See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/46527124

A requiem for the use of the geometric mean in evaluating portfolio

performance

Article  in  Applied Financial Economics Letters · November 2007

DOI: 10.1080/17446540601018964 · Source: RePEc

CITATIONS READS

8 804

3 authors:

Spyros Missiakoulis Dimitrios Vasiliou

Hellenic Open University National and Kapodistrian University of Athens
36 PUBLICATIONS   159 CITATIONS    54 PUBLICATIONS   690 CITATIONS   

SEE PROFILE SEE PROFILE

Nikolaos Eriotis
National and Kapodistrian University of Athens
57 PUBLICATIONS   818 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

A bargaining solution analysis for local government decision: From territorial to behavioral Community-the win-win-win papakonstantinidis situation Case study: the
Greek Case View project

Technical Analysis - Behavioral Finance - Portfolio Μanagement View project

All content following this page was uploaded by Nikolaos Eriotis on 24 August 2015.

The user has requested enhancement of the downloaded file.

This article was downloaded by: [Nat and Kapodistran Univ of Athens ]
On: 05 November 2012, At: 08:00
Publisher: Routledge
Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer House,
37-41 Mortimer Street, London W1T 3JH, UK

Applied Financial Economics Letters

Publication details, including instructions for authors and subscription information:
http://www.tandfonline.com/loi/rafl20

A requiem for the use of the geometric mean in

evaluating portfolio performance
a b c
Spyros Missiakoulis , Dimitrios Vasiliou & Nikolaos Eriotis
a
Department of Public Administration, Panteion University of Athens and Omega Insurance,
Alkyonidon 191, Voula, 166 73 Athens, Greece
b
School of Social Sciences, Hellenic Open University, Sahtouri 16 & Ag. Andreou Street, 262
22 Patras, Greece
c
Department of Business and Finance, National and Kapodistrian University of Athens,
Stadiou 5, 105 62 Athens, Greece
Version of record first published: 14 Nov 2007.

To cite this article: Spyros Missiakoulis, Dimitrios Vasiliou & Nikolaos Eriotis (2007): A requiem for the use of the geometric
mean in evaluating portfolio performance, Applied Financial Economics Letters, 3:6, 403-408

To link to this article: http://dx.doi.org/10.1080/17446540601018964

PLEASE SCROLL DOWN FOR ARTICLE

Full terms and conditions of use: http://www.tandfonline.com/page/terms-and-conditions

This article may be used for research, teaching, and private study purposes. Any substantial or systematic
reproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in any form to
anyone is expressly forbidden.

The publisher does not give any warranty express or implied or make any representation that the contents
will be complete or accurate or up to date. The accuracy of any instructions, formulae, and drug doses should
be independently verified with primary sources. The publisher shall not be liable for any loss, actions, claims,
proceedings, demand, or costs or damages whatsoever or howsoever caused arising directly or indirectly in
connection with or arising out of the use of this material.
 Applied Financial Economics Letters, 2007, 3, 403–408

A requiem for the use of the

geometric mean in evaluating
portfolio performance
Spyros Missiakoulisa, Dimitrios Vasilioub,* and Nikolaos Eriotisc
a
Department of Public Administration, Panteion University of Athens and
Omega Insurance, Alkyonidon 191, Voula, 166 73 Athens, Greece
Downloaded by [Nat and Kapodistran Univ of Athens ] at 08:00 05 November 2012

b
School of Social Sciences, Hellenic Open University, Sahtouri 16
& Ag. Andreou Street, 262 22 Patras, Greece
c
Department of Business and Finance, National and Kapodistrian University
of Athens, Stadiou 5, 105 62 Athens, Greece

Although the geometric mean procedure is very popular among financial

analysts, it is shown that when it is applied on rates of returns for
evaluating portfolio performance it does not produce efficient results.
Valuable past performance information is ignored since the geometric
mean procedure applied on rates of returns uses only three specific pieces
of information, namely the initial value, the terminal value and the total
number of time periods under evaluation.

