Hindawi

Complexity
Volume 2020, Article ID 1435356, 13 pages
https://doi.org/10.1155/2020/1435356

Research Article
Portfolio Optimization with Asset-Liability Ratio
Regulation Constraints

1
De-Lei Sheng and Peilong Shen2
1
School of Applied Mathematics, Shanxi University of Finance and Economics, Taiyuan 030006, China
2
School of Finance, Shanxi University of Finance and Economics, Taiyuan 030006, China

Correspondence should be addressed to De-Lei Sheng; [email protected]

Received 12 June 2019; Accepted 8 February 2020; Published 23 March 2020

Guest Editor: Thiago Christiano Silva

Copyright © 2020 De-Lei Sheng and Peilong Shen. This is an open access article distributed under the Creative Commons
Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is
properly cited.
This paper considers both a top regulation bound and a bottom regulation bound imposed on the asset-liability ratio at the
regulatory time T to reduce risks of abnormal high-speed growth of asset price within a short period of time (or high investment
leverage), and to mitigate risks of low assets’ return (or a sharp fall). Applying the stochastic optimal control technique, a
Hamilton–Jacobi–Bellman (HJB) equation is derived. Then, the effective investment strategy and the minimum variance are
obtained explicitly by using the Lagrange duality method. Moreover, some numerical examples are provided to verify the ef-
fectiveness of our results.

1. Introduction variance problem. Fu et al. [6] consider the continuous-time

mean-variance portfolio selection under different borrowing
Taking liabilities into the traditional portfolio models, interest rates and lending interest rates with the Lagrange
Sharpe and Tint [1] put forward the asset-liability problem duality method. Pan and Xiao [7] also investigate the
for pension fund management under the mean-variance continuous-time asset-liability management problem by
framework. Kell and Müller [2] point out that liabilities considering the stochastic interest rates and inflation risks.
affect the efficient frontier of this asset-liability problem. But Considering a financial market consists of a risk-free bond, a
early research on asset-liability problems was limited to the stock, and a derivative, Li et al. [8] give the optimal in-
standard single-period mean-variance criterion; Leippold vestment strategies of a continuous-time mean-variance
et al. [3] obtain an analytical optimal strategy and efficient asset-liability management in presence of stochastic vola-
frontier for asset-liability problems by using embedding tility. For other related literature studies, you can refer to
technique proposed by Li and Ng [4] under a multiperiod Wei and Wang [9]; Zhang et al. [10]; Duarte et al. [11]; and
mean-variance framework. Leippold et al. [3] also consider so on.
the multiperiod mean-variance asset-liability problem and As everyone knows, the assets scale pursued by invest-
show that the optimal strategy can be decomposed into ment growing too fast over a relatively short period of time,
orthogonal sets and the efficient frontier spanned by an which indicates that the growth rate increasing faster than
orthogonal basis of dynamic returns. For the continuous- the normal rate of growth, is itself accompanied by high
time asset-liability problem, Chiu and Li [5] derive the potential risks. The most typical example in practice is the
optimal policy and the mean-variance efficient frontier by stock market crash which has happened many times almost
applying the technique of stochastic linear-quadratic con- in all countries with stock markets. The earliest stock market
trol. In comparison with the embedding technique and the crash occurred in 1720, the Mississippi Stock Disaster in
stochastic linear-quadratic control method, the Lagrange France and the South Sea Stock Disaster in Britain. The share
duality method is more convenient to solve the mean- price of the Mississippi company rose from around 500 livres
2 Complexity

