Dynamic Trading with Predictable Returns and

Transaction Costs∗
Nicolae Gârleanu and Lasse Heje Pedersen†
Current Version: March 27, 2009

Abstract

This paper derives in closed form the optimal dynamic portfolio policy when trading
is costly and security returns are predictable by signals with different mean reversion
speeds. The optimal updated portfolio is a linear combination of the existing portfolio,
the optimal current portfolio absent trading costs, and the optimal portfolio based
on future expected returns. Predictors with slower mean reversion (alpha decay) get
more weight since they lead to a favorable positioning both now and in the future.
We implement the optimal policy for commodity futures and show that the resulting
portfolio has superior returns net of trading costs relative to more naive benchmarks.
Finally, we derive natural equilibrium implications, including that demand shocks with
faster mean reversion command a higher return premium.

∗
We are grateful for helpful comments from Pierre Collin-Dufresne, Esben Hedegaard, Ananth Madhavan
(discussant), and Humbert Suarez, as well as from seminar participants at Stanford Graduate School of
Business, University of California at Berkeley, and the Journal of Investment Management Conference.
†
Gârleanu is at Haas School of Business, University of California, Berkeley, NBER, and CEPR;
e-mail: garleanu@[Link]. Pedersen (corresponding author) is at New York Univer-
sity, NBER, and CEPR, 44 West Fourth Street, NY 10012-1126; e-mail: lpederse@[Link],
[Link]
1 Introduction
Active investors and asset managers — such as hedge funds, mutual funds, and proprietary
traders — try to predict security returns and trade to profit from their predictions. Such
dynamic trading often entails significant turnover and trading costs. Hence, any active
investor must constantly weigh the expected excess returns from trading against its risk and
trading costs. An investor often uses different return predictors, e.g., value and momentum
predictors, and these have different prediction strengths and mean-reversion speeds, or, said
differently, different “alphas” and “alpha decays.” The alpha decay is important because it
determines how long time the investor can enjoy high expected returns and, therefore, affects
the trade-off between returns and transactions costs. For instance, while a momentum signal
may predict that the IBM stock return will be high over the next month, a value signal might
predict that Cisco will perform well over the next year. The optimal trading strategy must
consider these dynamics.
This paper addresses how the optimal trading strategy depends on securities’ current
expected returns, the evolution of expected returns in the future, their risks and correlations,
and their trading costs. We present a closed-form solution for the optimal portfolio re-
balancing rule taking these considerations into account.
The optimal trading strategy is intuitive: The best new portfolio is a combination of 1)
the current portfolio (to reduce turnover), 2) the optimal portfolio in the absence of trading
costs (to get part of the best current risk-return trade-off), and 3) the expected optimal
portfolio in the future (a dynamic effect). Said differently, the best portfolio is a weighted
average of the current portfolio and a “target portfolio” that combines portfolios 2) and 3).
Consistent with this decomposition, an investor facing transaction costs trades more
aggressively on persistent signals than on fast mean-reverting signals: the benefits from the
former accrue over longer periods, and are therefore larger. As is natural, transaction costs
inhibit trading, both currently and in the future. Thus, target portfolios are conservative
given the signals, and trading towards the target portfolio is slower when transaction costs
are large.

2
 The key role played by each return predictor’s mean reversion is an important implication
of our model. It arises because transaction costs imply that the investor cannot easily change
his portfolio and, therefore, must consider his optimal portfolio both now and in the future.
In contrast, absent transaction costs, the investor can re-optimize at no cost and needs to
consider only the current investment opportunities (and possible hedging demands) without
regard to alpha decay.
We first solve the model in discrete time. One may wonder, however, whether the discrete-
time trading interval is important for the model, how different models with different time
intervals fit together, and what happens as the trading intervals approach zero, that is, with
continuous trading. To answer these questions, we present a continuous-time version of the
model and show how the discrete-time solutions approach the continuous-time solution. An
additional benefit of the continuous-time model is that the solution is even simpler, making
applications of the model even easier.
As one such application, we embed the continuous-time model in an equilibrium setting.
Rational investors facing transaction costs trade with several groups of noise traders who
provide a time-varying excess supply or demand of assets. We show that, in order for the
market to clear, the investors must be offered return premia depending on the properties of
the noise-traders’ positions. In particular, the noise trader positions that mean revert more
quickly generate larger alphas in equilibrium, as the rational investors must be compensated
for incurring more transaction costs per time unit. Long-lived supply fluctuations, on the
other hand, give rise to smaller and more persistent alphas.
Finally, we illustrate our results empirically in the context of commodity futures mar-
kets. We use returns over the past 5 days, 12 months, and 5 years to predict returns. The
5-day signal is quickly mean reverting (fast alpha decay), whereas the 5-year signal is per-
sistent, and the 12-month signal is naturally in between. We calculate the optimal dynamic
trading strategy taking transaction costs into account and compare its performance to the
optimal portfolio ignoring transaction costs and to a class of strategies that makes a static
(one-period) transaction-cost optimization. Our optimal portfolio performs the best net of

3
transaction costs among all the alternatives that we consider. Its net Sharpe ratio is about
20% better than that of the best strategy among all the static strategies. Our strategy’s
superior performance is achieved by trading at an optimal speed and by trading towards a
target portfolio that is optimally tilted towards the more persistent return predictors.
We also study the impulse-response of the security positions following a shock to return
predictors. While the no-transaction-cost position immediately jumps up and mean reverts
with the speed of the alpha decay, the optimal position increases more slowly to minimize
trading costs and, depending on the alpha decay speed, may eventually become larger than
the no-transaction-cost position as the optimal position is sold more slowly.
Our paper is related to several large strands of literature. First, a large literature stud-
ies portfolio selection with return predictability in the absence of trading costs (see, e.g.,
Campbell and Viceira (2002) and references therein). A second strand of literature derives
the optimal trade execution, treating what to trade as given exogenously (see, e.g., Perold
(1988), Bertsimas and Lo (1998), Almgren and Chriss (2000), and Engle and Ferstenberg
(2007)). A third strand of literature, starting with Constantinides (1986), considers the
optimal allocation of capital between cash and a single risky security in settings with i.i.d.
returns with trading costs.1 Without trading costs the optimal policy is to keep a con-
stant fraction of wealth in the risky asset, but since this requires trading, with trading costs
the optimal solution keeps the fraction within a bound. While a partial equilibrium model,
Constantinides (1986) also considers trading-cost implications for the equity premium. Equi-
librium models with trading costs include Amihud and Mendelson (1986), Vayanos (1998),
Vayanos and Vila (1999), Lo, Mamaysky, and Wang (2004), Gârleanu (2009), and Acharya
and Pedersen (2005) who consider an equilibrium model with time-varying trading costs.
Liu (2004) determines the optimal trading strategy for an investor with constant absolute
risk aversion (CARA) and many independent securities with both fixed and proportional
1
Davis and Norman (1990) provide a more formal analysis of Constantinides’ model. Also, Gârleanu
(2009) and Lagos and Rocheteau (2006) show how search frictions and payoff mean-reversion impact how
close one trades to the static portfolio. Our continuous-time model with with bounded-variation trading
shares features with Longstaff (2001) and, in the context of predatory trading, by Brunnermeier and Pedersen
(2005) and Carlin, Lobo, and Viswanathan (2008).

