Lecture within the Colloquium of Empirical Macroeconomics and

Quantitative Economic Policy

An Evolutionary Portfolio Theory

by

Prof. Dr. Thorsten Hens and

Klaus Reiner Schenk-Hoppé

Institute for Empirical Research in Economics

University of Zurich

University of Bielefeld
Department of Economics
Center for Empirical Macroeconomics
P.O. Box 100 131
33501 Bielefeld, Germany
Institute for Empirical Research in Economics
University of Zurich

Working Paper Series

ISSN 1424-0459

Working Paper No. 74

An Evolutionary Portfolio Theory

Thorsten Hens and Klaus Reiner Schenk-Hoppé

May 2001
 An Evolutionary Portfolio Theory∗
Thorsten Hens
Klaus Reiner Schenk-Hoppé

Institute for Empirical Research in Economics

University of Zurich, Switzerland

Abstract
The purpose of this paper is to suggest a new theory of portfolio
selection which is based on evolutionary reasoning in simple repeated
market situations. According to this new point of view the ultimate
success of a portfolio strategy is measured by the wealth share the
strategy is eventually able to conquer in an evolutionary process of
market selection. We identify a simple portfolio strategy as being the
unique evolutionary stable strategy.

JEL-Classification: D52, D81, D83, G11.

Keywords: portfolio theory, evolutionary finance, incomplete markets.

1 Introduction
The standard theory of portfolio selection (Markowitz 1952) provides an in-
tuitive model of return and risk in terms of mean and variance. This model
is very useful to guide one’s intuition, and because of its appealing simplic-
ity it is also commonly used in practical financial decisions. Moreover, ever
since the work of Sharpe (1964), Lintner (1965) and Mossin (1966) the re-
sulting consequences for capital market equilibrium have been explored and
culminated in the celebrated CAPM.
∗
We are grateful to Igor Evstigneev and seminar participants at Ecofin, NHH-Bergen,
University of Munich, and ETH-Zurich. Thorsten Hens wants to thank Sandra Güth for
collaboration on this topic in an early stage.
Contact address: [email protected], [email protected]

1
 While the Markowitz model is often used in two period portfolio decisions
its generalization to multiple periods, see e.g. Merton (1973), Breeden (1979),
and Magill and Quinzii (2000), is still under debate. Alternatively it has
been suggested to maximize the expected growth rate of wealth. In a series
of papers, Hakansson (1970), Thorp (1971), Algoet and Cover (1988), and
Karatzas and Shreve (1998), among others, have explored the maximum
growth perspective. Computing the maximum growth portfolio is a non-
trivial problem. Even if one restricts attention to simple trading strategies
when markets are incomplete there is no explicit solution to this problem. By
now numerical algorithms to compute the maximum growth portfolio have
been provided by Algoet and Cover (1988) and Cover (1984, 1991), but so
far practical decisions are rarely based on these ideas. Moreover, as usual in
mathematical finance, the price process is taken to be exogenously given and
consequences on the equilibrium are ignored following this approach.
We intend to contribute to this challenging problem by applying recent
ideas from evolutionary game theory. The evolutionary approach to portfo-
lio selection also takes a long-run perspective, however, it emphasizes market
interaction. According to this approach portfolio strategies compete for mar-
ket capital—the endogenous price process is thus a market selection mecha-
nism along which some strategies gain market capital while others lose. The
equilibrium notion this approach provides is a distribution of market capital
(wealth shares) that is invariant under the market selection process. The
species among which the endogenous price process selects are portfolio rules
which have also been called simple trading strategies. These are non-negative
vectors of expenditure shares for assets which are held constant over time. It
is easy to see that every monomorphic population, i.e. a collection of traders
using the same portfolio rule, gives rise to an invariant distribution of wealth
shares. Hence the notion of an invariant distribution does not a priori select
any particular portfolio rule. However, if one also checks the robustness of
the invariant distributions with respect to the innovation of new portfolio
rules (called mutations), then one can identify a single portfolio rule that
is the unique evolutionary stable strategy, i.e. that is able to drive out any
mutations.
According to this rule one should divide wealth proportionally to the
expected relative returns of the assets. In the case of diagonal securities1
the unique evolutionary stable strategy boils down to a well known trading
strategy that in this case is known to be the best trading strategy in models
with exogenous prices as well as in those with with endogenous prices: With
1
We call a system of securities diagonal when in each state exactly one asset has a
non-zero payoff. An example for these are Arrow-securities.

2
exogenous prices (fixed to one for every asset) it has been shown by Breiman
(1961) that dividing income proportionally to the probability of the states
maximizes the growth rate of wealth. This rule is known as “betting your
beliefs.” It can be generated by maximizing the expected logarithm of rel-
ative returns which in turn is known as the Kelly rule, Kelly (1956). With
endogenous prices and diagonal securities, Blume and Easley (1992) have
shown that “betting your beliefs” is the global attractor of the market selec-
tion mechanism. That is to say, this rule will have the highest growth rate
of wealth in any population of portfolio rules. Blume and Easley (1992)’s
result lays the foundation for an evolutionary portfolio theory because it is
the first result that takes market selection seriously.
Our paper on the one hand extends Blume and Easley’s model to any
complete or incomplete payoff structure. On the other hand our result relies
on the idea of evolutionary stability which so far has not been used in any
portfolio theory. As in Blume and Easley (1992) we consider an economy
with short lived assets. With regards to possible applications of the theory,
this is an unsatisfactory assumption that will have to be generalized in future
research. Moreover, we have borrowed the idea of simple trading strategies
from Blume and Easley (1992), and—as in their paper—we keep the idea of
constant and identical savings rates. Simple trading strategies with constant
savings rates are the first set of rules we are able to consider. Later research
will have to extend this analysis, for example to Markovian portfolio rules.
With general trading strategies, recently, Blume and Easley (2000) and San-
droni (2000) have investigated the case of endogenous savings in a complete
market. They show that among all infinite horizon expected utility maximiz-
ers those who happen to have rational expectations will eventually dominate
the market. This interesting result holds irrespectively of the investors’ risk
aversion characteristics. The intuition goes that differences in investors ex-
pectations determine their savings rates which are essential for the growth
of the investors’ wealth.
Since we belief that every utility function is debatable we do not want
to base portfolio rules on (infinite horizon) utility maximization problems.
Instead we take the notion of the portfolio rule as the primitive concept and
then investigate the evolution of any finite set of portfolio rules. Hence for
our analysis it is not important whether the portfolio rules could possibly
be generated by utility maximization. What counts from an evolutionary
point of view is not the utility level but the chances of survival; and, indeed,
the unique evolutionary stable strategy that we identify cannot be generated
by maximization of a stationary utility function over an infinite horizon. In
contrast to the Kelly rule, the unique evolutionary stable portfolio strategy
does not maximize expected logarithm of its own wealth but it minimizes the

3
expected logarithm of all potential invaders to a monomorphic population in
which the incumbent’s portfolio strategy itself determines market prices.
The generalization of Blume and Easley (1992) towards incomplete mar-
kets brings the evolutionary portfolio theory a bit further to a possible appli-
cation. Recall from elementary statistics that the sample mean of the realized
relative payoffs is an efficient and unbiased estimator of the expected relative
payoffs, provided the process of asset payoffs is ergodic. We assume ergodic-
ity. Hence it is very easy to compute our new strategy on real markets. In
Hens and Stalder (2001) we compute the performance of the new strategy on
SWX 1999 data. It turns out that our new strategy out-performs the SMI.
Moreover, in that paper we also consider an artificial market with short lived
asset having returns as generated by the SWX 1999 data. It turns out that in
competition with strategies from classical finance (mean-variance optimiza-
tion) and from behavioral finance (prospect theory) the evolutionary finance
strategy gains the biggest wealth share within a couple of weeks. Hence even
though the theoretical results are derived for idealized returns and in the long
run (i.e. they are asymptotic results on P-almost all paths), in contrast to
Samuelson’s (1979) critique, these results seem also to be relevant for realistic
returns even in the medium run.
In the next section we present the economic model which has the math-
ematical structure of a random dynamical system. Then we define the equi-
librium concepts and the stability notions, Section 3. In Section 4 we discuss
and generalize Blume and Easley’s result in the case of incomplete markets.
Section 5 presents our main result which will be proved using a series of
propositions that are also of independent interest. In section 6 we analyze
the evolutionary fitness of portfolio rules based on mean-variance optimiza-
tion to study the issue of under-diversified portfolios. Section 7 concludes.