I. Introduction
In finance, as in all social sciences, the application of
quantitative theory and methods focuses not on the
estimate itself, but on the estimator. That is, the
formula by which the data are transformed into
an actual estimate. Because there are no
‘superestimators’, researchers are looking for those
estimators satisfying specific criteria (e.g. unbiased-
ness, asymptotic properties, etc.) as well as being
algebraic functions of all available data and informa-
tion. To disregard or ignore some of the known or
given information will result ‘bad’ estimators.
This note considers the inappropriateness, due to
the oversight of actual data, of the geometric mean as
an estimator of the average return on a financial
asset, or index of returns on a portfolio of such assets,
over a series of time periods.
Performance evaluation is an essential part of the
process of managing investment portfolios. Superior
performance in the past may have resulted from good
luck or from the actions of a highly skilled investment
manager. Conversely, inferior performance in the past
may have been the result of bad luck, or it may have
been derived from excessive turnover, high manage-
ment fees or other costs associated with an unskilled
investment manager. These possibilities suggest that
the first task in evaluating portfolio performance is to
calculate the return realized by the investment
manager over the evaluation period. Because different
methodologies for calculating a portfolio’s return are
available and the methodologies can lead to quite
disparate results, it was difficult to compare the
performance of investment managers. In consequence,
there was a great deal of confusion concerning the
meaning of data provided by investment managers to
their clients. This has led to abuses by some managers
in reporting performance results that were better than
actual performance. To mitigate this problem the
Chartered Financial Analyst Institute (CFAI) has
established standards both for calculating perfor-
mance results and for presenting those results.
Mean returns are usually estimated by averaging
past returns. A procedure which is logical, simple and
in many cases, appropriate. Average returns are
affected by the start and terminal periods used

*Corresponding author. E-mail: vasiliou@eap.gr

Applied Financial Economics Letters ISSN 1744–6546 print/ISSN 1744–6554 online ß 2007 Taylor & Francis 403
http://www.tandf.co.uk/journals
DOI: 10.1080/17446540601018964
 404
in measurement. Among financial professionals, the
most popular mean return used is the geometric one.
II. The Geometric Mean Return
In 1993, the Association for Investment Management
and Research (AIMR) in order to prevent fraud in
money management performance advertisements,
as well as to promote comparability, has adopted a
series of standards providing guidelines and recom-
mendations for reporting. These standards, known as
the AIMR Performance Presentation Standards, are
‘a set of guiding ethical principles intended to promote
Downloaded by [Nat and Kapodistran Univ of Athens ] at 08:00 05 November 2012
full disclosure and fair representation by investment
managers in reporting their investment results’.
Furthermore, the AIMR Performance
Presentation Standards require that the under
evaluation time-period should be divided in specific
subperiods (e.g. into months or quarters) and for
each subperiod, its return is calculated. Then
the geometric average of all subperiod returns is
calculated.
In May of 2004, AIMR changed its name to CFAI.
Nowadays, CFAI has issue the revised Global
Investment Performance Standards (GIPS), which
‘create a single global standard of investment perfor-
mance reporting and increase minimum standards
worldwide. The revised GIPS standards represent
the most comprehensive and significant upgrade to
the Standards since their introduction.’
CFAI is very explicit in the use of geometric mean
in evaluating rates of return. It is characteristic of
CFAI’s views that in seven times where an average
must be computed, the phrase ‘returns must be
geometrically linked ’ is found in the 2006 revision
of CFAI’s Guidance Statement on Calculation
Methodology.
One may ask why use the geometric mean and not
some other average statistic such as for example the
arithmetic mean. There are numerous articles and
books in financial literature explaining and support-
ing the use of geometric mean in evaluating financial
rates of return. To quote a few:
‘The geometric mean has considerable appeal
because it represents the constant rate of return
we would have needed to earn in each year to
match actual performance over some past invest-
ment period. It is an excellent measure for past
performance.’ (Bodie et al., 2002, p. 810)
‘. . . returns were averaged, with the preferred
method being geometric averaging’ (Fabozzi,
1999, p. 713).
S. Missiakoulis et al.
‘If we want to use a compound return on value to
estimate a forward-looking cost of capital, the
geometric mean of the simple returns, which treats
all years the same, is probably the better choice.’
(Fama and French, 1999, p. 1964)
‘Geometric mean return should be used for
measuring historical returns that are compounded
over multiple time periods.’ (Francis and
Ibbotson, 2002, p. 283)
‘The correct method to determine the annual rate
of return is to use a geometric average.’
(Mayo, 2006, p. 313)
‘Some practitioners prefer the geometric mean
because it reflects their view that outperforming a
benchmark by a fixed amount in a down market is
indeed more difficult than doing so in an up
market.’ (Menchero, 2004, p. 77)
‘Investors are typically concerned with long-term
performance when comparing alternative invest-
ments. Geometric mean is considered a superior
measure of the long-term.mean rate of return.’
(Reilly and Brown, 2003, p. 9)
In financial literature, a lot of articles are
concerned with the comparison of arithmetic and
geometric mean in evaluating portfolio performance.
Also, many more considered the problem of predict-
ing unbiased forecast of the terminal value. See, for
example, Blume (1974), Cooper (1996), Indro and
Lee (1997), or more recent articles such as Menchero
(2004, 2005), Jacquier et al. (2003, 2005). None of
them, however, states clearly the deficiency associated
with geometric mean.
We are certain, that if we ask any financial analyst
about using geometric mean in portfolio evaluation,
he will most probably answer ‘it is known that
geometric mean is an imperfect performance measure’
without stating why it is imperfect and what are the
consequences of using it.
When estimating the average rates of return
based on historical data, all approaches that exist
to calculate the historical mean assume a standard
and constant method of transforming the initial
value of given historical data to the terminal value.
Given annual data, the geometric mean assumes
that a constant annualized measure of the propor-
tional change in value occurs until the terminal
value. The geometric mean therefore reflects the
compound rate of return that would have been
earned by an investor who invested at the start of a
 Using the geometric mean in evaluating portfolio performance 405
number of time periods. In other words, it assumes Equation 4 is also derived from a straightforward
that the initial value grew at a constant rate of application of the geometric mean formula, i.e.
return. vffiffiffiffiffiffiffiffiffiffi vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
uT uT  
It is well known that the mathematics of u Y uY Xi
Xi
1
g ¼ t ri ¼ t
T
1 T
1

