F. LisiB. Maillet
24 / 25
2
1
3
5
6
4
Extensions
On our on-going research agenda, we have planned to add new
performance measures (20 extra or so), in order to complete our
actual Survey
Next step will be to perform some empirical applications in order to
compare the properties of collected measures (rank correlation
tests, persistence, lucky versus star funds studies, etc.)
We would like also to complement our intuitions about a new
performance measure (called “Generalized Performance Measure”),
which can be seen as a generalization of the four main categories
presented in this paper.
A Survey on theFour Families
of PerformanceMeasures
M. CaporinG. Jannin
F. LisiB. Maillet
25 / 25
We are grateful to Christophe Boucher and Patrick Kouontchou for help and
encouragement in preparing this work.
Grégory M. Jannin
Thank you for your attention...
Vth on Computational and Financial Econometrics
- University of London, December 2011 -
“A Survey on the Four Families
of Performance Measures”