2 The Model
Time is discrete and indexed by t. The possible states of nature are de-
termined in each period in time by the realization of a stochastic process
ξ = (ξt )t∈Z with values in some measurable set (E, E). Let (Ω, F, P) denote
the canonical realization of ξ on the path space, i.e. Ω is the sample path
space with representative element ω = (..., ω−1 , ω0 , ω1 , ...), F = E Z is the cor-
responding σ-algebra, and P is the associated probability measure induced by
ξ. ωt denotes the state of nature at time t, and the sequence of observations
up to the end of period t is referred to as ω t . Further let F t = σ{ωu | u < t}
denote the information available at the beginning of period t. By definition,
an F t -measurable random variable can only depend on ω t−1 .

4
 There are finitely many investors i ∈ I = {1, ..., I}, I ≥ 1, endowed with
wealth w0i > 0 at time 0. Assets Akt (ω) := Ak (ωt ), k = 1, ..., K, with K ≥ 2,
live for one period only but are identically re-born in every period. We make
the following
Assumption 1 For all t ∈ Z, (1) Akt : Ω → R+ for all k; (2) for each k
k
PKa setk Ωk ∈ F with P(Ωk ) > 0 such that At (ω) > 0 for all ω ∈ Ωk ;
there exists
and (3) k=1 At : Ω → R++ .
This assumption ensures that all assets yield non-negative payoffs in all states
of nature, each asset yields a strictly positive payoff for a set of states of non-
zero measure, and total payoff of all assets is strictly positive.
In each period in time t every investor selects a portfolio ait = (ai1,t , ..., aiK,t )
with values in RK i
+ . at is an F-measurable random variable. We can make
more specific assumptions (and actually do later on) on the measurability of
the variables such as all ait being F t -measurable for each period in time t.
Given the portfolio ait at time t, the investor’s wealth in period t + 1 is given
by,
XK
i
wt+1 = Akt (ω) aik,t (1)
k=1
Letting ρk,t denote the price of asset k in period t, then—provided that the
agent’s wealth is positive—his budget shares are given by,
ρk,t aik,t
λik,t :=
wti
We define the trading strategy of investor i as a sequence of budget shares
λit = (λi1,t , ..., λiK,t )t≥0 . If all ait and ρk,t are F t -measurable for all t, then each
budget share λit is also F t -measurable.
Assuming that any investor exhausts his budget in all periods in time, i.e.
the portfolio is chosen such that K i i
P
k=1 k,t ak,t = wt for all t ≥ 0, every trading
ρ
PK
strategy λit takes values in the unit simplex ∆K := {x ∈ RK +| k=1 xk = 1}.
The market-clearing prices are given by,
I
1 X i
ρk,t = k λ wi (2)
āt i=1 k,t t

where ākt > 0, assumed to be F-measurable, is the total supply of asset k at

time t.
For the market selection process to be well defined, we need to guarantee
that market prices qt are always positive. A sufficient condition for this is that
some trading strategy with positive initial wealth is completely mixed, i.e. it
has only strictly positive budget shares. Therefore we make the assumption,

5
Assumption 2 In every market there is some trading strategy λ with posi-
tive initial wealth w0 > 0 that is completely mixed, i.e. λ(ω) ∈ int∆K for all
ω ∈ Ω.

It is clear from Assumption 1 and equation (1) that any completely mixed
trading strategy with positive wealth in a period of time maintains a positive
wealth for all future periods.
Taking into account how equilibrium prices are determined, we obtain a
recursive formula for the total wealth of each consumer. Consumer i’s wealth
in period t + 1 is given by,
K
i
X λik,t wti
wt+1 = Akt (ω) ākt PI j j
(3)
k=1
λ
j=1 k,t wt
P i
and the total market wealth in period t + 1, wt+1 := i wt+1 , can be equated
as,
XI XK
i
wt+1 = wt+1 = Akt (ω) ākt (4)
i=1 k=1

From (3) and (4) we obtain a recursive formula for the evolution of the wealth
(or market) shares rti := wti /wt ,
K
i
X Akt (ω) ākt λik,t rti
rt+1 = PK PI (5)
k=1 l=1 Alt (ω) ālt j=1 λjk,t rtj

Finally we define the relative payoff of asset k as,

Ak (ω) āk
Rtk (ω) := PK t l t l
l=1 At (ω) āt

Assumption 3 The supply of each asset is deterministic and independent

of time. By appropriate normalization of the expected payoff of each asset we
can and do assume ākt ≡ 1 for all k.

Assumption 3 obviously ensures that the relative payoff of each asset,

Rtk (ω), also depends only on the state of ωt . Further, Assumption 3 ensures
that market wealth wt+1 (ω t ) ≡ w(ωt ) is completely determined by the current
state of nature.
The prices of the assets normalized by the market wealth are given by,
I
X
qk,t := ρk,t /wt = λik,t rti (6)
i=1

6
i.e. the normalized price is a convex combination of the trading strategies for
asset k over the wealth shares of investors. qk,t is an F-measurable random
variable; it is F t -measurable, if so are all trading strategies.
Note that in the definition of next period wealth we have assumed that no
investor saves or withdraws any money along the process. Our results carry
over to the case of saving rates which are constant over time and identical
among investors.
We will restrict our study of the evolution of wealth shares to the case of
trading strategies that are deterministic and constant over time. Following
the terminology of Blume and Easley (1992) we define,

Def inition 1 A simple trading strategy—also called portfolio rule—is a

deterministic vector of budget shares that is hold constant over time, i.e.
λit (ω) ≡ λi ∈ ∆K .

The restriction to simple trading strategies will permit clear-cut results

on the success of trading strategies. The main results of Blume and Easley
(1992) are also derived under this assumption.2
The model considered in Blume and Easley (1992) is derived from the
above model by assuming that there are finitely many states of nature and
that the state in each period is determined according to an i.i.d. random
draw.
The budget shares that an investor with a simple trading strategy allo-
cates to each asset is independent of the observed history and of time. Under
this assumption we can re-formulate the process of the evolution of wealth
shares (5) for a given set of simple trading strategies (λi )i∈I . Before doing
so, we need to introduce the appropriate framework for our analysis. First,
we need an ergodic-theoretic representation of uncertainty. Second, we need
a dynamical description of the evolution of wealth shares.

Assumption 4 The states of nature are determined by an ergodic process.3

The shift map θ on the space of all sample paths Ω is defined by θω(·) =
ω(1 + ·). Denote by θt the t-times iterate of θ. The family θt , t ∈ Z defines
a measurable flow on Ω, i.e. θt+u = θt ◦ θu for all u, t ∈ Z, θ0 = idΩ , and θ
is measurable and measurably invertible. By Assumption 4, θ and θ−1 are
2
In our discussion of the Blume and Easley (1992) result with diagonal securities we
allow for general F-measurable trading strategies.
3
This assumption also discharges us from using the common “almost surely,” because
any condition or result with this additional restriction can be transferred into a “for all ω”
statement by restricting the space Ω to an invariant subset of full P-measure. This claim
only holds in that generality because time is discrete here.