4
costs. The assumptions of CARA and independence across securities imply that the opti-
mal position for each security is independent of the positions in the other securities. Using
calibrated numerical solutions, Heaton and Lucas (1996) consider incomplete markets, Jang,
Koo, Liu, and Loewenstein (2007) introduce a time-varying investment opportunity set, and
Lynch and Tan (2008) consider return predictability, wealth shocks, and state-dependent
transaction costs. Our model is most similar to that of Grinold (2006), who derives the
optimal steady-state position with quadratic trading costs and a single predictor of returns
per security.
We contribute to the literature in several ways. We provide a closed-form solution for
a model with multiple correlated securities and multiple return predictors with different
mean-reversion speeds, derive new equilibrium implications, and demonstrate the model’s
empirical importance using real data.
We end our discussion of the related literature by noting that quadratic programming
techniques are also used in macroeconomics and other fields, and, usually, the solution comes
down to algebraic matrix Riccati equations (see, e.g., Ljungqvist and Sargent (2004) and
references therein). We solve our model explicitly, including the Riccati equations, in both
discrete and continuous time.
The paper is organized as follows. Section 2 lays out a general discrete-time model, pro-
vides a closed-form solution, and various results and examples. Section 3 solves the analogous
continuous-time model and shows how it is approached by the discrete-time model as the
time between periods becomes small. Section 4 studies the model’s equilibrium implications.
Section 5 applies our framework to a trading strategy for commodity futures, and Section 6
concludes.

2 Discrete-Time Model
We first present the model, then solve it and provide additional results and examples.

5
2.1 General Discrete-Time Framework

We consider an economy with S securities traded at each time t = 1, 2, 3, .... The securities’
price changes between times t and t + 1, pt+1 − pt , are collected in a vector rt+1 given by

rt+1 = µt + αt + ut+1 , (1)

where µt is the “fair return,” e.g., from the CAPM, ut+1 is an unpredictable noise term
with variance vart (ut+1 ) = Σ, and αt (alpha) is the predictable excess return known to the
investor already at time t, and given by

αt = Bft (2)

∆ft+1 = −Φft + εt+1 . (3)

Here, f is a K ×1 vector of factors that predict returns, B is a S×K matrix of factor loadings,
Φ is a K ×K positive-definite matrix of mean-reversion coefficients for the factors, and εt+1 is
the shock affecting the predictors with variance vart (εt+1 ) = Ω. Naturally, ∆ft+1 = ft+1 − ft
is the change in the factors.
The interpretation of these assumptions is straightforward: the investor analyzes the se-
curities and his analysis results in forecasts of excess returns. The most direct interpretation
is that the investor regresses the return on security s on factors f which could be past returns
over various horizons, valuation ratios, and other return-predicting variables:

X
s
rt+1 = µst + β sk ftk + ust+1 , (4)
k

and thus estimates each variable’s ability to predict returns as given by β sk (collected in the
matrix B). Alternatively, one can think of the factors as an analyst’s overall assessment of
a security (possibly based on a range of qualitative information) and B as the strength of
these assessments in predicting returns.
We note that each factor f k can in principle predict each security, but this is can be easily

6
simplified to the special case in which there are different factors for different securities, as
we discuss in Example 2 below. We further note that Equation (3) means that the factors
and alphas mean-revert to zero. This is a natural assumption since an excess return that is
always present should be viewed as compensation for risk, not reward for security analysis.
Hence, such average returns are part of the fair return µt . That said, an intercept term can
be accommodated, e.g., as a constant factor ft1 = 1.
Trading is costly in this economy and the transaction cost (T C) associated with trading
∆xt = xt − xt−1 shares is given by

1
T C(∆xt ) = ∆x> Λ∆xt , (5)
2 t

where Λ is a symmetric positive-definite matrix measuring the level of trading costs.2 Trading
costs of this form can be thought of as follows. Trading ∆xt shares moves the (average) price
by 12 Λ∆xt , and this results in a total trading cost of ∆xt times the price move, which gives
T C. Hence, Λ (actually 1/2 Λ for convenience) is a multi-dimensional version of Kyle’s
lambda.
The investor’s objective is to choose the dynamic trading strategy (x0 , x1 , ...) to maximize
the present value of all future expected alphas, penalized for risks and trading costs:
"  #
X γ > 1 >
max E0 (1 − ρ)t x>
t αt − xt Σxt − ∆xt Λ∆xt , (6)
x0 ,x1 ,...
t
2 2

where ρ ∈ (0, 1) is a discount factor, and γ is the risk aversion coefficient.

There are several natural interpretations of this objective. First, we can envision an
investor who is compensated based on his performance relative to a benchmark b. Under
this interpretation, xt is the deviation of the total portfolio, which we can denote by x∗ ,
from the benchmark portfolio, that is, xt = x∗t − b. Hence, x>
t α measures the excess return

over the benchmark, and x>

t Σxt measures the variance of the tracking error relative to the
2
The assumption that Λ is symmetric is without loss of generality. To see this, suppose that T C(∆xt ) =
>
1
2 ∆xt Λ̄∆xt ,where Λ̄ is not symmetric. Then, letting Λ be the symmetric part of Λ̄, i.e., Λ = (Λ̄ + Λ̄> )/2,
Λ generates the same trading costs as Λ̄.

7
benchmark.
Another interpretation is a hedge fund manager who cares about his total net return (so
xt is the total portfolio), but is committed to achieving “alpha” (i.e., as opposed to making
constant bets based on constant risk premia).
A final interpretation is a “standard” investor who considers the return of his total
portfolio (i.e., not just the alpha above the fair return µ or over a benchmark). Under this
interpretation, xt is the total portfolio, and we eliminate µ from (4), letting instead the first
factor be constant, ft1 = 1, and β s1 = µs . Thus, β s1 ft1 = µs captures the average security
returns. It follows that α incorporates the entire return in the objective function and, in fact,
this objective can be justified in a standard set-up with exponential utility for consumption
and normally-distributed price changes, under certain conditions.
These interpretations are linked naturally. To see this, we can take the final interpretation
and find an optimal “total portfolio” x∗t (i.e., the solution with ft1 = 1, β s1 = µs , and no µ
term in 4), and compare it to the alpha-maximizing portfolio xt from the first interpretation
(i.e., the solution of (6) as is). It can be shown that, if x∗0 = (γΣ)−1 µ + x0 , then x∗t =
(γΣ)−1 µ + xt for all t, that is, the total-return-maximizing portfolio x∗t is the “benchmark”
plus the optimal deviation xt from this benchmark, where the benchmark is given as the
constant Markowitz portfolio relative to the average returns µ.

2.2 Solution and Results

We solve the model using dynamic programming. We start by introducing a value function
V (xt−1 , ft ) measuring the value of entering period t with a portfolio of xt−1 securities and
observing return-predicting factors ft . The value function solves the Bellman equation:
 
γ > 1 >
V (xt−1 , ft ) = max x>
t αt − xt Σxt − ∆xt Λ∆xt + (1 − ρ)Et [V (xt , ft+1 )] . (7)
xt 2 2

We guess, and later verify, that the solution has the following quadratic form:

1 1 >
V (xt , ft+1 ) = − x> >
t Axx xt + xt Axf ft+1 + ft+1 Af f ft+1 + a0 , (8)
2 2

8
where we need to derive the scalar a0 , the symmetric matrices Axx and Af f , and the matrix
Axf .
The model in its most general form can be solved explicitly as we state in the following
proposition. The expressions for the coefficient matrices (Axx , Axf ) are somewhat long, so
we leave them in the Appendix, but they become simple in the special cases discussed below,
and, in continuous time, they are relatively simple even in the most general case.