in May 1719, to nearly 10,000 livres in February 1720, but it Financial leverage needs to be considered because it is
declined to 500 livres in September 1721. The stock prices of ubiquitous in practice, and the excessive leverage is an issue
South Sea Company soar from £128 in January, £175 in tackled in the Basel III requirements. In the narrow sense,
February, £330 in March, and £550 in May, and then, the financial leverage refers to a measure of operating large-scale
shares leaped to £1000 per share by August 1720 and finally business with less money, which is widely used in various
peaked at this level but soon it plunged and triggered an economic activities. For example, the transaction relying on
avalanche of selling. As for the most devastating stock market futures margin is commonly used in futures markets and the
crashes of the United States, such as the American stock margin rate usually varies between 5% and 15% of the total
market crash in 1929 and the American stock market crash in contract value. Only a down payment of 5% of the full
1987, all ended in a catastrophic decline after a period of contract value is required, and an investor can carry out the
drastic growth. When the China stock market crash broke out financial transaction equivalent to 20 times the amount of
in 2007, the Shanghai stock exchange composite index surged the margin. Another example is the way to invest in real
from around 2700 points to the peak value 6124 points, which estate with installment repayment, and investors only need
took only a few months. But then it went down sharply to the to pay the down payment, usually 10% to 30%, but they can
lowest value 1644 points. Another China stock market crash leverage the investment scales of the total market value.
in 2015, the Shanghai stock exchange composite index rise Quite evidently, investors generally use smaller self-owned
from around 3000 points to 5178 points on June 12, but only funds to operate businesses of great market value with fi-
in the next two months, it plunged to around 2800 points. nancial leverage. If these assets appreciate, investors can
The sustained high-speed growth of stock prices within a immediately obtain considerable investment returns.
short period of time is actually accumulating enough de- However, once the value of these assets shrinks, the losses are
structive energy which may erupt at any time. Sophisticated magnified and the corresponding leverage multiples are
investors will actively sell their risky stocks in time when the catastrophic and unbearable. Therefore, the essence of fi-
stock returns reach a certain high level, instead of blindly nancial leverage lies in magnifying the gains and losses to
pursuing much higher returns, so as to avoid the sharp drop meet the needs of investors who have insufficient funds but
before these assets are successfully sold. This high return level want to do large-scale businesses. There is a large body in the
targeted by these traders to sell assets actively is a practical top literature of financial stability which estimates systemic risk
bound used in investment management practice and deter- and explicitly takes into account excessive leverage. Related
mined by investment managers initiatively, which is actually works can be referred to the study bySilva et al. [16] and
the concrete example for the application of top investment many others.
regulation bound in business. Thus, it confirms the real ex- However, financial markets themselves also have the
istence of the investment regulation top bound in real in- function of enlarging profits and losses. As long as investors
vestment practice. The requirement of investment regulation invest in risky assets in financial markets, they are actually
bottom bound is necessary for the pursuit of much higher using the financial market to enlarge their scale of assets,
investment returns. Only if the assets scale pursued by in- which is essentially consistent with the core meaning of
vestment is larger than liability, then it can ensure that the financial leverage in narrow sense. Therefore, in a broad
investment activities are really valuable and vigorous. sense, a financial market itself can be regarded as a financial
These kinds of investment risks of asset price collapse leverage. In other words, the financial leverage has already
resulting from the abnormally rapid rise within a limited been used by investors as long as they invest in financial
time period have extremely destructive, which may even be markets in pursuit of higher investment returns, so it is also
the real manifestation of systemic financial risks. As for necessary to consider leverage regulation even if you only
systemic financial risks, Guerra et al. [12] and Souza et al. invest in ordinary financial markets.
[13] conduct in-depth and innovative research studies and Different from the previous research on the modeling
propose a novel methodology to measure systemic risk in method of the asset-liability problem, both a bottom reg-
networks composed of financial institutions. They define the ulation bound and a top regulation bound are imposed on
bank’s probability of default and calculate this probability the asset-liability ratio to control the variation range so as to
using Merton’s method inspired by Black and Scholes [14] prevent the risks of asset price collapse resulted from an
and Merton [15]. Interestingly, the probability of default abnormal high-speed growth (such as stock market crash),
defined by Guerra et al. [12] and Souza et al. [13] can also be asset shrink, and the collapse of investment leverage. The
used to research the asset-liability management problem dynamic process of asset-liability ratio is defined as an asset
with unbearable collapse risk resulted from the abnormally process being divided by a liability process after eliminating
rapid rise of asset price within a limited time period. For the influence of inflation. A stochastic optimal control model
instance, you can take the horizon on which the entity (firm) is formulated under the framework of mean variance with
may default to be short enough and then large declines in the variance being minimized, given a determined expec-
assets or large increases in liabilities can be researched tation of asset-liability ratio at regulatory time T. Using the
similar to the probability of default calculated by Merton’s Lagrange multiplier method, the original problem is
method. However, based on different considerations, we transformed into an unconstrained optimization problem,
realize our ideas about the asset-liability management and then a Hamilton–Jacobi–Bellman (HJB) equation is
problem with unbearable investment risk using a widely established by adopting technique of stochastic control. At
used mathematical method in this paper. last, using Lagrange duality between the original problem
Complexity 3