7
ergodic with respect to P. The list (Ω, F, P, θ) is called an ergodic dynamical
system.
For any fixed set of simple strategies (λi )i∈I the evolution of wealth shares
(5) can be written as,
rt+1 = f (θt ω, rt ) (7)
where
K
X λi r i
t
fi (θ ω, r) = Rk (θt ω) PI k j (8)
j
k=1 j=1 λk r

and Rk (ω) := R0k (ω) (and thus Rk (θt ω) = Rtk (ω)). By ergodicity ERk (θt ω) ≡
R̄k (≤ 1), i.e. the expected relative payoff of each asset is constant over time.
We refer to equation (7) as the market selection process in the following.
The dynamical description of the market selection process employs the
framework of a random dynamical system, Arnold (1998), see also Schenk-
Hoppé (2001). (7) generates a random dynamical system in the following
sense. Let f (ω) := f (ω, ·) : ∆I → ∆I , and define

 f (θt−1 ω) ◦ . . . ◦ f (ω)r for t ≥ 1
ϕ(t, ω, r) := r for t = 0 (9)
t −1 −1 −1
f (θ ω) ◦ . . . ◦ f (θ ω) r for t ≤ −1


In words, ϕ(t, ω, r) is the vector of wealth shares of all investors at time t when
the initial distribution of wealth shares is r and the sequence of realizations
of states ω prevails.
f −1 (ω) is well-defined for all ω ∈ Ω if all strategies are completely mixed
and there are no redundant assets. In this case ϕ is defined for all t ∈ Z;
otherwise t ∈ N. In the following we restrict attention to the case in which
f is invertible.
The family of maps ϕ(t, ω, r) is a random dynamical system on the unit
simplex ∆I . That is, ϕ : Z × Ω × ∆I → ∆I , (t, ω, r) 7→ ϕ(t, ω, r) is a
B(Z) ⊗ F ⊗ B(∆I ), B(∆I ) measurable4 mapping such that ϕ(0, ω) = id∆I
and ϕ(s + t, ω) = ϕ(t, θs ω) ◦ ϕ(s, ω) for all s, t ∈ Z, and ω ∈ Ω.
We refer the reader to the monograph by Arnold (1998) for any additional
information.
It is important to emphasize that the random dynamical system generated
by (7) depends on the trading strategies pursued by the investors. That is,
for any set of simple strategies (λi )i∈I there is a unique random dynamical
system generated by (7). We will refer to such a random dynamical system
as being associated to the set of strategies (λi )i∈I .
4
B denotes the Borel σ-algebra, and B(∆I ) := B(RI ∩ ∆I ) is the trace σ-algebra.

8
3 Evolutionary Stability
In this section we introduce the stability concepts needed to analyze the long
term behavior of the wealth shares under the market selection process (7).
Given a random dynamical system for a set of simple trading strategies
i
(λ )i∈I , I > 0, one is particularly interested in those wealth shares that evolve
in a stationary fashion over the infinite time-horizon. Here we restrict our-
selves to deterministic distributions of wealth shares that are fixed under the
market selection process (7).5 To specify this notion, we recall the definition
of a deterministic fixed point in the framework of random dynamical systems.

Def inition 2 Given a set of strategies (λi )i∈I with associated random dy-
namical system ϕ, r̄ ∈ ∆I is called a (deterministic) fixed point of ϕ if for
all ω ∈ Ω,
r̄ = ϕ(1, ω, r̄) (≡ f (ω, r̄)) . (10)
The distribution of wealth shares r̄ is said to be invariant under the market-
selection process (7) for the set of strategies (λi )i∈I .

Condition (10) is equivalent to r̄ = ϕ(t, ω, r̄) for all t and all ω. (See
footnote 3 for a justification to use “all ω” instead of “almost surely.”)
If ri = 0, then ϕi (t, ω, r) = 0 by (7). Therefore, in any set of trading
strategies each unit vector in ∆I is a fixed point, i.e. the state in which one
investor possesses the entire market does not change over time.
We are particularly interested in those invariant distributions of wealth
shares which are stable under the market selection process. Roughly speak-
ing, stability means that small perturbations of the initial distribution of
wealth shares do not have a long-run effect. If an invariant distribution of
wealth shares is stable, all sample paths starting in a neighborhood of this
distribution at time zero and the sample path of the invariant distribution of
the wealth shares are asymptotically identical. We will need different notions
of stability; they are defined as follows.

Def inition 3 Given a set of strategies (λi )i∈I with associated random dy-
namical system ϕ, an invariant distribution of wealth shares r̄ ∈ ∆I is called
(locally) stable, if there exists a random open set U (ω) containing r̄ such that
for all ω, limt→∞ kϕ(t, ω, r) − r̄k = 0 for all r ∈ U (ω).

U (ω) is a random open set, if it is an open set for all ω and {ω | (∆I \U (ω)) ∩
G 6= ∅} ∈ F for all open sets G, cf. Arnold (1998, Chap. 1.6).
5
See e.g. Schenk-Hoppé (2001) for applications of the general concept of a random fixed
point in economic growth.

9
 Given a set of strategies (λi )i∈I and a locally stable invariant distribution
of wealth shares r̄, then any initial distribution of wealth shares in a small
neighborhood of r̄ is asymptotically identical to r̄ as time tends to infinity.
It is straightforward to see that this notion of stability does not make
sense on the level of individual investors in general. For example, suppose r
is an invariant distribution of wealth shares for the random dynamical system
ϕ associated to the pair of strategies (λ1 , λ2 ). Then r̄α := (r1 , (1 − α) r2 , α r2 )
is invariant for the random dynamical system, say ϕα , on ∆3 associated to
(λ1 , λ2 , λ2 ) for all α ∈ [0, 1]. However, r̄α can never (even if r is locally stable
for ϕ) be locally stable for the random dynamical system ϕα on ∆3 .
We will therefore interpret a distribution of wealth shares r as a distribu-
tion over populations of players where all players within each group play the
same strategy. Thus the wealth share ri denotes that fraction of the total
market wealth belonging to the players of strategy λi . Under this assumption
0
it is clear that all strategies λi are different from each other, i.e. λi 6= λi for
all i, i0 ∈ I, i 6= i0 .
The above definition refers to the stability of a distribution of wealth in
a population with given strategies. However, one would also like to have
a notion of stability in the case that new strategies occur on the market.
We first note that the structure of the market selection process (7) implies
the following extension property. Let (λi )i∈I , I > 0, be any set of simple
strategies. Suppose r̄ is an invariant distribution of wealth shares for the
corresponding random dynamical system on ∆I . Then for any set (λj )j∈J ,
J = {1, ..., J} with J ≥ 0 (J = ∅, if J = 0), of simple strategies, (r̄, 0, ..., 0) ∈
∆I+J (J-times zero) is an invariant distribution of wealth shares for the
random dynamical system on ∆I+J associated to the set of simple strategies
((λi )i∈I , (λj )j∈J ).
Given a set of strategies (λi )i∈I another set of strategies (λj )j∈J is called
0
new, if λj 6= λi for all j ∈ J , i ∈ I and λj 6= λj for all j, j 0 ∈ J , j 6= j 0 ,
i.e. adding a set of new strategies yields a market in which no redundant
strategies are present.
Def inition 4 Given a set of simple strategies (λi )i∈I with associated random
dynamical system ϕ on ∆I . An invariant distribution of wealth shares r̄ ∈ ∆I
is called evolutionary stable, if for all J ≥ 0, (r̄, 0, ..., 0) ∈ ∆I+J is stable for
all sets of strategies ((λi )i∈I , (λj )j∈J ) with (λj )j∈J being new.
For each evolutionary stable distribution of wealth shares there exits an
entry barrier (a random variable here) below which an arbitrary number of
new strategies do not drive out the incumbent players. Any perturbation,
if sufficiently small, does not change the long-run behavior of the distribu-
tion of wealth shares. The market selection process asymptotically leaves

10
the mutants with no wealth share while the market is shared between the in-
cumbents as unchanged. As discussed above, we do not allow for redundant
strategies to be introduced.
Unlike in evolutionary game theory, this notion of stability refers to the
distribution of wealth shares and not to the set of strategies. It may well
be the case that for a given set of strategies there are two different stable
invariant distributions of wealth shares one of which being evolutionary stable
and the other not.
If there is only one strategy present in the market this strategy is assigned
the total wealth. It therefore makes sense to define,

Def inition 5 A strategy is called evolutionary stable, if the invariant dis-

tribution of wealth shares 1 ∈ ∆1 is evolutionary stable.