Proposition 1 The optimal dynamic portfolio xt is a “matrix weighted average” of the

current position and a target portfolio:

xt = (I − Λ−1 Axx )xt−1 + Λ−1 Axx target t , (9)

with

target t = A−1
xx Axf ft . (10)

The matrix Axx is positive definite; Axx and Axf are stated explicitly in (A.14) and (A.20).
An alternative characterization of the optimal portfolio is a weighted average of the
current position, xt−1 , the current optimal position in the absence of transaction costs,
(γΣ)−1 Bft , and the expected target next period, Et (target t+1 ) = A−1
xx Axf (I − Φ)ft :

xt = [Λ + γΣ + (1 − ρ)Axx ]−1

× Λxt−1 + γΣ (γΣ)−1 Bft + (1 − ρ)Axx A−1




xx Axf (I − Φ)ft . (11)

The proposition provides expressions for the optimal portfolio that are natural and relatively
simple. The optimal trade ∆xt follows directly from the proposition as

∆xt = Λ−1 Axx (target t − xt ) . (12)

The optimal trade is proportional to the difference between the current portfolio and the
target portfolio, and the trading speed decreases in the trading cost Λ.
We discuss the intuition behind the result further under the additional assumption that

9
Λ = λΣ for some number 0 < λ ∈ R, which simplifies the solution further. This means
that the trading-cost matrix is proportional to the return variance-covariance matrix. This
trading cost is natural and, in fact, implied by the model of dealers in Gârleanu, Pedersen,
and Poteshman (2008). To understand this, suppose that a dealer takes the other side of
the trade ∆xt for a single period and can “lay it off” thereafter, and that alpha is zero
conditional on the dealer’s information. Then the dealer’s risk is ∆x>
t Σ∆xt and the trading

costs is the dealer’s compensation for risk, depending on the dealer’s risk aversion λ. Under
this assumption, we derive the following simple and intuitive optimal trading strategy.

Proposition 2 When the trading cost is proportional to the amount of risk, Λ = λΣ, then
the optimal new portfolio xt is a weighted average of the current position xt−1 and a moving
“target portfolio”
 a a
xt = 1− xt−1 + target t (13)
λ λ
a
where λ
< 1 and
 −1
−1a(1 − ρ)
target t = (γΣ) B I + Φ ft (14)
γ
p
−(γ + λρ) + (γ + λρ)2 + 4γλ(1 − ρ)
a = (15)
2(1 − ρ)

The target is the current optimal position in the absence of trading costs if the return-
 −1
predictability coefficients were B I + a(1−ρ)
γ
Φ instead of B.
Alternatively, xt is a weighted average of the current position, xt−1 , the current optimal
position in the absence of trading costs, static t = (γΣ)−1 Bft , and the expected target in the
 −1
−1 a(1−ρ)
future, Et (target t+1 ) = (γΣ) B I + γ Φ (I − Φ)ft :

λ γ (1 − ρ)a
xt = xt−1 + static t + Et (target t+1 ). (16)
λ + γ + (1 − ρ)a λ + γ + (1 − ρ)a λ + γ + (1 − ρ)a

This result provides a simple and appealing trading rule. Equation (13) states that the
optimal portfolio is between the existing one and an optimal target, where the weight on the

10
target a/λ decreases in trading costs λ because higher trading costs imply that one must
trade more slowly. The weight on the target increases in γ because a higher risk aversion
means that it is more important not to let one’s position stray too far from its optimal level.
The alternative characterization (16) provides a similar intuition and comparative statics,
and separates the target into the current Markowitz static optimal position without trans-
action costs and the expected future target. We note that while the weights on the current
position xt−1 appear different in (13) and (16), they are, naturally, the same. The optimal
trading policy is illustrated in Figure 1.
The optimal trading is simpler yet under the additional (and rather standard) assumption
that the mean reversion of each factor f k only depends on its own level (not the level of the
other factors), that is, Φ = diag(φ1 , ..., φK ) is diagonal, so that Equation (3) simplifies to
scalars:

k
∆ft+1 = −φk ftk + εkt+1 . (17)

Under these assumptions we have:

Proposition 3 If Λ = λΣ and Φ = diag(φ1 , ..., φK ), then the optimal portfolio is the

weighted average (13) of the current portfolio xt−1 and a target portfolio, which is the optimal
portfolio without trading costs with each factor ftk scaled depending on its alpha decay φk :
>
ft1 ftK

−1
target t = (γΣ) B , . . . , (18)
1 + φ1 (1 − ρ)a/γ 1 + φK (1 − ρ)a/γ

We see that the target portfolio is very similar to the optimal portfolio without transaction
costs (γΣ)−1 Bft . The transaction costs imply first that one optimally only trades part of the
way towards the target, and, second, that the target down-weights each return-predicting
factor more the higher is its alpha decay φk . Down-weighting factors reduces the size of the
position, and, more importantly, changes the relative importance of the different factors as
illustrated in Figure 2. Naturally, giving more weight to the more persistent factors means
that the investor trades towards a portfolio that not only has a high alpha now, but also is

11
expected to have a high alpha for a longer time in the future.
We next provide a few examples.

Example 1: Timing a single security

An interesting and simple case is when there is only one security. This occurs when an
investor is timing his long or short view of a particular security or market. In this case,
the assumption that Λ = λΣ from Propositions 2–3 is without loss of generality since all
parameters are just scalars. In the scalar case, we use the notation Σ = σ 2 and B =
(β 1 , ..., β K ). Assuming that Φ is diagonal, we can apply Proposition 3 directly to get the
optimal timing trade:

K
 a a 1 X 1
xt = 1− xt−1 + 2 i
β i fti . (19)
λ λ γσ i=1 1 + φ (1 − ρ)a/γ

Example 2: Relative-value trades based on security characteristics

It is natural to assume that the agent uses certain characteristics of each security to predict
its returns. Hence, each security has its own return-predicting factors (whereas, in the general
model above, all the factors could influence all the securities). For instance, one can imagine
that each security is associated with a value characteristic (e.g., its own book-to-market)
and a momentum characteristic (its own past return). In this case, it is natural to let the
alpha for security s be given by

X
αts = β i fti,s , (20)
i

where fti,s is characteristic i for security s (e.g., IBM’s book-to-market) and β i be the pre-
dictive ability of characteristic i (i.e., how book-to-market translates into future expected
return, for any security) which is the same for all securities s. Further, we assume that

12
characteristic i has the same mean-reversion speed for each security, that is, for all s,

i,s
∆ft+1 = −φi fti,s + εi,s
t+1 . (21)

 >
We collect the current values of characteristic i for all securities in a vector fti = fti,1 , ..., fti,S ,
e.g., the book-to-market of security 1, book-to-market of security 2, etc.
This setup based on security characteristics is a special case of our general model. To
map it into the general model, we stack all the various characteristic vectors on top of each
other into f :
 
ft1
 . 
 
ft =  ..  . (22)
 
I
ft

Further, we let IS×S be the S-by-S identity matrix and can express B using the kronecker
product:
 
1 I
β 0 0 β 0 0
 
>
B = β ⊗ IS×S = 0 .. ..
. ··· . 0 . (23)
 
0 0
 
0 0 β1 0 0 β I

Thus, αt = Bft . Also, let Φ = diag(φ ⊗ 1S×1 ) = diag(φ1 , ..., φ1 , ..., φI , ..., φI ). With these
definitions, we apply Proposition 3 to get the optimal characteristic-based relative-value
trade as
I
 a a X 1
xt = 1− xt−1 + (γΣ)−1 i
β i fti . (24)
λ λ i=1
1 + φ (1 − ρ)a/γ

Example 3: Static model

When the investor completely discounts the future, i.e., ρ = 1, then he only cares about the

13
current period and the problem is static. The investor simply solves

γ > λ >
max x>
t αt − xt Σxt − ∆xt Σ∆xt (25)
xt 2 2

with a solution that specializes Proposition 2:

λ γ
xt = xt−1 + (γΣ)−1 αt . (26)
γ+λ γ+λ

To recover the optimal dynamic weight on the current position xt−1 from (16), one must
1
lower the trading cost λ to 1+(1−ρ)a/γ
λ to account for the future benefits of the position.
Alternatively, one can increase risk aversion, or do some combination.
Interestingly, however, with multiple return-predicting factors, no choice of risk aversion
γ and trading cost λ recovers the dynamic solution. This is because the static solution treats
all factors the same, while the dynamic solution gives more weight to factors with slower
alpha decay. We show empirically in Section 5 that even the best choice of γ and λ in a
static model performs significantly worse than our dynamic solution.
To recover the dynamic solution in a static setting, one must change not just γ and λ,
but additionally the alphas αt = Bft by changing B as described in Propositions 2–3.