geometric mean require nonnegative numbers. On 1þ

i¼2 i¼2
Xi
1
the other hand, a financial rate of return could be vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
uT 
any number, either positive, or negative, or zero. uY Xi
1 þ Xi
Xi
1 
¼ t
T
1
Therefore, a transformation is required in order to
i¼2
Xi
1
make our data applicable. We transform each
vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
uT  rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi rffiffiffiffiffiffi
rate of return to an index of return, i.e. if Rt is uY Xi  T
1 X2 X3 . . . XT T
1 XT
the rate of return for year t, we compute the index ¼ t
T
1
¼ ¼
rt as rt ¼ 100 þ Rt and we base all our computations i¼2
Xi
1 X1 X2 . . . XT
1 X1
on rt.
Mathematically speaking, if rt, t ¼ 1, . . . , T denote
Downloaded by [Nat and Kapodistran Univ of Athens ] at 08:00 05 November 2012

the T observed values of random annual returns

(index form), the geometric mean of r is defined as III. The Inappropriateness of the
the Tth root of the product of all the values of the Geometric Mean
set, i.e.
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi Equation 4 tells us that the geometric mean return is
g ¼ T r1 r2 . . . rT ð1Þ independent of the middle periods. It depends only on
three specific parameters, the initial value, the
A second and more practical way to
terminal value and the total number of time periods.
calculate the geometric mean is to take the
What was happened between initial and terminal
antilog of the arithmetic average of the logarithms
periods is irrelevant and does not affect neither our
of r’s, i.e.
computation nor our evaluation. So, valuable infor-
! mation about the in between performance has been
1X T PT
g ¼ anti log logðrt Þ ¼ eð1=T Þ t¼1 logðrt Þ ð2Þ totally lost. Some, they argue that ‘the reason for
T t¼1 using the three points of data is the limitation on
providing interim data points’. Obviously such a
As we said it earlier, the geometric mean return statement is outdated. Nowadays with the extreme
assumes that the initial value grew at a constant rate availability of information no one can argue for lack
of return. If g is the geometric mean return as of data.
aforementioned and X1 and XT represent the initial For our discussion, consider an investment starts at
and terminal values respectively, then the following a market value of 1000, falls to 800 (20% fall or an
equation holds index return of 80) and then increases again to 1000
(25% rise or an index return of 125). The geometric
XT ¼ X1 gT
1 ð3Þ
mean of index return is 100. This outcome will always
which in turn gives be the same (100) no matter the investment perfor-
rffiffiffiffiffiffi mance during the first time period. To illustrate it
XT
T
1 further, consider the following four cases presented
g¼ ð4Þ
X1 in Table 1.