Finally, we define a corresponding local stability criterion.

Def inition 6 Given a set of simple strategies (λi )i∈I with associated random
dynamical system ϕ on ∆I . An invariant distribution of wealth shares r̄ ∈ ∆I
is called locally evolutionary stable, if for all J ≥ 0 there exists a constant
δ > 0 such that (r̄, 0, ..., 0) ∈ ∆I+J is locally stable for all sets of strategies
((λi )i∈I , (λj )j∈J ) with (λj )j∈J being new and mini∈I maxj∈J kλi − λj k < δ.

A locally evolutionary stable distribution of wealth shares is evolutionary

stable with respect to local mutations. That is, the strategies that can be
pursued by all mutants are limited to small deviations from existing strate-
gies.
We can now turn to the study of the long-run outcome of the market
selection process.

4 Blume and Easley’s result revisited

We briefly outline Blume & Easley’s (1992) findings in the framework of
random dynamical systems theory, extending their result to the case in which
the set of all possible states of nature is an arbitrary set and the random draw
is ergodic.
Analogous to Blume and Easley (1992) we assume that the payoffs of the
assets are diagonal. That is, there exists a measurable partition (Ωk )k=1,...,K
of Ω into sets with strictly positive measure such that Rk (ω) > 0 if and only
if ω ∈ Ωk . (In fact Rk (ω) ∈ {0, 1} for diagonal securities.) Due to this
property we can unambiguously denote the relative payoffs in the market
selection process (7) at time t by R(θt ω).

11
 A trading strategy of investor i is an F-measurable random variable
λi (ω) ∈ ∆K . We denote by λiθt ω the wealth share invested in that asset
k with θt ω ∈ Ωk .
Due to the assumption of diagonal payoffs (7) simplifies to,
λi t r i
i
rt+1 = R(θt ω) PI θ ω j t j
j=1 λθt ω rt

for any sample path ω. The evolution of the ratio of the wealth shares of any
two investors, say i and j, can then be written as,
i
rt+1λiθt ω rti
j =
rt+1 λjθt ω rtj

In other words, the normalized asset price, Ij=1 λjθt ω rtj , cancels out for di-
P
agonal securities.
Fix any deterministic initial wealth shares r0i > 0 and r0j > 0. Then the
asymptotic behavior of the ratio of the two wealth shares is given by,
T −1
1 rTi 1 X λi t λi
lim log j = lim log jθ ω = E log jω
T →∞ T rT T →∞ T λθ t ω λω
t=0

for almost all sample paths ω. The equality on the far right-hand side of
the last equation holds by the (Birkhoff–Chintchin) ergodic theorem. The
expected value is finite if the strategies are completely mixed.
Consequently we obtain that along almost any sample path ω,
1 ri
lim log Tj > 0 if and only if E log λiω > E log λjω (11)
T →∞ T rT
The equation on the left-hand side implies that for almost all ω, log rTi (ω) ≥
T ε + ln rTj (ω) for all sufficiently large T , where ε > 0. Since log rTi (ω) ≤ 0 for
all T and all ω, ln rTj (ω) → −∞ as T → ∞. Thus, we find that rTi (ω) → 1
almost surely.
This result implies the following asymptotic behavior of the market se-
lection process. Those investors who are closest to maximizing the expected
logarithm of the wealth shares will eventually dominate the market. This
result holds regardless of the initial distribution of wealth shares in the pop-
ulation. However, note that even in the case discussed here the surviving
population depends on the strategies present in the population.
The best choice an investor can make in a period t is to set λiθt ω = 1 if
and only if θt ω ∈ Ωk . However, this requires knowledge of the state ωt prior
to the revelation of the random draw at time t.

12
 If the information available at time t is given by F t , i.e. λi (θt ω) has to
be F t -measurable (and thus can only depend on ω t−1 at time t), the optimal
portfolio rule depends on the specific stochastic process that determines the
state of the world. For instance if the underlying ergodic process stems for a
Markov process, then the optimal portfolio rule is one that depends on the
last observation ωt−1 . Further information from the observed history is not
helpful in that case.
Let us turn to the case being closest to Blume and Easley (1992). Suppose
the random draw is i.i.d. and that all trading strategies λi (θt ω), i ∈ I, are
F t -measurable. Then the state ωt is independent of the observed history up
to time t, ω t−1 . In this case the past does not contain any information on
the future and thus it is rational to play a simple trading strategy.
If trading strategies are simple and the state of nature can only take
finitely many values, s = 1, ..., S with P{ω | ωt = s} ≡ ps > 0, the right-
hand side of equation (11) becomes
S
X S
X
ps log λis > ps log λjs
s=1 s=1

Consequently—as in Blume and Easley (1992)—we obtain that those in-

vestors who are closest to maximizing the expected logarithm of the wealth
shares λs (under the distribution of the one-period random draw) will eventu-
ally dominate the market. The strategy maximizing the expected logarithm
of the budget shares λs is “betting your beliefs”, i.e. λs = ps with s = 1, ..., S.
As already pointed out above, if this strategy is present in the population
then it is the unique long-run outcome of the market-selection process. In a
market in which only simple strategies are present, equation (11) becomes an
absolute fitness criterion in the sense that it is independent of the population
under consideration. With diagonal securities and an i.i.d. random draw
maximizing the expected logarithm means maximizing the growth rate of
wealth in any population.

5 Main Result
In this section we present our main result. We first characterize the set of
stationary wealth shares as being monomorphic populations. That is to say,
any population of traders in which every trader uses the same trading strat-
egy gives rise to an invariant distribution of wealth shares and if there are
no redundant assets then only such monomorphic populations correspond to
stationary solutions. Then we show that only the portfolio rule that we pro-
pose is evolutionary stable. Any other portfolio rule can be driven out even

13
by portfolio rules arbitrary close to it, i.e. it is not even locally evolutionary
stable. The central mathematical tool is Oseledets’s multiplicative ergodic
theorem.
With a general payoff matrix we can no longer benefit from the can-
cellation of prices in the evolution of relative wealth shares and there are
some important conceptual differences to the case of diagonal securities. In
contrast to that case there is no longer an absolute fitness criterion for the
survival of trading strategies. The growth rate of any trading strategy now
depends essentially on the population in which it lives. Restricting attention
to the question of local stability of deterministic invariant distributions cir-
cumvents these problems and is still sufficient to single out a unique trading
strategy. Before presenting the main result, we derive two auxiliary results
that are also of independent interest.6
We have already noted that every distribution of wealth shares in which
the players of only one strategy possess the entire market wealth is invari-
ant under the market selection process (and is a deterministic fixed point).
Moreover, if there are no redundant assets there is also a converse to this
observation as the following result shows.

Assumption 5 There are no redundant assets.