Example 4: Today’s first signal is tomorrow’s second signal

Suppose that the investor is timing a single market using each of the several past daily
returns to predict the next return. In other words, the first signal ft1 is the daily return
for yesterday, the second signal ft2 is the return the day before yesterday, and so on, so
that the last signal used yesterday is ignored today. In this case, the trader already knows
today what some of her signals will look like in the future. Today’s yesterday is tomorrow’s
day-before-yesterday:

1
ft+1 = ε1t+1
k
ft+1 = ftk−1 for k > 1

14
The matrix Φ is therefore not diagonal, but has the form
 
0 0
 
 
 1 0 
I −Φ= .. ..
.
. .
 
 
 
0 1 0

Suppose for simplicity that all signals are equally important for predicting returns B =
(β, ..., β) and use the notation Σ = σ 2 . Then we can use Proposition 2 to get the optimal
trading strategy
a 1
xt = 1−
xt−1 + B((γ + λ + (1 − ρ)a)I − λ(1 − ρ) (I − Φ))−1 ft
λ σ2
 a β X
K+1−k

k
= 1− xt−1 + 1 − z ft
λ σ 2 (γ + λ + (1 − ρ)a) k
 a β(λ − a) X K+1−k

k
= 1− xt−1 + 1 − z ft
λ λ2 σ 2 k

λ(1−ρ)
where z = γ+λ+(1−ρ)a
< 1. Hence, the optimal portfolio gives the largest weight to the
first signal (yesterday’s return), the second largest to the second signal, and so on. This is
intuitive, since the first signal will continue to be important the longest, the second signal
the second longest, and so on.

3 Continuous-Time Model
We next present the continuous-time version of our model. The continuous-time model
is convenient since it has an even simpler solution and, therefore, it provides a powerful
workhorse for further applications, e.g., our equilibrium model. We show below that the
continuous-time model obtains naturally as the limit of discrete-time models.
The securities have prices p with dynamics

dpt = (µt + αt ) dt + dut (27)

15
where, as before, µt is the “fair return,” u is a Brownian motion with drift zero and instan-
taneous variance covariance matrix vart (dut ) = Σdt,3 and alpha is given by

αt = Bft (28)

dft = −Φft dt + dεt . (29)

The vector f contains the factors that predict returns, B contains the factor loadings, Φ is
the matrix of mean-reversion coefficients, and ε is a Brownian motion with drift zero and
instantaneous variance-covariance matrix vart (dεt ) = Ωdt.
The agent chooses his trading intensity τt ∈ RS , which determines the rate of change of
his position xt :

dxt = τt dt. (30)

The cost per time unit of trading τt shares per time unit is

1
T C(τt ) = τt> Λτt (31)
2

and the investor chooses his optimal trading strategy to maximize the present value of the
future stream of alphas, penalized for risk and trading costs:
Z ∞  
−ρ(s−t) > γ > 1 >
max Et e xs αs − xs Σxs − τs Λτs ds. (32)
(τs )s≥t t 2 2

The value function V (x, f ) of the investor solves the Hamilton-Jacoby-Bellman (HJB)
equation

∂ 2V
  
> γ > 1 > ∂V ∂V 1
ρV = sup x Bf − x Σx − τ Λτ + τ+ (−Φf ) + tr Ω . (33)
τ 2 2 ∂x ∂f 2 ∂f ∂f >
3
That is, the quadratic variation is [u]t = Σ t.

16
Maximizing this expression with respect to the trading intensity results in
>
−1 ∂V
τ = Λ .
∂x

It is natural to conjecture a quadratic form for the value function:

1 1
V (x, f ) = − x> Axx x + x> Axf f + f > Af f f + A0 .
2 2

We verify the conjecture as part of the proof to the following proposition.

Proposition 4 The optimal portfolio xt tracks a moving “target portfolio” A−1

xx Axf ft with a

tracking speed of Λ−1 Axx . That is, the optimal trading intensity τt = dxt
dt
is

τt = Λ−1 Axx A−1

xx Axf ft − xt , (34)

where the positive definite matrix Axx and the matrix Axf are given by

ρ 1

− 12 − 12 ρ  21 1
Axx = − Λ + Λ γΛ ΣΛ + I Λ 2
2 (35)
2 4

−1
vec(Axf ) = ρI + Φ ⊗ IK + IS ⊗ (Axx Λ−1 )
>
vec(B). (36)

As in discrete time, the optimal trading strategy has a particularly simple form when
trading costs are proportional to the variance of the fundamentals:

Proposition 5 If trading costs are proportional to the amount of risk, Λ = λΣ, then the
dxt
optimal trading intensity τt = dt
is

a
τt = (target t − xt ) (37)
λ

with
 −1
−1a
target = (γΣ) B I + Φ ft (38)
γ
p
−ρλ + ρ2 λ2 + 4γλ
a = . (39)
2

17
In words, the optimal portfolio xt tracks target t with speed λa . The tracking speed decreases
with the trading cost λ and increases with the risk-aversion coefficient γ.
If each factor’s alpha decay only depends on itself, Φ = diag(φ1 , ..., φK ), then the target
is the optimal portfolio without transaction costs with each return-predicting factor ft down-
weighted more the higher is the trading cost λ and the higher is its alpha decay speed φk :
>
ft1 ftK

−1
target t = (γΣ) B , ..., . (40)
1 + aφ1 /γ 1 + aφK /γ

3.1 Connection between Discrete and Continuous Time

The continuous-time model, and therefore solution, are readily seen to be the limit of their
discrete-time analogues when parameters are chosen consistently, adjusted for the length of
the time interval between successive trading opportunities.

Proposition 6 Consider the discrete-time model of Section 2 with parameters defined to

depend on the time interval ∆t in the following way:

Σ̂(∆t ) = Σ∆t (41)

Ω̂(∆t ) = Ω∆t (42)

Λ̂(∆t ) = ∆t −1 Λ or λ̂(∆t ) = ∆t −2 λ (43)

B̂(∆t ) = B∆t (44)

Φ̂(∆t ) = 1 − e−Φ∆t (45)

ρ̂(∆t ) = 1 − e−ρ∆t (46)

γ̂(∆t ) = γ. (47)

Then, given the initial position x0 , the discrete-time solution converges to the continuous-
time solution as ∆t approaches zero: The optimal discrete-time position converges to the
continuous-time one, i.e., x̂t → xt a.s., as does the optimal trade per time unit, i.e.,
∆x̂t /∆t → τt a.s.

18
We note that Equations (41)–(42) simply state that the variance is proportional to time.
The adjustment to the trading cost in Equation (43) is different for the following reason.
Suppose that one can trade twice as frequently and consider trading over two time periods.
The same total amount as previously can be traded now by splitting the order in half.
With a quadratic trading cost, this leads to a total trading cost over the two periods of
2 · T C(∆x/2) = 2 · T C(∆x)/4 = T C(∆x)/2. Hence, in order for the total trading costs to be
independent of the trading frequency, Λ must double when the frequency doubles, explaining
the equation for Λ. When trading costs are proportional to Σ, the equation for λ simply
follows from the previous analysis and Λ = λΣ.

