Journal of Financial Risk Management, 2014, 3, 19-28

Published Online June 2014 in SciRes. http://www.scirp.org/journal/jfrm

http://dx.doi.org/10.4236/jfrm.2014.32003

Continuous-Time Mean-Variance Portfolio

Selection with Inflation in an Incomplete
Market
Yingying Xu1, Zhuwu Wu1,2
1

School of Science, China University of Mining and Technology, Xuzhou, China

School of Management, China University of Mining and Technology, Xuzhou, China
Email: [email protected], [email protected]

Received 14 April 2014; revised 10 May 2014; accepted 3 June 2014

Copyright 2014 by authors and Scientific Research Publishing Inc.
This work is licensed under the Creative Commons Attribution International License (CC BY).
http://creativecommons.org/licenses/by/4.0/

Abstract
This paper concerns a continuous-time portfolio selection problem with inflation in an incomplete
market. By using the approach of more general stochastic linear quadratic control technique (SLQ),
we obtain the optimal strategy and efficient frontier to this problem. Furthermore, a numerical
example is also provided.

Keywords
Portfolio Selection, Efficient Frontier, Optimal Strategy, Stochastic Linear-Quadratic Control

1. Introduction
Portfolio selection problem is a key topic in the modern finance. The seminal work of Markowitz (1952, 1959)
addressed the issue of allocation of wealth in order to obtain the optimal return-risk trade-off. Since then, the
mean-variance model has been extended in many aspects. Merton (1969, 1971) introduced a continuous-time
model for maximizing the expected utility from investors consumption and terminal wealth. Zhou & Li (2000)
investigated a continuous-time mean-variance portfolio problem and obtained the optimal strategy and efficient
frontier by using the stochastic LQ technique, which opened up possible approach to solve the problem for more
constraints. Following Zhou & Li (2000), many scholars extended this model to the more complicated market
situations, such as liability, bankruptcy prohibition and incomplete market. See more details in Bielecki, Jin,
Pliska, & Zhou (2005), Xie, Li, & Wang (2008) and Ji (2010).
In a real world, investors must deal realistically with the problems of inflation with the growth of economy
when adopting a long-term but finite horizon investment strategy. Therefore, the consideration of inflation risk
How to cite this paper: Xu, Y. Y., & Wu, Z. W. (2014). Continuous-Time Mean-Variance Portfolio Selection with Inflation in
an Incomplete Market. Journal of Financial Risk Management, 3, 19-28. http://dx.doi.org/10.4236/jfrm.2014.32003

Y. Y. Xu, Z. W. Wu

in a portfolio selection model will make it more practical. However, to our knowledge, the research on mean-variance portfolio selection under inflation is limited. The existing literature on this topic is not much as can be
seen Brennan & Xia (2002) and Bensoussan, Keppo, & Sethi (2009).
The main goal of this paper is to investigate a continuous-time portfolio selection problemunder inflation in
an incomplete market. It is clear that this model is more suitable and practical in most of the real-world situations, especially for long-term investors. Therefore, our focus will be on two cases. On the one hand, we investigate the incomplete market with inflation, in which there are m risky assets and one risk-free asset. The price
processes of risky assets are driven by an m-dimensional Brownian motion. We also assume that the inflation
factors affected by the market are random, which can be described by m + 1 Brownian motion. In general, the
changes in the nominal price index are not just correlated with the risky assets nominal prices, but also with
other uncertainties. It is reasonable that the other uncertainties can be represented by one Brownian motion,
which is our (m + 1)-th Brownian motion. The original idea can be seen in Brennan & Xia (2002). On the other
hand, we employ a stochastic linear quadratic (LQ) technique introduced by Zhou & Li (2000) to solve this
problem. It should be pointed out that the introduction of inflation is by no means routine and does give rise to
difficulties which are not encountered in Zhou & Li (2000). However, by using the more general stochastic LQ
control technique in Yong & Zhou (1999), we can also obtain the optimal strategy and efficient frontier in
closed forms.
The paper proceeds as follows. In Section 2, the model is formulated. Section 3 provides a closed-form solution of our model by using the more general stochastic LQ approach. Section 4 presents a numerical example.
Finally, concluding remarks and suggestions for future work are given in Section 5.