Assumption 5 requires that for any two portfolios a1 , a2 ∈ ∆K with a1 =

6
2
P k 1 2
a, k R (ω)(a k − a k ) 6
= 0 on a set of strictly positive measure. That is
different portfolios cannot generate the same stream of dividends.

Proposition 1 Under the maintained assumptions only one strategy can

have strictly positive wealth in every population of strategies with a (deter-
ministic) invariant distribution of wealth shares.

The proof is relegated to the Appendix.

In Proposition 1 all deterministic invariant distributions of wealth shares
are characterized. We next derive a sufficient condition for the stability of
such fixed points. The following Proposition is the central auxiliary result of
the main Theorem.

Proposition 2 Given a set of strategies (λi )i∈I . Under the maintained as-
sumptions the invariant distribution of wealth shares r̄ = en being concen-
trated on the players of the n-th strategy (which is completely mixed by As-
sumption 2) is
6
Güth in (Güth and Ludwig 2000) has also studied the issues addressed in the two
following propositions for the case of a finite state space.

14
 (i) stable, if
K
!
X λi
E log Rk (ω) nk <0 for all i 6= n; (12)
k=1
λk

(ii) unstable, if for some i 6= n

K
!
X λi
E log Rk (ω) nk > 0. (13)
k=1
λk

The proof is given in the Appendix.

The results in Proposition 2 have the following interpretation. In a sit-
uation in which the prices of all assets are determined by the portfolio rule
λn we can measure the exponential growth rate of other, competing portfolio
rules. If the invariant distribution of wealth shares r̄ = en is stable then the
strategy λn which fixes the prices has a higher growth rate in a neighborhood
of this distribution of market wealth than all other portfolio rules in the pop-
ulation. However, if there is at least one strategy that has a higher growth
rate for these prices, r̄ = en is unstable and the λn -player does not reob-
tain total market wealth after a slight deviation from the possing-everything
situation.
The main result of our paper is based on the observation that, allowing
for all possible mutations, only one particular strategy satisfies the necessary
condition for stability derived in Proposition 2.

Theorem 1 Under the maintained assumptions, the simple strategy λ? de-

fined by,
λ?k = ERk (ω),
for k = 1, ..., K is evolutionary stable, and no other strategy is locally evolu-
tionary stable.

Again the proof can be found in the Appendix.

The portfolio rule λ? divides wealth according to the expected relative
payoffs of the assets. For a given asset market of the structure discussed in
this paper, the strategy is very simple to compute; it requires a minimum of
easily accessible information.
Let us consider the case in which the state of nature can only take finitely
many values s = 1, ..., S in detail. Under the assumption of ergodicity P{ω |
ωt = s} ≡ ps >P0 for all s. Therefore, the portfolio rule λ? in Theorem 1
becomes λ?s = Ss=1 ps Rk (s). It is straightforward to see that we reobtain

15
the result by Blume and Easley (1992, Section 3) in the case of diagonal
securities. In this case, λ? corresponds to the Kelly rule of “betting one’s
beliefs.”
We next relate our result in Theorem 1 to the standard concept of evo-
lutionary game theory. In the following it is shown that the strategy λ? can
be interpreted as a Nash equilibrium.
For this purpose we recall the definition of the auxiliary function (15)
which is used in the proof of Theorem 1,
K
!
X α k
ĝ(α, β) := gβ (α) = E log Rk (ω)
k=1
βk

ĝ(α, β) measures the asymptotic exponential growth rate of a strategy α in

a population in which all asset prices are determined by strategy β. Using
Proposition 2, the assertion of Theorem 1 can be stated as:
For all α 6= λ? ,

ĝ(λ? , λ? ) > ĝ(α, λ? ) and;

ĝ(α, α) < ĝ(β, α) for some β in every neighborhood of α.

That is to say λ? is the unique symmetric Nash equilibrium in a game with

payoff function ĝ. Moreover, λ? is also a strict Nash equilibrium. Therefore
λ? is the unique evolutionary stable strategy in the sense of Maynard Smith
and Price (1973), i.e. the population in which all investors play λ? is resistant
to the invasion of any new portfolio rule.

6 Mean-Variance Optimization
In this section we analyze the evolutionary fitness of portfolio rules based on
mean-variance optimization. According to the CAPM, mean-variance opti-
mization leads to a well diversified portfolio, the market portfolio, provided
every investor bases his portfolio choice on the mean-variance principle—an
assumption that is clearly not true in practice. It is also well known that in
practice mean-variance portfolios are often under diversified, i.e. they typi-
cally put positive weight on very few assets only. To cure this defect it is
then usually suggested to modify the mean-variance portfolio by devoting
some positive but small share of the budget on every asset in the portfolio,
ensuring that the portfolio is completely mixed. The next result shows that
this commonly used “quick fix” of the under-diversification problem is indeed
an improvement of the mean-variance portfolio.

16
Corollary 1 Suppose λ̂ is an under-diversified simple strategy, i.e. λ̂k = 0
for at least one k. Denote by λ̂εk := (1 − ε)λ̂k + ε/S, 0 < ε ≤ 1, the
corresponding ε-completed strategy. Then λ̂ε is robust against λ̂-mutants for
all sufficiently small ε > 0, i.e. the distribution of wealth shares that assigns
total wealth to the λ̂ε -player is stable in the population (λ̂ε , λ̂).

See the Appendix for a proof.

Even though using the “quick fix” to prevent under-diversification is bet-
ter than investing according to the under-diversified portfolio rule, it is clear
from the main result Theorem 1, that ε-completed under-diversified simple
strategies are not locally stable (if they do not coincide with λ? ). However,
we next show that the situation for ε-completed portfolio rules λ̂ε is even
worse. Any completely mixed simple strategy drives out λ̂ε for all small
enough ε > 0.

Corollary 2 Given any completely mixed simple strategy λc and any under-
diversified simple strategy λ̂. Then λ̂ε , defined in Corollary 1, is not robust
against λc -mutants for all sufficiently small ε > 0, i.e. the distribution of
wealth shares that assigns total wealth to the λ̂ε -player is not stable in the
population (λ̂ε , λc ).

The proof of this corollary is given in the Appendix.

7 Conclusions
In this paper we have suggested an evolutionary portfolio theory which is
based on an evolutionary process of market selection and mutations of simple
trading strategies. We identify a portfolio rule as the unique evolutionary
stable strategy in a possibly incomplete market of short-lived assets. The
strategy divides wealth according to the expected relative payoffs of the as-
sets. It is very simple to compute and (under the assumption of ergodicity
of the payoffs) the information needed to implement the strategy is easily
available. The success of the strategy characterized in this paper is based
on two principles: frequent repetition of an elementary situation and evolu-
tionary market interaction. Using the first principle well-known results on
the convergence of stochastic dynamical systems (such as the multiplicative
ergodic theorem) can be applied to gain structure for an otherwise intricate
problem. The second principle emphasizes that portfolio choice is not only a
matter of individual optimization but also of strategic interaction in a com-
petition for market capital. As we have seen the strategy we derive here is
the only strategy that is fit enough to drive out any other simple strategies.

17
 We consider the results of this paper as a first step toward a rigorous
evolutionary theory of portfolio selection. Further research has to be done
for the case of more general trading strategies, such as adapted or Marko-
vian strategies. Moreover, the model should be extended to long-lived assets.
The latter extension is extremely important because capital gains are often
more relevant for portfolio decisions than dividend payments. We view the
assumption of short-lived assets as the major obstacle for real-world appli-
cations of evolutionary portfolio theory.