4 Equilibrium Implications
In this section we study the restrictions placed on a security’s return properties by the market
equilibrium. More specifically, we consider a situation in which an investor facing transaction
costs absorbs a residual supply specified exogenously and analyze the relationship implied
between the characteristics of the supply dynamics and the return alpha.
For simplicity, we consider a model with one security in which L ≥ 1 groups of (exoge-
nously given) noise traders hold positions ztl (net of the aggregate supply) given by

dztl = κ ftl − ztl dt

(48)

dftl = −ψl ftl dt + dWtl . (49)

In addition, the Brownian motions W l satisfy vart (dWtl )/dt = Ωll . It follows that the aggre-
gate noise-trader holding, zt = l ztl , satisfies
P

L
!
X
dzt = κ ftl − zt dt. (50)
l=1

We conjecture that the investor’s inference problem is as studied in Section 3, where f given
by f ≡ (f 1 , ..., f L , z) is a linear return predictor and B is to be determined. We verify the

19
conjecture and find B as part of Proposition 7 below.
Given the definition of f , the mean-reversion matrix Φ is given by
 
ψ1 0 ··· 0
 
···
 
 0 ψ2 0 
Φ =  .. .. .. ..
. (51)
.
 
 . . . 
 
−κ −κ · · · κ

Suppose that the only other investors in the economy are the investors considered in
Section 3, facing transaction costs given by Λ = λσ 2 . In this simple context, an equilibrium
is defined as a price process and market-clearing asset holdings that are optimal for all
agents given the price process. Since the noise traders’ positions are optimal by assumption
as specified by (48)–(49), the restriction imposed by equilibrium is that the dynamics of the
price are such that, for all t,

xt = −zt (52)

dxt = −dzt . (53)

Using (37), these equilibrium conditions lead to

a −2 a
σ B(aΦ + γI)−1 + eL+1 = −κ(1 − 2eL+1 ), (54)
λ λ

where eL+1 = (0, · · · , 0, 1) ∈ RL+1 and 1 = (1, · · · , 1) ∈ RL+1 . It consequently follows that,
if the investor is to hold −ftL at time t for all t, then the factor loadings must be given by
 
2λ
B = σ − κ(1 − 2eL+1 ) − eL+1 (aΦ + γI)
a
 γ 
= σ 2 [−λκ(1 − 2eL+1 ) − aeL+1 ] Φ + I . (55)
a

20
For l ≤ L, we calculate Bl further as

Bl = −σ 2 κ(λψk + λγa−1 + λκ − a)

= −λσ 2 κ(ψk + ρ + κ), (56)

while

BL+1 = σ 2 (ρλκ + λκ2 − γ). (57)

Using this, it is straightforward to see the following key equilibrium implications:

Proposition 7 The market is in equilibrium if and only if x0 = −z0 and the security’s alpha
is given by

L
X
αt = λσ 2 κ(ψl + ρ + κ)(−ftl ) + σ 2 (ρλκ + λκ2 − γ)zt . (58)
l=1

The coefficients λσ 2 κ(ψk + ρ + κ) are positive and increase in the mean-reversion parameters
ψk and κ and in the trading costs λσ 2 . In other words, noise trader selling (ftk < 0) increases
the alpha, and especially so if its mean reversion is faster and if the trading cost is larger.

Naturally, noise trader selling increases the expected excess return (alpha), while their
buying lowers the alpha since the arbitrageurs need to be compensated to take the other side
of the trade. Interestingly, the effect is larger when trading costs are larger and for noise-
trader shocks with faster mean reversion because such shocks are associated with larger
trading costs for the arbitrageurs.

5 Application: Dynamic Trading of Commodity Fu-

tures
In this section we apply our framework to trading commodity futures using real data.

21
5.1 Data

We consider 15 different liquid commodity futures, which do not have tight restrictions on
the size of daily price moves (limit up/down). In particular, we collect data on Aluminum,
Copper, Nickel, Zinc, Lead, Tin from London Metal Exchange (LME), Gas Oil from the
Intercontinental Exchange (ICE), WTI Crude, RBOB Unleaded Gasoline, Natural Gas is
from New York Mercantile Exchange (NYMEX), Gold, Silver is from New York Commodities
Exchange (COMEX), and Coffee, Cocoa, Sugar from New York Board of Trade (NYBOT).
(This excludes futures on various agriculture and livestock that have tight price limits.)
We consider the sample period from 01/01/1996 – 2009/01/23 where we have data on
all commodities.4 Every day, we compute for each commodity the price change of the most
liquid futures contract (among the available contract maturities), and normalize the series
such that each commodity’s price changes have annualized volatility of 10%. We abstract
from the cost of rolling from one futures contract to the next. (In the real world, there is a
separate roll market with a small transaction costs, far smaller than the cost of independently
selling the “old” contract and buying the “new” one.)

5.2 Predicting Returns and Other Parameter Estimates

We use the characteristic-based model described in Example 2 in Section 2, where each

commodity characteristic is its own past returns at various horizons. Hence, to predict
returns, we run a pooled panel regression:

s
rt+1 = 0.000 + 0.011 ft5D,s + 0.037 ft1Y,s − 0.015 ft5Y,s +ust+1
(59)
(−0.02) (1.4) (4.6) (−1.85)

where the left hand side is the commodity price changes and the right hand side has the
return predictors: f 5D is the average past 5 days’ price changes, divided by the past month’s
standard deviation of price changes, f 1Y is the past year’s average price change divided by
4
Our return predictors uses moving averages of price data lagged up to five years, which is available for
most commodities except some of the LME base metals. In the early sample when some futures do not have
a complete lagged price series we simply average we have.

22
the past year’s average standard deviation, and f 5Y is the same over the past 5 years. We
report the OLS standard errors in brackets.
We see that price changes show continuation at short and medium frequencies and re-
versal over long horizons.5 The goal is to see how an investor could optimally trade on this
information, taking transaction costs into account. Of course, these (in-sample) regression
results are only available now and a more realistic analysis would consider rolling out-of-
sample regressions. However, using the in-sample regression allows us to focus on portfolio
optimization. (Indeed, taking out-of-sample estimation errors into account would add noise
to the evaluation of the optimization gains of our method.)
The return predictors are chosen so that they have very different mean reversion:

5D,s
∆ft+1 = −0.1977ft5D,s + εt+1
5D,s
(60)
1Y,s
∆ft+1 = −0.0034ft1Y,s + ε1Y,s
t+1 (61)
5Y,s
∆ft+1 = −0.0010ft5Y,s + ε5Y,s
t+1 (62)

These mean reversion rates correspond to a 3-day half life of the 5-day signal, a 205-day half
life of the 1-year signal, and a 701-day signal of the 5-year signal.6
We estimate the variance-covariance matrix Σ using daily price changes over the full
sample. We set the absolute risk aversion to γ = 10−9 which we can think of as corresponding
to a relative risk aversion of 1 for an agent with 1 billion dollars under management. We set
the time discount rate to ρ = 1 − exp(−0.02/260) corresponding to a 2 percent annualized
rate. Finally, we set the transaction cost matrix to Λ = λΣ, where we consider λ = 5 × 10−7
as well as a higher λ which is double that.
5
Asness, Moskowitz, and Pedersen (2008) document 12-month momentum and 5-year reversals of com-
modities and other securities. These result are robust and holds both for price changes and returns. The
5-day momentum is less robust. For instance, for certain specifications using percent returns, the 5-day
coefficient switches sign to reversal. This robustness is not important for our study due to our focus on
optimal trading rather than out-of-sample return predictability.
6
The half life is the time it is expected to take for half the signal to disappear. It is computed as
log(0.5)/ log(1 − .1977) for the 5-day signal.