2. Problem Formulation
We consider a market in which m + 1 assets are traded continuously within the time horizon [ 0,T ] . One of the
assets is the risk-free whose nominal price process St0 is subject to the following ordinary differential equation:
dSt0 rt St0 dt , t [ 0, T ] ,
=
(1)
0
S0= s0 > 0,
where rt > 0 is the nominal interest rate of the risk-free asset. The remaining m assets are risky and their nominal price processes St1 , , Stm satisfy the following stochastic differential equations:

dSti =
Sti bti dt + ti dWt , t [ 0, T ] ,
i
S0= si R

(2)

where Wt := Wt1 , , Wt m is a m-dimensional standard Brownian motion, which represents the random factors
that affect risky assets nominal prices. bti is the appreciation rate of the ith ( i = 1, , m ) risky asset, let
bt := ( bt1 , bt2 , , btm ) . ti := ( ti1 , ti 2 , , tim ) is the volatility associated with the ith risky asset. Thus, the covariance matrix of risky assets is as follows:

=
(t ) :

, , , ) ( )
(=
1
t

2
t

m
t

ij
t mm

(3)

where the superscript represents the transpose of a vector or a matrix. As widely adopted in the literature,
we assume the non-degeneracy condition of

( t ) ( t ) I , t [ 0, T ] , > 0.

(4)

The nominal price of real consumption goods in the economy at time t is denoted by t , which follows a
diffusion process:
d t
= t dt + t dWt =: t dt + t1dWt1 + + tm +1dWt m +1 ,

(5)
t
=
1,
0
1
m +1
where Wt := Wt , , Wt
is a (m + 1)-dimensional Brownian motion, which represents the random factors

20

Y. Y. Xu, Z. W. Wu

that affect the price index. t is the expected rate of inflation, and t := (t1 , , tm +1 ) is the volatility of the
price index.
Remark 1. In general, the driving factors of inflation include but do not equal to the ones of the risky assets
nominal prices. We describe randomness of the price index with Wt1 , , Wt m +1 , in which the foregoing m Brownian motions are the same ones that drive the risky assets nominal price, and the (m + 1)-th Brownian motion
Wt m +1 represents other randomness. Moreover, we assume that Wt m +1 and Wt j ( j = 1, 2, , m ) are independent.
F { Ft ; t 0} and
Let ( , F , P, { Ft }t 0 ) be a complete filtered probability space, where=

=
Ft Ws , Wsm +1 ;0 s t . We assume that all the coefficient functions are continuous bounded deterministic
functions on [ 0,T ] . We denote by C [ 0, T ] ; R nk the class of R nk -valued continuous bounded deterministic functions on [ 0,T ] , and by L2F [ 0, T ] ; R m the class of all R m -valued, progressively measurable and

square integral random variables on [ 0,T ] under P with norm

=
L2 :
F

E t dt
0
2

1
2

< , t L2F

([0, T ]; R ) .
m

We denote by X t the nominal wealth of the investor at time t [ 0, T ] . Suppose the investor decides to hold
N ti shares of ith asset ( i = 0,1, , m ) at time t. Then
Xt
=

Nti Sti ,

0 t T.

