Billio et al. / International Review of Applied Financial Issues and Economics, Vol. 3, No. 4, 2011
700
Abstract- The aim of this paper is to propose an innovative
score measuring the relative performance – in terms of
return – of an asset allocation with respect to the alternative
allocations offered to the manager. Intuitively, this score is
defined as the percentage of alternative allocations
outperformed by the manager’s allocation. In particular,
considering the case of a manager investing according to the
zero-dollar long/short equally weighted strategy, we study in
details the computation and the properties of this score and
we deal with the related combinatorial issues when the
number of assets is large.
Keywords: Performance measure; Order statistics; Generalized
hyperbolic distribution
JEL Codes: C13, C16, C46, C62
1. Introduction
During the last two decades, the explosive growth of the
asset management industry came with an increasing interest in
the analysis of investment performance. The research in this
area is axed on Sharpe-like ratios proposed in the 60’s [Sharpe
(1966), Treynor (1965), Jensen (1968)] and it is expanding by
developing the notion of performance as a reward counter-
balanced by some risk. The main innovations focused on the
definition and modeling of risk [Shadwick and Keating
(2002), Darolles et al. (2009)]. Practically, the performance of
a portfolio manager, over a given period, is usually computed
as the ratio of his excess return over a risk measure [Grinblatt
et al. (1994)]. The managers are then ranked according to
these ratios, and the manager providing the highest and
steadiest returns receives the best score. These measures are
convenient because they require no assumption on the strategy
of the portfolio managers. However, they suffer major
drawbacks. First, these measures are relative to a peer’s
performance and irrelevant if no peer is found. We generally
© 2011 The S.E.I.F Press. All Rights Reserved.
assume that the best score corresponds to a “good” portfolio
allocation, with no guarantee on the goodness of this
allocation. Secondly, as they are a ratio of two random
variables, they suffer significant estimation errors [Lo (2002)
and Christie (2007) among others] which prevent any
performance comparison to be significant.
In this paper, we differentiate three elements affecting the
performance of a manager: the set of investable portfolios
offered to this manager, the period of investment and the
allocation method used. In general, when we measure the
performance of a manager we aim to measure the performance
of his allocation method, i.e. his ability to correctly choose his
allocation among his set of investable portfolios and for any
period. We can achieve this goal by fixing the two first
elements. First, the set of investable portfolios is defined by
the set of constraints on the portfolio positions. Those are
usually stated by the investment policy of the manager. For
instance, they can be affected by short-sale or diversification
restrictions. Remark that, using Sharpe-like ratios, this issue is
dealt with by comparing managers having the same constraints
and investing with similar strategies, in other words these
managers share the same set of investable portfolios.
Secondly, the period of investment affects the return provided
by an allocation method. Indeed, over two different periods,
the use of the returns in order to compare the performance of
two managers is known to be delicate. A manager using a
certain allocation method would obtain different returns over
the different periods while his allocation method may not be in
question. This issue can be tackled by normalizing the returns
with a proxy of the risk for the different periods. However, in
the following, we propose an alternative way to deal with the
normalization of the returns. Basically, we normalize an
invested portfolio’s return by providing the percentage of
investable portfolios that it outperforms. This normalized
return, called score, has the advantage to require no peer to be
interpreted and it is cross-sectional in the asset returns as it is
cross-sectional in the set of the investable portfolio. This score
is the subject of this paper. Finally, the performance of an
allocation method over several periods can be obtained by
computing its average score over these periods. This
http://www.irafie.com
A Cross-Sectional Score for the Relative Performance of an
Allocation
Monica Billio
University Ca’Foscari of Venice, Italy
Ludovic Calès
University of Lausanne, Switzerland
Dominique Guégan
University Paris-1 Panthéon-Sorbonne, France