Table 1. Example of four different cases of returns

Case 1 Case 2 Case 3 Case 4

Value Return Value Return Value Return Value Return

1000 1000 1000 1000
1000 100 2000 200 800 80 2000 200
1000 100 3000 150 600 75 1000 50
1000 100 4000 133 400 67 500 50
1000 100 5000 125 200 50 1500 300
1000 100 1000 20 1000 500 1000 67
G.M.R. 1 100 G.M.R. 2 100 G.M.R. 3 100 G.M.R. 4 100
 406 S. Missiakoulis et al.
Table 2. The CFA example

Date Market value Cash flow Market value post cash flow
12/31/99 500 000
1/31/00 509 000
2/19/00 513 000 50 000 563 000
2/28/00 575 000
3/12/00 585 000
20 000 565 000
3/31/00 570 000

As we can see, in all four cases, the geometric mean at 1 000 000. Suppose further that February was a
return is the same simply because all four initial disaster. On 19th of February the market value fell to
values and all four terminal values are exactly 513 000, then the investor added the 50 000 but
Downloaded by [Nat and Kapodistran Univ of Athens ] at 08:00 05 November 2012

the same. on 28th of February the market value reached at

A more interested illustration is the example given
on the last page of CFA Guidance Statement on
Calculation Methodology (revised edition January
2006).
Given the following information, the analyst is
asked to calculate the rate of return for this portfolio
for January, February, March and the first quarter
of 2000, using a true time-weighted (i.e. geometric)
rate of return (See Table 2 above).
100. We now have Table 3.
It is crystal clear that the two percentages are
exactly the same. What had happened during the
quarter, no mater how extreme or not it was, played
no role at all in the final outcome.
From the aforementioned it is obvious that the
geometric mean has a deficiency that can lead to less
than completely satisfactory results. It does not take
into account the sub-period returns, but only the

Solution:
ð509 000
500 000Þ
January RJAN ¼ ¼ 1:80%
500 000
ð513 000
509 000Þ
February 1=31=00
2=19=00 R¼ ¼ 0:79%
509 000
ð575 000
563 000Þ
2=19=00
2=28=00 R¼ ¼ 2:13%
563 000
1=31=00
2=28=00 RFEB ¼ ðð1 þ 0:008Þ  ð1 þ 0:021ÞÞ
1 ¼ 2:92%
ð585 000
575 000Þ
March 2=28=00
3=12=00 R¼ ¼ 1:74%
575 000
ð570 000
565 000Þ
3=12=00
3=31=00 R¼ ¼ 0:88%
565 000
2=28=00
3=31=00 RMAR ¼ ðð1 þ 0:017Þ  ð1 þ 0:009ÞÞ
1 ¼ 2:62%
Quarter 1 RQT1 ¼ ðð1 þ 0:018Þ  ð1 þ 0:029Þ  ð1 þ 0:026ÞÞ
1 ¼ 7:48%

Before proceeding any further, we should note that
the correct figure for the first quarter is 7.55% and
not 7.48%. This difference occurs due to the round-
ing of the figures that was applied by the author of
the Guidance at each calculation.
Let us now examine two extreme changes to the
aforementioned example. Suppose that January was a
big success. The manager has doubled the market
value of the portfolio so on the 31st of January it was
beginning and the ending values of the evaluation
period. An inferior money manager will be granted
with the same performance return with a superior
money manager, if the latter was misfortunate
enough to end up with the same portfolio
value with the former (provided of course that
both portfolios started with the same value).
The superior manager’s portfolio performance may
be affected by factors which were beyond his/her
 Using the geometric mean in evaluating portfolio performance 407
Table 3. The modified CFA example

Date Market value Cash flow Market value post cash flow
12/31/99 500 000
1/31/00 1 000 000
2/19/00 513 000 50 000 563 000
2/28/00 100
3/12/00 585 000
20 000 565 000
3/31/00 570 000

ð1 000 000
500 000Þ

January RJAN ¼ ¼ 100:00%
Downloaded by [Nat and Kapodistran Univ of Athens ] at 08:00 05 November 2012