(6)

i =0

Let ti = N ti Sti be the total nominal market value of the ith ( i = 0,1, , m ) asset held by the investor at time

t, and let t := t1 , , tm . We call the process=

: { t : t [ 0, T ]} a portfolio or a strategy of the investor.
We assume that the trading of shares takes place continuously in a self-financing fashion and there are no
transaction costs or taxes. We also assume that short-selling is allowable. Then we have

dX t = rt X t dt + Bt t dt + t t dWt

X 0= x > 0

(7)

where Bt =
( bt1 rt ,, btm rt ) is the risk premium.
With the consideration of the inflation, the real value of any asset in the economy at time t is determined by
deflating by the price index t . The real value of the investors wealth is given by X t t . Let =
ht : X t t .
Applying Its formula to ht , we obtain

Bt t t
t ht dWt tm +1ht dWt m +1
dht = rt t + tt ht + t
dt + t
t

t t t
t

B
t ht dWt
dt + t
= rt t + tt ht + t
t

h0 = x

1
where t = t , , tm , 0 , Bt =
( bt1 rt ,, btm rt , 0 ) and

(8)

=
t

t11 t12
21
t22
t
ij

=

t m +1 m +1
( ) ( )
m1
m2

t
t
0
0

( )

t1m
t2 m

tmm
0

0
.

0
1

Remark 2. In order to facilitate the following mathematical treatment, we give t , Bt , t . In fact, we can also
think that: it is assumed that there are m + 1 risky assets in the market, and their nominal prices are driving by
m + 1 Brownian motions. The first m risky assets are the same ones we assumed before, and the (m + 1)-th risky
asset is a fictitious risky asset. The=
ith ( i 1, 2, , m + 1) risky assets volatility is given by the ith rank of t .
Moreover, we assume the shares of the (m + 1)-th risky asset held by the investor remains 0. Therefore, we can

21

Y. Y. Xu, Z. W. Wu

get t by deriving t .
The admissible strategy set under inflation with initial wealth x is defined as

U=
( x) :

{ L ([0, T ]; R ) , ( h , ) satisfies Equation (8)} .

t

m +1

2
F

The objective of the investor is to maximize the expected terminal real wealth, EhT , and at the same time to
minimize the variance of the terminal real wealth, VarhT ,
Var hT =
E ( hT EhT ) =
EhT2 ( EhT ) .
2

This is the mean-variance model which can be expressed by the bi-objective optimization problem:

min ( EhT ,VarhT ) .

(9)

U ( x )

It is known from Li & Ng (2000) that Equation (9) is equivalent to the following single objective optimization
problem:

P( )

min ( EhT + VarhT ) ,

U ( x )

(10)

where the parameter > 0 represents the weight imposed by the investor on the objective VarhT . Define

P( ) :=
{ is an optimal strategy of P ( )} .

(11)

3. Solution to the Problem

In this section, we will apply the more general stochastic linear quadratic (LQ) control technique in Yong &
Zhou (1999) to our model. Firstly, we will introduce a stochastic LQ auxiliary control problem and derive its
optimal feedback control. Eventually the optimal portfolio strategy and the efficient frontier for the original
mean-variance portfolio optimization problem under inflation are obtained in closed form.

3.1. Auxiliary Problem

Similar to Zhou & Li (2000), we introduce an auxiliary problem as follows:
A( , )

min E hT2 hT

U ( x )

(12)

where > 0, < < + . Define

A( , ) =
{ is an optimal control of A ( , )}

(13)

Recall Theorem 3.1 in Zhou & Li (2000) which shows the relationship between problems P ( ) and
A( , ) .
Theorem 1. For any > 0 , one has

P( )

< <+

A( , ) .

Moreover, if P( ) , then A( , ) with = 1 + 2 EhT , where ht is the wealth process corres-

ponding to the strategy .