23
5.3 Dynamic Portfolio Selection with Trading Costs

We consider three different trading strategies: the optimal trading strategy given by Equa-
tion (24) (“optimal”), the optimal trading strategy in the absence of transaction costs
(“no-TC”), and the optimal portfolio in a static (i.e., one-period) model with transaction
costs given by Equation (26) (“static”). For the static portfolio we use a modified λ such
that the coefficient on xt−1 is the same as for the optimal portfolio which is numerically
almost the same as choosing λ to maximize the static portfolio’s net Sharpe Ratio.
The performance of each strategy as measured by the Sharpe Ratio (SR) is reported in
Table 1. The cumulative excess return of each strategy scaled to 10% annualized volatility is
depicted in Figure 3, and Figure 4 shows the cumulative net returns. We see that the highest
SR before transaction cost is naturally achieved by the no-TC strategy, and the optimal and
static portfolios have similar drops in gross SR due to their slower trading. After transaction
costs, however, the optimal portfolio is the best, significantly better than the best possible
static strategy, and the no-TC strategy incurs enormous trading costs.
It is interesting to consider the driver of the superior performance of the optimal dynamic
trading strategy relative to the best possible static strategy. The key to the out-performance
is that the dynamic strategy gives less weight to the 5-day signal because of its fast alpha
decay. The static strategy simply tries to control the overall trading speed, but this is not
sufficient: trading fast implies large transaction costs, especially due to the 5-day signal.
Trading slowly means that the strategy does not capture the alpha. The dynamic strategy
overcomes this by trading somewhat fast, but trading mainly towards the more persistent
signals.
To illustrate the difference in the positions of the different strategies, Figure 5 shows the
positions over time of two of the commodity futures, namely Crude and Gold. We see that
the optimal portfolio is a much more smooth version of the no-TC strategy thus reducing
trading costs.

24
5.4 Response to New Information

It is instructive to trace the response to a shock to the return predictors, namely to εi,s
t

in Equation (60). Figure 6 shows the responses to, respectively, shocks to each return-
predictor factor, namely the 5-day factor (i = 1), the 1-year factor (i = 2), and the 5-year
factor (i = 3).
The first panel shows that the no-TC strategy immediately jumps up after a shock to the
5-day factor and slowly mean reverts as the alpha decays. The optimal strategy trades much
slower and never accumulates nearly as large of a position. Interestingly, since the optimal
position also trades more slowly out of the position as the alpha decays, the lines cross as
the optimal strategy eventually has a larger position than the no-TC strategy.
The second panel shows the response to the 1-year factor. The no-TC jumps up and
decays, whereas the optimal position increases more smoothly and catches up as the no-TC
starts to decay. The third panel shows the same for the 5Y signal, except that the effects
are slower and with opposite sign since 5-year returns predict future reversals.

6 Conclusion
This paper provides a highly tractable framework for studying optimal trading strategies in
light of various return predictors, risk and correlation considerations, as well as transaction
costs. We derive an explicit closed-form solution for the optimal trading policy and several
useful and intuitive results arise. The optimal portfolio tracks a “target portfolio,” which is
analogous to the optimal portfolio in the absence of trading costs in its tradeoff between risk
and return, but different since more persistent return predictors are weighted more heavily
relative to return predictors with faster alpha decay. The optimal strategy is not to trade
all the way to the target portfolio, since this entails too high transaction costs. Instead, it is
optimal to take a smoother and more conservative portfolio that moves in the direction of
the target portfolio while limiting turnover.
Our framework provides a powerful tool to optimally combine various return predictors

25
taking into account their evolution over time, their decay rate, and their correlation, and
trading their benefits off against risks and transaction costs. Such trade-offs are at the heart
of the decisions of “arbitrageurs” that help make markets efficient as per the efficient market
hypothesis. Arbitrageurs’ ability to do so is limited, however, by transaction costs, and
our framework constitutes a powerful tool for the study of the dynamic implications of this
limitation.
We illustrate this feature by embedding our setting in an equilibrium model with several
“noise traders” who trade in and out of their positions with varying mean-reversion speeds.
In equilibrium, a rational arbitrageur – with trading costs and using the methodology that we
derive – needs to take the other side of these noise-trader positions to clear the market. We
solve the equilibrium explicitly and show how noise trading leads to return predictability and
return reversals. Further, we show that noise-trader demand that mean-reverts more quickly
leads to larger return predictability. This is because a fast mean reversion is associated with
high transaction costs for the arbitrageurs and, consequently, they must be compensated in
the form of larger return predictability.
We implement our optimal trading strategy for commodity futures. Naturally, the opti-
mal trading strategy in the absence of transaction costs has a larger Sharpe ratio gross of fees
than our trading policy. However, net of trading costs our strategy performs significantly
better since it incurs far lower trading costs while still capturing much of the return pre-
dictability and diversification benefits. Further, the optimal dynamic strategy is significantly
better than the best static strategy – taking dynamics into account significantly improves
performance.
In conclusion, we provide a tractable solution to the dynamic trading strategy in a rele-
vant and general setting that we believe to have many interesting applications.

26
A Further Analysis and Proofs
Given the linear dynamics of x, the position xt can be expressed easily as a function of the
initial condition and the exogenous path. These results can be used to provide a simple
proof of Proposition 6.

Proposition 8 In discrete time, the optimal dynamic portfolio xt can be written as a func-
tion of the initial position x0 and the return-predicting factors fs between time 0 and the
current time t:
t
X
xt = M1t x0 + M1t−s M2 fs , (A.1)
s=1

where

M1 = ((1 − ρ)Axx + γΣ + Λ)−1 Λ = I − Λ−1 Axx (A.2)

M2 = ((1 − ρ)Axx + γΣ + Λ)−1 (B + (1 − ρ)Axf (I − Φ)) = Λ−1 Axf . (A.3)

Proposition 9 In continuous time, the optimal dynamic portfolio xt can be written in terms
of the initial position x0 and the path of realized factors fs between 0 and the current time t:
Z t
−Λ−1 Axx t −1 A
xt = e x0 + e−Λ xx (t−s)
Λ−1 Axf fs ds. (A.4)
s=0

Proof of Propositions 1, 2, and 3. We calculate the expected future value function as

1 1 >
Et [V (xt , ft+1 )] = − x> > >
t Axx xt + xt Axf (I − Φ)ft + ft (I − Φ) Af f (I − Φ)ft (A.5)
2 2
1
+ Et (ε>
t+1 Af f εt+1 ) + a0 .
2

The agent maximizes the quadratic objective − 21 x> Jt xt + x>

t jt with

Jt = γΣ + Λ + (1 − ρ)Axx (A.6)

jt = (B + (1 − ρ)Axf (I − Φ))ft + Λxt−1 . (A.7)

27
The maximum value is attained by

xt = Jt−1 jt (A.8)

which proves (11).

The maximum value is equal to V (xt−1 , ft ) = 21 jt> Jt−1 jt and combining this with (8) we
obtain an equation that must hold for all xt−1 and ft . This implies the following restrictions
on the coefficient matrices:

1 1 1
− Axx = Λ(γΣ + Λ + (1 − ρ)Axx )−1 Λ − Λ (A.9)
2 2 2
−1
Axf = Λ(γΣ + Λ + (1 − ρ)Axx ) (B + (1 − ρ)Axf (I − Φ)) (A.10)
1 1
Af f = (B + Axf (I − Φ))> (γΣ + Λ + (1 − ρ)Axx )−1 (B + Axf (I − Φ)) (A.11)
2 2
1−ρ
+ (I − Φ)> Af f (I − Φ).
2

We next derive the coefficient matrices Axx , Axf , Af f by solving these equations. For
1 1 1 1
this, we first rewrite Equation (A.9) by letting Z = Λ− 2 Axx Λ− 2 and M = Λ− 2 ΣΛ− 2 , which
yields

Z = I − (γM + I + (1 − ρ)Z)−1 ,

which is a quadratic with an explicit solution. Since all solutions Z can written as a limit of
polynomials of M , Z and M commute and the quadratic can be sequentially rewritten as