A Appendix
Proof of Proposition 1. We prove the statement by contraposition. Let
λi , i ∈ I, be a family of simple trading strategies such that λi 6= λj for some
i, j ∈ I, i 6= j. Let r ∈ ∆I with ri rj > 0. We will show that r cannot be
invariant.
Since λi , λj ∈ Pint∆K , and ri rj > 0, λi ri 6=Pλj rj . This further implies
that a := (λk r / Il=1 λjk rl )k=1,...,K 6= (λjk rj / Il=1 λjk rl )k=1,...,K =: aj , i.e.
i i i

the ‘portfolios’ ai and aj are different. Due to the non-redundancy Assump-

tion 5, fi (ω, r) 6= fj (ω, r) in equation (7) on a set Ω̃ ⊂ Ω of strictly positive
measure. Hence r is not invariant in the sense of Definition 2. 

Proof of Proposition 2. The proof is mainly an application of Oseledets’s

multiplicative ergodic theorem for random dynamical systems on manifolds,
see Arnold (1998, Chapter 4).
The random dynamical system describing the evolution of wealth shares is
defined on the simplex ∆I , an I −1-dimensional manifold with boundary. We
therefore transform the system and consider a conjugate random dynamical
system on a subset of the Euclidean space.
Define the projection of the unit simplex,
n I−1
X o
I−1 I−1
D := y ∈ R | yi ≥ 0, yi ≤ 1 ⊂ RI−1
+ .
i=1

Further, for each n ∈ I define the map

 I−1
X 
hn : DI−1 → ∆I , hn (y1 , ..., yI−1 ) := y1 , ..., yn−1 , 1 − yi , yn , ..., yI−1 .
i=1

with inverse
h−1
n (x1 , ..., xI ) := (x1 , ..., xn−1 , xn+1 , ..., xI ).

18
hn is a C ∞ -diffeomorphism. We obtain the conjugate random dynamical
system on DI−1 ,
ψn (t, ω) := h−1
n ◦ φ(t, ω) ◦ hn .

Due to the definition of the space DI−1 , we can take directional derivatives
in the direction of all unit vectors at all points in the interior of DI−1 relative
to RI−1
+ . That is we can determine the Jacobian of the conjugate system at
PI−1
all points in {y ∈ RI−1 + | i=1 yi < 1} (which is an invariant set for the
random dynamical system ψn ).
Note that the origin of RI−1 corresponds to the nth edge of the unit sim-
plex. The stability properties of these two fixed points are identical because
of the C ∞ -equivalence of both random dynamical systems.
For notational simplicity we assume without loss of generality that n = I.
Then the partial derivatives of ψI (1, ω, y) = h−1 I ◦ φ(1, ω) ◦ hI (y) are given
by,
K
∂ψIi (1, ω, y) X Rk (ω) (λm I i
k − λk ) λk y i
=−
( j=1 λk yj + λIk (1 − I−1
PI−1 j
∂ym
P 2
k=1 j=1 yj ))

for all i 6= m, and by

K
∂ψIi (1, ω, y) X Rk (ω) (λm I m
k − λk ) λk y m
= −
( j=1 λk yj + λIk (1 − I−1
PI−1 j
∂ym
P 2
k=1 j=1 yj ))
K
X Rk (ω) λm
k
+ PI−1 PI−1
k=1 j=1 λjk yj + λIk (1 − j=1 yj )

for all i = m.
The stability properties of eI ∈ ∆I can be determined by evaluating the
Jacobian of ψI at the origin and applying the multiplicative ergodic theorem
of Oseledets. It will be shown that condition (12) resp. (13) ensures that the
top Lyapunov exponent of this linear system is strictly negative resp. positive.
Results by Wanner (1995), see Arnold (1998, Theorem 7.5.6), ensure that
the dynamic behavior of the linearized system carries over (locally) to the
nonlinear stochastic system.
From the above expressions, we obtain the Jacobian of ψI at y = (0, ..., 0).
It is a diagonal matrix with entry,
K
X λm
Am,m (ω) := R (ω) kI
k

k=1
λk

The multiplicative ergodic theorem, Arnold (1998, Theorem 4.2.6), implies

that the Lyapunov exponents of the fixed point y = 0 of ψI are given by

19
limT →∞ T1 log | Tt=0 Am,m (θt ω)|, m = 1, ..., I − 1. The integrability condition
Q
of the multiplicative ergodic theorem is satisfied because the space ∆I is
compact. By the (Birkhoff–Chintchin) ergodic theorem, we find that this
limit is equal to
   
T K m K m
1X X λ k
X λ k
lim log  Rk (θt ω) I  = E log  Rk (ω) I  (14)
   
T →∞ T  λk   λk 
t=0 k=1 k=1

Zero is a stable fixed point of ψI if the term in (14) is strictly negative for all
m = 1, ..., I − 1. If (14) is strictly positive for some m, then zero is locally
unstable. Due to the diagonal structure of the Jacobian, the eigenspaces
correspond to the linear spaces spanned by the unit vectors (restricted to the
positive orthant RI−1
+ ).
The stability of the original system on ∆I at the fixed point eI ∈ ∆I is
determined by the Lyapunov exponents (14). The corresponding eigenspaces
are given by the vertices. All summands in (14) are positive and we thus
have obtained conditions (12) and (13) of the proposition. 

Proof
PK ?of Theorem 1. Obviously, λ? is a completely mixed strategy, i.e.
?
k=1 λk = 1 and λk > 0 for all k. Next we define the auxiliary function,

K
!
X α k
gβ (α) := E log Rk (ω) (15)
k=1
βk

in accordance with Proposition 2. For each fixed strategy β ∈ int∆K ⊂ RK ,

gβ : int∆K → R. gβ (α) is the Lyapunov exponent of the distribution of
wealth that assigns total wealth to the ‘status quo’ population that plays
strategy β in a market in which α is the only the alternative strategy.
By Proposition 2 the first assertion of the theorem follows if we can show
that gλ? (α) < 0 for all α ∈ int∆K with α 6= λ? .
We prove that gβ (α) is strictly concave for all β ∈ int∆K and that gλ? (α)
takes its maximum value at α = λ? .
To ensure strict concavity it suffices to show that α 7→ gβ (α) is strictly
concave on the space RK ++ , because restriction of the domain to the linear sub-
space int∆ preserves strict concavity. The function log K
K k
P
k=1 (R (ω) αk /βk )
is concave for all ω and—due to the no-redundancy Assumption 5—strictly
concave on a set of positive measure. Therefore gβ (α) is strictly concave for
each fixed β ∈ int∆K .
We can now employ that λ? is the unique maximum of gλ? (α) on int∆K
if all directional derivatives at this point are zero.

20
 The partial derivative of gβ (α) with respect to the i-th component αi is
given by
∂gβ (α) Ri (ω)/βi
= E PK αk
∂αi k
k=1 R (ω) β k

Observe
PK that interchanging integration and differentiation is allowed because
log( k=1 R (ω) αk /βk ) is integrable for each fixed α (follows from ERk (ω) ≤
k

1 < ∞ for all k) and E(Ri (ω)/ K k

P
k=1 R (ω)) ≤ 1 < ∞ (this follows from the
fact that Rk (ω) ≥ 0 for all k and all ω by assumption). The last equation
implies
∂gλ? (λ? ) Ri (ω) Ri (ω)
=E ? =E ≡1
∂αi λi ERi
for all i = 1, ..., K, since K k
P
k=1 R (ω) ≡ 1 for all ω.
The directional derivative of gλ? in the direction (dα1 , ..., dαK ) with the
restriction K
P
k=1 dαk = 0 (which is a vector in the simplex) is equated as

K
X ∂gλ? (λ? )
dαi = 0.
i=1
∂αi

Let us next prove that any simple strategy β 6= λ? with β ∈ int∆K is not
locally evolutionary stable. That is, for any neighborhood of β there exists
an α such that gβ (α) > 0. It suffices to show that the directional derivative
of gβ at β is strictly positive in one direction.
Since β 6= λ? and both are points in the simplex there exists i 6= j with
βi > λ?i and βj < λ?j . Note that we have assumed a minimum of two assets.
The directional derivative of gβ at β in the direction dα given by dαi =
−1/2, dαj = 1/2, and zero otherwise, is given by,
K K
ERk 1 λ?j λ?i
 