Let = ( 2 ) , Y=
ht . Then Equation (8) becomes the following stochastic differential equation:
t

Vt
dt + t t t Yt t dWt
dYt = rt t + tt Yt dt + rt t + tt dt + t
t

=
=

:
Y
y
x

(14)

where V=
Bt t t , and the objective function of the auxiliary problem A ( , ) becomes E ( YT2 2 ) .
t
Hence, the auxiliary problem A ( , ) is equivalent to minimizing
1

J ( ; ) := E YT2
2

22

(15)

Y. Y. Xu, Z. W. Wu

Furthermore, the admissible strategy set U ( x ) can be written as

( y ) :=

{ L ([0, T ]; R ) , (Y , ) satisfies Equation (14)} .

t

m +1

2
F

Thus the auxiliary problem A ( , ) is equivalent to the following stochastic LQ control problem:

A ( )

min J ( ; ) .

( y )

3.2. Solution to the Auxiliary Problem

A solution of the stochastic LQ control problem A ( ) will involve, in an essential way, the following Riccati
equation:

1

0
Pt + 2 At + t t ( M t + Dtt ) ( DtDt ) ( M t + Dtt ) Pt =

PT =
D P D > 0, t 0, T ,
[ ]
t t t

(16)

along with the following adjoint ordinary differential equation:

1
1

0
g t + At M t ( DtDt ) ( M t + Dtt ) gt + t Dt ( DtDt ) ( M t + Dtt) t + At Pt =

gT = 0,

(17)

where
dPt
dg t
,
=
Pt : =
, g t :
dt
dt

At =rt t + t t , M t =

Dt j
=

1
t1 j , , tm +1 j , Dt
=
t

Vt
,
t

( D ,, D ) .
m +1
t

1
t

Theorem 2. Let Pt C ([ 0, T ] ; R + ) and gt C ([ 0, T ] ; R ) be the solution of Equations (16) and (17), respectively, such that

=
t :
=
t :

( DtDt ) ( M t + Dtt) ,
1

( DtPD
t t ) ( M t g t + Dt Pt t ) .
1

Then Problem A ( ) is solvable with the optimal control being in a state feedback form,

* (Yt ) =
tYt t , t [ 0, T ] .

(18)

Moreover, the optimal cost value is

=
J*

1
1 T

2
2 At gt + Pt t t ( DtPt Dt ) 2 t

0
2

2
dt + P0 y + g 0 y ,
2

(19)

where M = i , j mij2 for any matrix or vector M = ( mij ) and y= x .

Proof: We first prove that the control given by Equation (18) is an admissible control. Substituting Equation
(18) into Equation (14), we have

dYt = ( At M tt ) Yt + At M t t dt + (t tDt ) Yt + t tDt dWt

Y0 = y.

23

(20)

Y. Y. Xu, Z. W. Wu

Noting that Pt , gt are continuous, and the appreciation coefficient and the diffusion coefficients are bounded
continuous within t. Hence, we deduce that Equation (20) admits a unique strong solution Yt which yields

E sup Yt* KT 1 + y
t[ 0,T ]

),

where KT > 0 is a constant associated with the terminal time. Therefore, we have shown that (Yt ) ( y ) .
Next, we prove that * (Yt ) is an optimal feedback control of state variable Yt . For any t ( y ) , let Yt
2
be the state variable associated with the control vector t . By applying It formula to PY
t t 2 and g t Yt , and
integrating them from 0 to T, taking expectations, add them together, we get
1
1
1
YT2 P0Y02 g 0Y0 = J ( ; ) P0Y02 g 0Y0
2
2
2
2
1

1 T

=
E 0 ( + t Yt + t ) ( DtPt Dt )( t + t Yt + t ) ( DtPt Dt ) 2 t + 2 At gt + 2 Pt t t dt.
2

(21)

Because
1 T
E ( t + tYt + t ) ( DtPt Dt )( t + t Yt + t ) dt 0 .
2 0

We obtain
J ( ; )

1
1 T
P0Y02 g 0Y0 0 2 At gt + 2 Pt t t ( DtPt Dt ) 2 t
2
2

dt ,

and the equality holds if and only if (Yt ) tYt t , t [ 0, T ] . It shows that the feedback control given by
Equation (18) is an optimal control and the optimal cost function can be obtained by Equation (19). The proof is
completed.
Noting that the third constraint in Equation (16) is satisfied automatically since the assumption t t I m ,
t [ 0, T ] . Obviously, the solution of Equation (16) can be expressed by the following:
2
1

Pt= exp t 2 As + s s ( Ds Ds ) 2 s ds .