(1 − ρ)Z 2 + Z(I + γM − (1 − ρ)I) = γM

 2
1 γ 1
Z+ (ρI + γM ) = M+ (ρI + γM )2 ,
2(1 − ρ) 1−ρ 4(1 − ρ)2

28
resulting in
  21
γ 1 2 1
Z = M+ (ρI + γM ) − (ρI + γM ) (A.12)
1−ρ 4(1 − ρ)2 2(1 − ρ)
"  12 #
1 γ 1 2 1 1
Axx = Λ 2 M+ (ρI + γM ) − (ρI + γM ) Λ 2 , (A.13)
1−ρ 4(1 − ρ)2 2(1 − ρ)

that is,
  12
γ 1 1 1 1 1 1 1
Axx = Λ 2 ΣΛ 2 + 2
(ρ2 Λ2 + 2ργΛ 2 ΣΛ 2 + γ 2 Λ 2 ΣΛ−1 ΣΛ 2 )
1−ρ 4(1 − ρ)
1
− (ρΛ + γΣ). (A.14)
2(1 − ρ)

Note that the positive definite choice of solution Z is only one that results in a positive
definite matrix Axx .
In the case Λ = λΣ for some scalar λ > 0, the solution for Axx = aΣ, where a solves

λ2
−a= − λ, (A.15)
γ + λ + (1 − ρ)a

or

(1 − ρ)a2 + (γ + λρ)a − λγ = 0, (A.16)

with solution
p
(γ + λρ)2 + 4γλ(1 − ρ) − (γ + λρ)
a = . (A.17)
2(1 − ρ)

The other value-function coefficient determining optimal trading is Axf , which solves the
linear equation (A.10). To write the solution explicitly, we note first that, from (A.9),

Λ(γΣ + Λ + (1 − ρ)Axx )−1 = I − Axx Λ−1 . (A.18)

29
Using the general rule that vec(XY Z) = (Z > ⊗ X) vec(Y ), we re-write (A.10) in vectorized
form:

vec(Axf ) = vec((I − Axx Λ−1 )B) + ((1 − ρ)(I − Φ)> ⊗ (I − Axx Λ−1 )) vec(Axf ), (A.19)

so that

−1
vec(Axf ) = I − (1 − ρ)(I − Φ)> ⊗ (I − Axx Λ−1 ) vec((I − Axx Λ−1 )B). (A.20)

In the case Λ = λΣ, the solution is

Axf = λB((γ + λ + (1 − ρ)a)I − λ(1 − ρ)(I − Φ))−1

= λB((γ + λρ + (1 − ρ)a)I + λ(1 − ρ)Φ))−1 (A.21)

γ −1
= B + (1 − ρ)Φ . (A.22)
a

Finally, Af f is calculated from the linear equation (A.11), which is of the form

Af f = Q + (1 − ρ)(I − Φ)> Af f (I − Φ) (A.23)

with

Q = (B + Axf (I − Φ))> (γΣ + Λ + (1 − ρ)Axx )−1 (B + Axf (I − Φ))

a positive-definite matrix since 1 − ρ ≤ 1 and I − Φ ≤ I.

The solution is easiest to write explicitly for diagonal Φ, in which case

Qij
Af f,ij = . (A.24)
1 − (1 − ρ)(1 − Φii )(1 − Φjj )

In general,

−1
vec (Af f ) = I − (1 − ρ)(I − Φ)> ⊗ (I − Φ)> vec(Q). (A.25)

30
One way to see that Af f is positive definite is to iterate (A.23) starting with A0f f = 0, given
that I ≥ Φ.
Having computed the coefficient matrices, finishing the proof is straightforward. Equa-
tion (16) follows directly from (11). Equation (13) follows from (16) by using the equations
for Axf and a, namely (A.10) and (A.15).

Proof of Propositions 4 and 5. Given the conjectured value function, the optimal choice
τ equals

τt = −Λ−1 Axx xt + Λ−1 Axf ft ,

Once this is inserted in the HJB equation, it results in the following equations defining the
value-function coefficients (using the symmetry of Axx ):

− ρAxx = Axx Λ−1 Axx − γΣ (A.26)

ρAxf = −Axx Λ−1 Axf − Axf Φ + B (A.27)

ρAf f = A> −1
xf Λ Axf − Af f Φ. (A.28)

1
Pre- and post-multiplying (A.26) by Λ− 2 , we obtain

ρ2
− ρZ = Z 2 + I − C, (A.29)
4

that is,
 ρ 2
Z + I = C, (A.30)
2

where

1 1
Z = Λ− 2 Axx Λ− 2 (A.31)
2
1 1 ρ
C = γΛ− 2 ΣΛ− 2 + I, (A.32)
4

31
This leads to

ρ 1
Z = − +C2 ≥ 0, (A.33)
2

implying that
 1
2 2
ρ 1 1 1 ρ 1
Axx = − Λ + Λ2 γΛ− 2 ΣΛ− 2 + Λ2 . (A.34)
2 4

The solution for Axf follows from Equation (A.27), using the general rule that vec(XY Z) =
(Z > ⊗ X) vec(Y ):

−1
vec(Axf ) = ρI + Φ> ⊗ IK + IS ⊗ (Axx Λ−1 ) vec(B)

If Λ = λΣ, then Axx = aΣ with

1
− ρa = a2 −γ (A.35)
λ

with solution
r
ρ ρ2 2
a = − λ+ γλ + λ. (A.36)
2 4

In this case, (A.27) yields

 a −1
Axf = B ρI + I + Φ
λ
γ −1
= B I +Φ ,
a

where the last equality uses (A.35).

Then we have

a  −1
Σ B (aΦ + γI)−1 ft − xt

τt = (A.37)
λ
a
It is clear from (A.36) that λ
decreases in λ and increases in γ.

32
Proof of Proposition 6. We prove this proposition in two main steps. We use the
notation from Proposition 8.
(i) It holds that

M1 (∆t ) = I − Λ−1 Axx + O(∆t ) ∆t

M2 (∆t ) = Λ−1 Axf + O(∆t ) ∆t

as ∆t → 0.
t −1 A
(ii) M1 (∆t ) ∆t → e−Λ xx t
uniformly on [0, T ] for any T > 0. For any continuous path u,
x̂t → xt uniformly on [0, T ] for any T > 0. It then follows immediately from (9) and (34)
∆xt
that ∆t
→ τt .

Proof of Proposition 7. Suppose that αt = Bft with B given by (55) and apply
Proposition 5 to conclude that, if xt = −ftK+1 , then dxt = −dftK+1 . The comparative-static
results are immediate.
To calculate the price explicitly, we need to specify the “normal” return relative to which
the alpha is calculated. In the convenient case µt = µpt − δ for some constant µ, we have

dpt = (µpt − δ + αt )dt + dut

for some martingale u. It follows that

d(e−µt pt ) = −e−µt (δ − αt )dt + e−µt dut ,

so that
T
1 − e−µ(T −t)
Z
−µt −µT
e pt = Et [e pT ] + δ − e−µ(s−t) Et [αs ] ds. (A.38)
µ t

33
Letting T → ∞ and using a transversality condition, we get

δ
pt = − B(µ + Φ)−1 ft . (A.39)
µ

The coefficient f l , l ≤ L, in the equation above can be calculated to be

 
2 −1 µ γ
λσ (µ + ψl ) ψl + (ρ + κ) + > 0.
µ + κ λ(µ + κ)

34
References
Acharya, V., and L. H. Pedersen, 2005, “Asset Pricing with Liquidity Risk,” Journal of
Financial Economics, 77, 375–410.

Almgren, R., and N. Chriss, 2000, “Optimal Execution of Portfolio Transactions,” Journal
of Risk, 3, 5–39.

Amihud, Y., and H. Mendelson, 1986, “Asset Pricing and the Bid-Ask Spread,” Journal of
Financial Economics, 17, 223–249.

Asness, C., T. Moskowitz, and L. H. Pedersen, 2008, “Value and Momentum Everywhere,”
Working paper AQR, NYU, and Chicago.