X ∂gβ (β) X
dαk = dαk = − > 0.
k=1
∂αk k=1
βk 2 βj βi


Proof of Corollary 1. According to Proposition 2(i) it suffices to show
that
K
 X 
k λ̂k
E log R (ω) <0
(1 − ε)λ̂k + ε/S
k:λ̂k >0

for all small ε > 0. The left-hand side of this equation is strictly increased by
omitting ε/S in the denominator. We thus obtain the sufficient condition,
 X 
k
E log R (ω) ≤ log(1 − ε) (16)
k:λ̂k >0

21
Since there is at least one k such that λ̂k = 0, we find that k:λ̂k >0 Rk (ω) < 1
P
on a set of positive measure
P (the term
is bounded by 1 for all ω), the left-
k
hand side of (16) E log k:λ̂k >0 R (ω) < 0. Therefore (16) is satisfied for
all small enough ε. 

Proof of Corollary 2. Again we employ Proposition 2. The local instability

result says that the assertion of the Corollary is true, if
K
!
c
X λ k
E log Rk (ω) >0 (17)
k=1 (1 − ε)λ̂ k + ε/S

for all small ε > 0.

Noting that
K
X Rk (ω) λck X Rk (ω) λck X S λck
= + Rk (ω)
k=1 (1 − ε)λ̂k + ε/S (1 − ε)λ̂k + ε/S ε
k:λ̂k >0 k:λ̂k =0

λck > 0 for all k, and Rk (ω) > 0 on a set of positive measure for all k with
λ̂k = 0, we find that the left-hand side of (17) tends to infinity as ε → 0. 

References
Algoet, P. H., and T. M. Cover (1988): “Asymptotic Optimality and
Asymptotic Equipartition Properties of Log-Optimum Investment,” An-
nals of Probability, 16, 876–898.

Arnold, L. (1998): Random Dynamical Systems. Springer-Verlag, Berlin.

Blume, L., and D. Easley (1992): “Evolution and Market Behavior,”

Journal of Economic Theory, 58, 9–40.

(2000): “If You’re So Smart, Why Aren’t You Rich? Belief Selection
in Complete and Incomplete Markets,” mimeo, Department of Economics,
Cornell University.

Breeden, D. (1979): “An intertemporal capital asset pricing model with

stochastic consumption and investment opportunities,” Journal of Finan-
cial Economics, 7, 265–296.

Breiman, L. (1961): “Optimal gambling systems for favorable games,”

Fourth Berkely Symposium on Mathematical Statistics and Probability, 1,
65–78.

22
Cover, T. (1984): “An algorithm for maximizing expected log-investment
return,” IEEE Transformation Information Theory, 30, 369–373.

(1991): “Universal Portfolios,” Mathematical Finance, 1, 1–29.

Güth, S., and S. Ludwig (2000): “Evolution in Financial Markets,” Ph.D.

thesis, University of Bielefeld, Chapter 6: “Hope for Homo-Oeconomicus?
On the Evolutionary Fitness of Utility Maximization”.

Hakansson, N. (1970): “Optimal investment and consumption strategies

under risk for a class of utility functions,” Econometrica, 38, 587–607.

Hens, T., and M. Stalder (2001): “Testing the evolutionary portfolio

theory,” Working paper, Institute for Empirical Research in Economics,
University of Zurich, in preparation.

Karatzas, I., and S. Shreve (1998): Methods of Mathematical Finance.

Springer-Verlag, New York.

Kelly, J. (1956): “A New Interpretation of Information Rate,” Bell System

Technical Journal, 35, 917–926.

Lintner, J. (1965): “The Valuation of Risky Assets and the Selection of

Risky Investment in Stock Portfolios and Capital Budgets,” Review of
Economics and Statistics, 47, 13–37.

Magill, M., and M. Quinzii (2000): “Intertemporal CAPM,” Economic

Theory, 15, 103–138.

Markowitz, H. (1952): “Portfolio Selection,” Journal of Finance, 7, 77–91.

Maynard Smith, J., and G. Price (1973): “The Logic of Animal Con-
flict,” Nature, 246, 15–18.

Merton, R. (1973): “An Intertemporal Capital Asset Pricing Model,”

Econometrica, 41, 867–888.

Mossin, J. (1966): “Equilibrium in a Capital Asset Market,” Econometrica,

34, 768–783.

Samuelson, P. (1979): “Why we should not make mean-log of wealth big,

though years to act are long,” Journal of Banking and Finance, 3, 305–307.

Sandroni, A. (2000): “Do Markets Favor Agents Able to Make Accurate

Predictions?,” Econometrica, 68, 1303–1341.

23
Schenk-Hoppé, K. R. (2001): “Random Dynamical Systems in Eco-
nomics,” Stochastics and Dynamics, 1, 63–83.

Sharpe, W. (1964): “Capital Asset Prices: A Theory of Market Equilibrium

Under Conditions of Risk,” Journal of Finance, 19, 425–442.

Thorp, E. (1971): “Portfolio choice and the Kelly criterion,” In Stochastic

Models in Finance,W.T. Ziemba and R.G. Vickson, eds., 599–619.

Wanner, T. (1995): “Linearization of random dynamical systems,” in Dy-

namics Reported Vol 4, ed. by C. Jones, U. Kirchgraber, and H. O. Walther,
pp. 203–269. Springer-Verlag.