(22)

Let H t = gt Pt . Then noting Equation (16) and Equation (17), one has
1
T
As + s s s Ds ( Ds Ds ) ( M s + Ds s ) ds

Ht =
1 e t

(23)

Since the equivalence of problem A ( ) and A ( , ) , the optimal feedback control of the auxiliary problem A ( , ) is also given by Theorem 2:

* ( ht ) =
t ( ht ) t =
( DtDt )

( M + D ) ( h ) ( DD ) ( M H + D ) ,
1

t t

t t

t [ 0, T ] , (24)

Substituting Equation (23) into Equation (24), we have

* ( ht ) =
( DtDt )

( M + D ) h + ( DD )
t

t t

1
T
t As + s s s Ds ( Ds Ds ) ( M s + Ds s ) ds

+ Dtt Dtt .
M te

3.3. Solution to the Original Problem

Let ht* be the wealth process under the optimal feedback control * of the auxiliary problem A ( , ) . Substituting * ( ht ) =
t ( ht ) t into Equation (8) yields
*

*
*
*
dht= ( At M tt ) ht + ( M tt M t t ) dt + t ht t Dt + ht t dWt

h 0 = x
(
)
0

( (

24

(25)

Y. Y. Xu, Z. W. Wu

Applying Its formula to ht2 yields

dh2 = 2 ( A M ) + DD 2 D + h2 dt + 2 ( M M ) 2 DD
t
t t
t t t t
t t t
t t
t t
t t
t t t t

+ 2 tDtDtt + 2 ( t Dtt t Dt t ) ht*dt + 2tDtDtt 2 tDtDtt + tDtDt t dt

+ 2ht* t ht* t Dt + ht*t dWt

2
2
h0 = x

( (

(26)

Taking expectation on both side of Equations (25) and (26), which leads respectively to

dEht* =
( At M tt ) Eht* + ( M tt M t t ) dt

Eh0 = x

(27)

dEh2 =
2 A M ( DD )1 M 2 M ( DD )1 D D ( DD )1 D + Eh2 dt
t
t
t t
t
t
t t
t t
t t
t t
t t
t t

t
t

g 2

1
1
1
t M t ( DtDt ) M t + 2 M t ( DtDt ) M t 2 M t t ( DtDt ) M t dt
+

Pt
Pt

2
2
Eh0 = x

(28)

and

The solution of Equation (27) is

Eh
=
t

t
t
x0 0t ( As M ss )ds
( A M ) dz
e
+ 0 ( M s s M s s ) e s z z z ds, t [ 0, T ] .
q0

This leads to

EhT= + ,

(29)

where

:= xe 0 (
T

At M tt )dt

, := 1 e

1
T
0 M t ( Dt Dt ) M tdt

Similarly, by solving Equation (28) we have

1
1
T
2 At M t ( DtDt ) M t 2 M t ( DtDt ) Dtt dt

EhT2 = x 2 e 0

T g
g
1
1
1
+ 0 t M t ( DtDt ) M t + 2 M t ( DtDt ) M t 2 M t t ( DtDt ) M t
Pt
Pt

1
1
T
2 As M s ( Ds Ds ) M s 2 M s ( Ds Ds ) Ds s ds

e t

dt ,

(30)

t [ 0, T ] .

Substituting Equation (23) into Equation (30), we have

1
1
T
2 At M t ( DtDt ) M t 2 M t ( DtDt ) Dtt dt

=
EhT2 x 2 e 0

1
1
T
2 At M t ( Dt Dt ) M t 2 M t ( Dt Dt ) Dtt dt

= x 2 e 0

+ 2 0 M t ( DtDt ) M te
T

+ 2 1 e

1
T
0 M t ( Dt Dt ) M tdt

t M s ( Ds Ds ) M s ds
T

dt

, t [ 0, T ] .