Bertsimas, D., and A. W. Lo, 1998, “Optimal Control of Execution Costs,” Journal of
Financial Markets, 1, 1–50.

Brunnermeier, M. K., and L. H. Pedersen, 2005, “Predatory Trading,” Journal of Finance,

60(4), 1825–1863.

Campbell, J. Y., and L. M. Viceira, 2002, Strategic Asset Allocation Portfolio Choice for
Long-Term Investors. Oxford University Press.

Carlin, B. I., M. Lobo, and S. Viswanathan, 2008, “Episodic Liquidity Crises: Cooperative
and Predatory Trading,” Journal of Finance, 62, 2235–2274.

Constantinides, G. M., 1986, “Capital Market Equilibrium with Transaction Costs,” Journal
of Political Economy, 94, 842–862.

Davis, M., and A. Norman, 1990, “Portfolio selection with transaction costs,” Mathematics
of Operations Research, 15, 676–713.

Engle, R., and R. Ferstenberg, 2007, “Execution Risk,” Journal of Portfolio Management,
33, 34–45.

35
Gârleanu, N., 2009, “Portfolio Choice and Pricing in Imperfect Markets,” Journal of Eco-
nomic Theory, 144(2), 532–564.

Gârleanu, N., L. H. Pedersen, and A. Poteshman, 2008, “Demand-Based Option Pricing,”

Review of Financial Studies (forthcoming).

Grinold, R., 2006, “A Dynamic Model of Portfolio Management,” Journal of Investment

Management, 4, 5–22.

Heaton, J., and D. Lucas, 1996, “Evaluating the effects of incomplete markets on risk sharing
and asset pricing,” Journal of Political Economy, 104, 443–487.

Jang, B., H. Koo, H. Liu, and M. Loewenstein, 2007, “Liquidity Premia and Transaction
Costs,” Journal of Finance, 62, 2329–2366.

Lagos, R., and G. Rocheteau, 2006, “Search in Asset Markets,” Working paper, NYU.

Liu, H., 2004, “Optimal consumption and investment with transaction costs and multiple
assets,” Journal of Finance, 59, 289–338.

Ljungqvist, L., and T. Sargent, 2004, Recursive Macroeconomic Theory, 2nd edition. MIT
press.

Lo, A., H. Mamaysky, and J. Wang, 2004, “Asset prices and trading volume under fixed
transaction costs,” Journal of Political Economy, 112, 1054–1090.

Longstaff, F. A., 2001, “Optimal Portfolio Choice and the Valuation of Illiquid Securities,”
The Review of Financial Studies, 14, 407–431.

Lynch, A., and S. Tan, 2008, “Explaining the Magnitude of Liquidity Premia: The Roles of
Return Predictability, Wealth Shocks and State-dependent Transaction Costs.,” Working
paper, NYU.

Perold, A., 1988, “The Implementation Shortfall: Paper Versus Reality,” Journal of Portfolio
Management, 14, 4–9.

36
Vayanos, D., 1998, “Transaction Costs and Asset Prices: A Dynamic Equilibrium Model,”
Review of Financial Studies, 11, 1–58.

Vayanos, D., and J.-L. Vila, 1999, “Equilibrium Interest Rate and Liquidity Premium with
Transaction Costs,” Economic Theory, 13, 509–539.

37
Table 1: Performance of Trading Strategies Before and After Transaction Costs.
This table shows the annualized Sharpe ratio gross and net of trading costs for the optimal
trading strategy in the absence of trading costs (“no TC”), our optimal dynamic strategy
(“optimal”), and a strategy that optimizes a static one-period problem with trading costs
(“static”). Panel A illustrates this for a low transaction cost parameter, while Panel B has
a high one.

Panel A: Low Transaction Costs

no TC optimal static
gross SR 0.79 0.63 0.64
net SR −18 0.54 0.44

Panel B: High Transaction Costs

no TC optimal static
gross SR 0.79 0.57 0.58
net SR −22 0.45 0.33

38
 contracts of copper

Et(targett+1)

targett

statict

new position xt

old position xt-1

contracts of crude

Figure 1: Optimal Trading Strategy: Triangulating between Current, Static, and

Future Portfolios. This figure shows how the optimal trade moves from the existing posi-
tion xt−1 towards the target, trading only part of the way to the target to limit transactions
costs. The target is an average of the static Markowitz portfolio and the expected future
target, which depends on the optimal portfolio in the future including the return predictors’
expected alpha decay.

39
 contracts of copper

fast signal

targett

old position xt-1 new position xt

slow signal

contracts of crude

Figure 2: Optimal trading strategy: Down-Weight Fast-Decay Factors. This figure

shows how the optimal trade moves from the existing position xt−1 towards the target, which
put more weight on slow persistent factors relative to factors with fast alpha decay.

40
 Cumulative Gross Returns
1.2
No TC
Optimal
Static
1

0.8

0.6

0.4

0.2

−0.2
04/20/97 09/02/98 01/15/00 05/29/01 10/11/02 02/23/04 07/07/05 11/19/06 04/02/08

Figure 3: Cumulative Excess Returns Gross of Transactions Costs. This figure shows
the cumulative excess returns before transactions costs for the the optimal trading strategy
in the absence of trading costs (“no TC”), our optimal dynamic strategy (“optimal”), and
a strategy that optimizes a static one-period problem with trading costs (“static”).

41
 Cumulative Net Returns
1.2
No TC
Optimal
Static
1

0.8

0.6

0.4

0.2

−0.2
04/20/97 09/02/98 01/15/00 05/29/01 10/11/02 02/23/04 07/07/05 11/19/06 04/02/08

Figure 4: Cumulative Excess Returns Net of Transactions Costs. This figure shows
the cumulative excess returns after transactions costs for the the optimal trading strategy
in the absence of trading costs (“no TC”), our optimal dynamic strategy (“optimal”), and
a strategy that optimizes a static one-period problem with trading costs (“static”).

CRUDE GOLD
10 15
No TC No TC
Optimal Optimal
8 Optimal (high TC) Optimal (high TC)

10
6

4 5

0
0

−2 −5

−4

−10
−6

−8 −15
04/20/97 09/02/98 01/15/00 05/29/01 10/11/02 02/23/04 07/07/05 11/19/06 04/02/08 04/20/97 09/02/98 01/15/00 05/29/01 10/11/02 02/23/04 07/07/05 11/19/06 04/02/08

Figure 5: Positions in Crude and Gold Futures. This figure shows the positions in
crude and gold for the the optimal trading strategy in the absence of trading costs (“no
TC”) and our optimal dynamic strategy (“optimal”) using high and low transactions costs.

42
 11 Optimal Trading After Shock to Signal 1 (5−Day Returns) 11
x 10 x 10 Optimal Trading After Shock to Signal 2 (1−Year Returns)
3.5 12
No TC No TC
Optimal Optimal
Optimal (high TC) Optimal (high TC)
3
10

2.5
8

1.5

4
1

2
0.5

0 0
0 10 20 30 40 50 60 70 0 50 100 150 200 250 300
11 Optimal Trading After Shock to Signal 3 (5−Year Returns)
x 10
0
No TC
Optimal
−0.5 Optimal (high TC)

−1

−1.5

−2

−2.5

−3

−3.5

−4

−4.5

−5
0 100 200 300 400 500 600 700 800

Figure 6: Optimal Trading in Response to Shock to Return Predicting Signals.

This figure shows the response in the optimal position following a shock to a return predictor
as a function of the number of days since the shock. The top left panel does this for a shock
to the fast 5-day return predictor, the top right panel considers a shock to the 12-month
return predictor, and the bottom panel to the 5-year predictor. In each case, we consider
the response of the optimal trading strategy in the absence of trading costs (“no TC”) and
our optimal dynamic strategy (“optimal”) using high and low transactions costs.

43