500 000
ð513 000
1 000 000Þ
February 1=31=00
2=19=00 R¼ ¼
48:70%
509 000
ð100
563 000Þ
2=19=00
2=28=00 R¼ ¼
99:98%
563 000
1=31=00
2=28=00 RFEB ¼ ðð1
0:487Þ  ð1
0:9998ÞÞ
1 ¼ 0:00911%
ð585 000
100Þ
March 2=28=00
3=12=00 R¼ ¼ 584 900:00%
100
ð570 000
565 000Þ
3=12=00
3=31=00 R¼ ¼ 0:88%
565 000
2=28=00
3=31=00 RMAR ¼ ðð1 þ 5849:00Þ  ð1 þ 0:0088ÞÞ
1 ¼ 590 176:99%
Quarter1 RQT1 ¼ ðð1 þ 1:00Þ  ð1 þ 0:0000911Þ  ð1 þ 5901:77ÞÞ
1 ¼ 7:55%

control, or the inferior manager’s portfolio IV. Conclusions

performance may be achieved by chance or coin-
cidence. However, the currently return measure
requirement of the revised Global Investment
Performance Standards is not likely to be powerful
enough to clearly detect such a degree of superiority
or inferiority. Information that is vital to investors in
order to compare investment performance among
investment management is lost. In consequence,
the employment of the geometric mean is unable
to assure investors neither that the performance
information is both complete and fairly presented,
nor that the investment management firms compete
on an equal footing. What can we do to resolve
this problem? This question remains to be answered
by future research undertaken by scholars of
finance. Of course an obvious and easy answer to
our question is to use the arithmetic mean. A statistic
which under certain assumptions is unbiased
and most important it utilizes all available data
and information.
Financial literature is full of both theoretical and
applied texts on using the geometric mean in
evaluating financial rates of return. The geometric
mean procedure is applied on past returns in order
to produce comparable statistics. Although, theoret-
ically speaking, the said procedure is based on all
available data, in fact, it utilizes only the rate of
return at the starting period, the equivalent value at
the terminal period and the number of time periods
under evaluation. Therefore, the geometric mean of
rates of return is, computationally, independent of
the middle periods. What was happened between
initial and terminal periods is irrelevant and does not
affect neither our computations nor our evaluations.
So, valuable information about the in between
performance has been totally lost when the geometric
mean of rates of return is applied. In consequence, the
employment of the geometric mean procedure is
unable to assure investors neither that the
 408
performance information is both complete and fairly
presented, nor that the investment management firms
compete on an equal footing.
References
Blume, M. E. (1974) Unbiased estimates of long-run
expected rates of return, Journal of the American
Statistical Association, 69, 634–8.
Bodie, Z., Kane, A. and Markus, A. J. (2002) Investments,
5th edn., McGraw-Hill/Irwin, Boston.
CFA Institute (2006) Guidance Statement
on Calculation Methodology. Available online at:
http://www.cfainstitute.org/cfacentre/ips/pdf/GSCalc
Downloaded by [Nat and Kapodistran Univ of Athens ] at 08:00 05 November 2012
MethRevised.pdf
Cooper, I. (1996) Arithmetic versus geometric
mean estimators: setting discount rates for
capital budgeting, European Financial Management,
2, 157–67.
Fabozzi, F. J. (1999) Investment Management, 2nd edn.,
Prentice Hall, Upper Saddle River, NJ.
Fama, E. F. and French, K. R. (1999) The corporate cost of
capital and the return on corporate investment, The
Journal of Finance, 54, 1939–67.
View publication stats
S. Missiakoulis et al.
Francis, J. C. and Ibbotson, R. (2002)
Investments, A Global Perspective, Prentice Hall,
Saddle River, NJ.
Indro, D. and Lee, W. (1997) Biases in arithmetic and
geometric averages as estimates of long-run expected
returns and risk premia, Financial Management, 26,
81–90.
Jacquier, E., Kane, A. and Marcus, A. J. (2003) Geometric
or arithmetic mean: a reconsideration, Financial
Analyst Journal, 59, 46–53.
Jacquier, E., Kane, A. and Marcus, A. J. (2005)
Optimal estimation of the risk premium of long
run and asset allocation: a case of compounded
estimation risk, Journal of Financial Econometrics, 3,
37–55.
Menchero, J. (2004) Multiperiod arithmetic attribution,
Financial Analyst Journal, 60, 76–91.
Menchero, J. (2005) Optimized geometric attribution,
Financial Analyst Journal, 61, 60–9.
Mayo, H. B. (2006) Investments An Introduction, 8th edn.,
Thomson–South Western, Mason, Ohio.
Reilly, F. K. and Brown, K. C. (2003) Investment Analysis
& Portfolio Management, 7th edn. Thomson–South
Western, Mason, Ohio.
Performance Presentation Standards (1993) Association
for Investment Management and Research,
Charlottesville, VA.