Then we get

EhT2= + 2 ,
1
1
T
2 At M t ( DD ) M t 2 M t ( DD ) Dtt dt

where := x 2 e 0

25

(31)

Y. Y. Xu, Z. W. Wu

Based on the Theorem 1, if any optimal solution of problem P ( ) exists, it can be obtained by the solution
* of the auxiliary problem A ( , * ) with * = 1 + 2 EhT* . According to EhT*= + , EhT*= + *
with * = * 2 . The above two equations yield

* =

1 + 2
.
1

Thus, the optimal feedback control of the problem P ( ) can be expressed by

t ( ht ) t ,
* ( ht ) =

with

.
+
2 (1 ) 1

*
=
=

Correspondingly, the variance of the terminal wealth is

VarhT= EhT2 EhT = (1 ) 2 2 + 2 .

By substituting =

( Eh

(32)

and , , into Equation (32), we have

=
VarhT*

1 *

2
EhT
+
1
1

e

1
T
0 M t ( Dt Dt ) M tdt

1 e

0 M t ( Dt Dt )
T

M tdt

T
0 ( At M t t ) dt

EhT* xeT
1
M ( D D ) M dt

e 0 t t t t

=
t ( DtDt ) =
M t + Dtt , t ( DtPD
Substituting
t t)
(33), we finally obtain the efficient frontier as follows:
1

Varh =
*
T

1
T
0 M t ( Dt Dt ) M tdt

1 e

0 M t ( Dt Dt ) M tdt
T

( A M ) dt

xe 0 t t t
1
1

T
0 2 At M t ( D D ) M t 2 M t ( D D ) DtCt dt
2
+x e
T
1
M ( D D ) M dt

e 0 t t t t

( M g + DP )
t t

and Equation (23) into Equation

t t t

V (

)1V dt

1
T
*
0 ( At +Vt ( t t ) tt )dt
EhT xe
. (34)

1
T
*
e 0 t t t t
0 ( At M t ( DD ) D )dt
1
EhT xe
=
T

1 e 0 Vt ( t t ) Vt dt

(33)

Remark 3. If we let =
t 0,=
0 , and the market is complete, then Equation (34) would reduce to
0 Bt ( t t )
T

=
Varh
*
T

Btdt

1
T
0 Bt ( t t ) Btdt

EhT* xe 0

rt dt

1 e
Obviously, the result of Zhou & Li (2000) is a special case in our paper.

).
2

(35)

4. Numerical Example
In this section, we discuss a numerical example. Suppose that the market has four assets and a risk-free asset.
Lets assign the following parameters which are needed in our model: the risk-free asset rt = 0.0350 , the expected value of inflation t = 0.0310 , the time horizon T = 2, the volatility of the price index
= ( 0.2123;0.1396;0.0823;0.1562;0.1876 ) , the appreciation rate of assets b = ( 0.1621;0.1422;0.1013;0.1326 )
i, j
is as follows:
the initial wealth x0 = 1 . And also suppose that the covariance matrix t

( )

0.3121

0.1127
0.1362

0.1831

0.1127 0.1362 0.1831

0.2656 0.0956 0.1772

.
0.0956 0.1856 0.1432

0.1772 0.1432 0.2372

44

After some transformations and calculations by using the above parameters, the efficient frontier of our model
is obtained by the following.

=
Varh2* 6.0236 Eh2* 1.4426 .

26

Y. Y. Xu, Z. W. Wu

Figure 1. The efficient frontier with and without inflation.