24
Working Papers of the Institute for Empirical Research in Economics

No.
1. Rudolf Winter-Ebmer and Josef Zweimüller: Firm Size Wage Differentials in Switzerland: Evidence from Job
Changers, February 1999
2. Bruno S. Frey and Marcel Kucher: History as Reflected in Capital Markets: The Case of World War II, February
1999
3. Josef Falkinger, Ernst Fehr, Simon Gächter and Rudolf Winter-Ebmer: A Simple Mechanism for the Efficient
Provision of Public Goods – Experimental Evidence, February 1999
4. Ernst Fehr and Klaus M. Schmidt: A Theory of Fairness, Competition and Cooperation, April 1999
5. Markus Knell: Social Comparisons, Inequality, and Growth, April 1999
6. Armin Falk and Urs Fischbacher: A Theory of Reciprocity, July 2000
7. Bruno S. Frey and Lorenz Goette: Does Pay Motivate Volunteers?, May 1999
8. Rudolf Winter-Ebmer and Josef Zweimüller: Intra-firm Wage Dispersion and Firm Performance, May 1999
9. Josef Zweimüller: Schumpeterian Entrepreneurs Meet Engel’s Law: The Impact of Inequality on Innovation-
Driven Growth, May 1999
10. Ernst Fehr and Simon Gächter: Cooperation and Punishment in Public Goods Experiments, June 1999
11. Rudolf Winter-Ebmer and Josef Zweimüller: Do Immigrants Displace Young Native Workers: The Austrian
Experience, June 1999
12. Ernst Fehr and Jean-Robert Tyran: Does Money Illusion Matter?, June 1999
13. Stefan Felder and Reto Schleiniger: Environmental Tax Reform: Efficiency and Political Feasibility, July 1999
14. Bruno S. Frey: Art Fakes – What Fakes?, An Economic View, July 1999
15. Bruno S. Frey and Alois Stutzer: Happiness, Economy and Institutions, July 1999
16. Urs Fischbacher, Simon Gächter and Ernst Fehr: Are People Conditionally Cooperative? Evidence from a Public
Goods Experiment, July 2000
17. Armin Falk, Ernst Fehr and Urs Fischbacher: On the Nature of Fair Behavior, August 1999
18. Vital Anderhub, Simon Gächter and Manfred Königstein: Efficient Contracting and Fair Play in a Simple
Principal-Agent Experiment, September 2000
19. Simon Gächter and Armin Falk: Reputation or Reciprocity?, September 1999
20. Ernst Fehr and Klaus M. Schmidt: Fairness, Incentives, and Contractual Choices, September 1999
21. Urs Fischbacher: z-Tree - Experimenter’s Manual, September 1999
22. Bruno S. Frey and Alois Stutzer: Maximising Happiness?, October 1999
23. Alois Stutzer: Demokratieindizes für die Kantone der Schweiz, October 1999
24. Bruno S. Frey: Was bewirkt die Volkswirtschaftslehre?, October 1999
25. Bruno S. Frey, Marcel Kucher and Alois Stutzer: Outcome, Process & Power in Direct Democracy, November
1999
26. Bruno S. Frey and Reto Jegen: Motivation Crowding Theory: A Survey of Empirical Evidence, November 1999
27. Margit Osterloh and Bruno S. Frey: Motivation, Knowledge Transfer, and Organizational Forms, November 1999
28. Bruno S. Frey and Marcel Kucher: Managerial Power and Compensation, December 1999
29. Reto Schleiniger: Ecological Tax Reform with Exemptions for the Export Sector in a two Sector two Factor Model,
December 1999
30. Jens-Ulrich Peter and Klaus Reiner Schenk-Hoppé: Business Cycle Phenomena in Overlapping Generations
Economies with Stochastic Production, December 1999
31. Josef Zweimüller: Inequality, Redistribution, and Economic Growth, January 2000
32. Marc Oliver Bettzüge and Thorsten Hens: Financial Innovation, Communication and the Theory of the Firm,
January 2000
33. Klaus Reiner Schenk-Hoppé: Is there a Golden Rule for the Stochastic Solow Growth Model? January 2000
34. Ernst Fehr and Simon Gächter: Do Incentive Contracts Crowd out Voluntary Cooperation? February 2000
35. Marc Oliver Bettzüge and Thorsten Hens: An Evolutionary Approach to Financial Innovation, July 2000
36. Bruno S. Frey: Does Economics Have an Effect? Towards an Economics of Economics, February 2000
37. Josef Zweimüller and Rudolf Winter-Ebmer: Firm-Specific Training: Consequences for Job-Mobility, March 2000

The Working Papers of the Institute for Empirical Research in Economics can be downloaded in PDF-format from
http://www.unizh.ch/iew/wp/
Institute for Empirical Research in Economics, Blümlisalpstr. 10, 8006 Zurich, Switzerland
Phone: 0041 1 634 37 05 Fax: 0041 1 634 49 07 E-mail: [email protected]
Working Papers of the Institute for Empirical Research in Economics

No.
38. Martin Brown, Armin Falk and Ernst Fehr: Contract Inforcement and the Evolution of Longrun Relations, March
2000
39. Thorsten Hens, Jörg Laitenberger and Andreas Löffler: On Uniqueness of Equilibria in the CAPM, July 2000
40. Ernst Fehr and Simon Gächter: Fairness and Retaliation: The Economics of Reciprocity, March 2000
41. Rafael Lalive, Jan C. van Ours and Josef Zweimüller: The Impact of Active Labor Market Programs and Benefit
Entitlement Rules on the Duration of Unemployment, March 2000
42. Reto Schleiniger: Consumption Taxes and International Competitiveness in a Keynesian World, April 2000
43. Ernst Fehr and Peter K. Zych: Intertemporal Choice under Habit Formation, May 2000
44. Ernst Fehr and Lorenz Goette: Robustness and Real Consequences of Nominal Wage Rigidity, May 2000
45. Ernst Fehr and Jean-Robert Tyran: Does Money Illusion Matter? REVISED VERSION, May 2000
46. Klaus Reiner Schenk-Hoppé: Sample-Path Stability of Non-Stationary Dynamic Economic Systems, Juni 2000
47. Bruno S. Frey: A Utopia? Government without Territorial Monopoly, June 2000
48. Bruno S. Frey: The Rise and Fall of Festivals, June 2000
49. Bruno S. Frey and Reto Jegen: Motivation Crowding Theory: A Survey of Empirical Evidence, REVISED
VERSION, June 2000
50. Albrecht Ritschl and Ulrich Woitek: Did Monetary Forces Cause the Great Depression? A Bayesian VAR Analysis
for the U.S. Economy, July 2000
51. Alois Stutzer and Rafael Lalive: The Role of Social Work Norms in Job Searching and Subjective Well-Being, July
2000
52. Iris Bohnet, Bruno S. Frey and Steffen Huck: More Order with Less Law: On Contract Enforcement, Trust, and
Crowding, July 2000
53. Armin Falk and Markus Knell: Choosing the Joneses: On the Endogeneity of Reference Groups, July 2000
54. Klaus Reiner Schenk-Hoppé: Economic Growth and Business Cycles: A Critical Comment on Detrending Time
Series, August 2000
55. Armin Falk, Ernst Fehr and Urs Fischbacher: Appropriating the Commons – A Theoretical Explanation, September
2000
56. Bruno S. Frey and Reiner Eichenberger: A Proposal for a Flexible Europe, August 2000
57. Reiner Eichenberger and Bruno S. Frey: Europe’s Eminent Economists: A Quantitative Analysis, September 2000
58. Bruno S. Frey: Why Economists Disregard Economic Methodology, September 2000
59. Armin Falk, Ernst Fehr, Urs Fischbacher: Informal Sanctions, September 2000
60. Rafael Lalive: Did we Overestimate the Value of Health?, October 2000
61. Matthias Benz, Marcel Kucher and Alois Stutzer: Stock Options: the Managers’ Blessing. Institutional Restrictions
and Executive Compensation, October 2000
62. Simon Gächter and Armin Falk: Work motivation, institutions, and performance, October 2000
63. Armin Falk, Ernst Fehr and Urs Fischbacher: Testing Theories of Fairness – Intentions Matter, September 2000
64. Ernst Fehr and Klaus Schmidt: Endogenous Incomplete Contracts, November 2000
65. Klaus Reiner Schenk-Hoppé and Björn Schmalfuss: Random fixed points in a stochastic Solow growth model,
November 2000
66. Leonard J. Mirman and Klaus Reiner Schenk-Hoppé: Financial Markets and Stochastic Growth, November 2000
67. Klaus Reiner Schenk-Hoppé: Random Dynamical Systems in Economics, December 2000
68. Albrecht Ritschl: Deficit Spending in the Nazi Recovery, 1933-1938: A Critical Reassessment, December 2000
69. Bruno S. Frey and Stephan Meier: Political Economists are Neither Selfish nor Indoctrinated, December 2000
70. Thorsten Hens and Beat Pilgrim: The Transfer Paradox and Sunspot Equilibria, January 2001
71. Thorsten Hens: An Extension of Mantel (1976) to Incomplete Markets, January 2001
72. Ernst Fehr, Alexander Klein and Klaus M. Schmidt: Fairness, Incentives and Contractual Incompleteness,
February 2001
73. Reto Schleiniger: Energy Tax Reform with Excemptions for the Energy-Intensive Export Sector, February 2001
74. Thorsten Hens and Klaus Schenk-Hoppé: An Evolutionary Portfolio Theory, May 2001

The Working Papers of the Institute for Empirical Research in Economics can be downloaded in PDF-format from
http://www.unizh.ch/iew/wp/
Institute for Empirical Research in Economics, Blümlisalpstr. 10, 8006 Zürich, Switzerland
Phone: 0041 1 634 37 05 Fax: 0041 1 634 49 07 E-mail: [email protected]