Next, we compare our model with that of Zhou & Li (2000). The biggest difference is that our model is considered the factor of inflation in the decision-making process. Figure 1 shows the efficient frontier to continuous-time mean-variance model with and without inflation in a market. It can be seen that the frontier with inflation lies below the one without inflation. This means that the inflation plays as a penalty factor for portfolio
revision. Furthermore, it tells us that the impact of it cannot be ignored in the real world when portfolio managers choose investment strategy.

5. Conclusion
This paper extends the work of Zhou & Li (2000) to an incomplete market with inflation. In our model, the inflation process is assumed to be a geometric Brownian motion, which is correlated with those risky assets. The
driving factors of inflation are not the same ones which affect risky assets prices. This means that the random
factors affecting inflation include but do not equal to the ones of risky assets prices. By using the more general
stochastic LQ approach, we have provided a closed-form optimal strategy and efficient frontier. Comparing to
Zhou & Li (2000), our results in this paper are more general. In addition, a numerical example is also provided.
The search on the liability and bankruptcy prohibition in this problem is left for future work.

Acknowledgements
We acknowledge the contributions of Fundamental Research Funds for the Central Universities (2012QNB19)
and Natural Science Foundation of China (11101422, 11371362 and 71173216).

References
Bensoussan, A., Keppo, J., & Sethi, S. P. (2009). Optimal Consumption and Portfolio Decisions with Partially Observed
Real Prices. Mathematical Finance, 19, 215-236. http://dx.doi.org/10.1111/j.1467-9965.2009.00362.x
Bielecki, T. R., Jin, H., Pliska, S. R., & Zhou, X. Y. (2005). Continuous-Time Mean-Variance Portfolio Selection with Bankruptcy Prohibition. Mathematical Finance, 15, 213-244. http://dx.doi.org/10.1111/j.0960-1627.2005.00218.x
Brennan, M. J., & Xia, Y. (2002). Dynamic Asset Allocation under Inflation. The Journal of Finance, 57, 1201-1238.
http://dx.doi.org/10.1111/1540-6261.00459
Ji, S. (2010). Dual Method for Continuous-Time Markowitzs Problems with Nonlinear Wealth Equations. Journal of Mathematical Analysis and Applications, 366, 90-100. http://dx.doi.org/10.1016/j.jmaa.2010.01.044
Li, D., & Ng, W. L. (2000). Optimal Dynamic Portfolio Selection: Multiperiod Mean-Variance Formulation. Mathematical
Finance, 10, 387-406. http://dx.doi.org/10.1111/1467-9965.00100
Markowitz, H. (1952). Portfolio Selection. The Journal of Finance, 7, 77-91.
Markowitz, H. (1959). Portfolio Selection: Efficient Diversification of Investments. New York: Wiley.
Merton, R. C. (1969). Lifetime Portfolio Selection under Uncertainty: The Continuous-Time Case. Review of Economics and
Statistics, 51, 247-257. http://dx.doi.org/10.2307/1926560
Merton, R. C. (1971). Optimum Consumption and Portfolio Rules in a Continuous-Time Model. Journal of Economic

27

Y. Y. Xu, Z. W. Wu

Theory, 3, 373-413. http://dx.doi.org/10.1016/0022-0531(71)90038-X

Xie, S., Li, Z., & Wang, S. (2008). Continuous-Time Portfolio Selection with Liability: Mean-Variance Model and Stochastic LQ Approach. Insurance: Mathematics and Economics, 42, 943-953.
http://dx.doi.org/10.1016/j.insmatheco.2007.10.014
Yong, J., & Zhou, X. Y. (1999). Stochastic Controls: Hamiltonian Systems and HJB Equations (Vol. 43). New York: Springer. http://dx.doi.org/10.1007/978-1-4612-1466-3
Zhou, X. Y., & Li, D. (2000). Continuous-Time Mean-Variance Portfolio Selection: A Stochastic LQ Framework. Applied
Mathematics and Optimization, 42, 19-33. http://dx.doi.org/10.1007/s002450010003

28