and the unconstrained problem, the minimum variance and where μ􏽥t is the expected inflation rate at time t; σ􏽥 t > 0 is the
the effective investment strategy are obtained. 􏽥
volatility of inflation rate; and W(t) is a standard Brownian
The remainder of this paper is organized as follows. motion on the probability space (Ω, F, P).
Section 2 describes the models of asset-liability ratio and the The price process of the risk-free bond is modeled as
constrained control problem. In Section 3, a Lagrange un- B(t) � e􏽥rt , where the constant 􏽥r ≥ 0 represents the nominal
constrained problem is solved by technique of stochastic interest rate.
optimal control, and the results of the original problem are The inflation-linked index bond {I(t), t ≥ 0} has the same
obtained according to Lagrange duality. Meanwhile, a risk source with the price level process, and thus, it can be
special case is also solved at the end of this section. Some expressed as
interpretations of main results are presented in Section 4, t t
and numerical examples are illustrated in Section 5. At last, 1 2 􏽥 􏼩,
I(t) � exp􏼨􏽚 􏼒 r + μ􏽥s 􏼁 − σ􏽥 s 􏼓ds + 􏽚 σ􏽥s dW(s) (2)
Section 6 gives a conclusion of the research work. 0 2 0

where r is the real interest rate at time t.

2. Formulation of the Model The price process of risky stock is formulated as
t 1 2 t
Throughout this paper, (Ω, F, P, 􏼈Ft 􏼉0 ≤ t ≤ T) denotes a S(t) � S0 exp􏼨􏽚 􏼒μ􏽢s − 􏽢 􏼩,
σ􏽢 s 􏼓ds + 􏽚 σ􏽢s dW(s) (3)
complete probability space satisfying the usual condition. A 0 2 0
finite constant T > 0 represents the preselected investment
regulatory time; Ft is the smallest σ− field generated by all where μ􏽢t is the investment return rate satisfying μ􏽢t > 􏽥r and σ􏽢t
random information available until time t, and all random is the volatility of the risky asset.
variables and stochastic processes involved in this article are
Ft measurable for every t ∈ [0, T]. 2.2. Asset Process and Liability Process. A control vector
(π 0(t), π1(t), π2(t)) represents the investment strategy need
to be found, where π0(t) denotes the investment share of
2.1. Financial Market. Similar to the previous research work, risk-free bond, π1(t) represents the investment share of
this paper considers an inflation-affected financial market in inflation-linked index bond, and π2(t) signifies the invest-
continuous time, which consists of one risk-free bond, one ment share of risky stock, and they always satisfy the
inflation-linked index bond, and one risky stock. The in- equation π0(t) + π1(t) + π2(t) � 1.
flation rate P(t) can be regarded as the Consumer Price The company’s accumulated wealth is allowed to invest
Index, which is described by a price level process as follows: in the financial market; thus, the asset process can be de-
t 1 2 t scribed as an investment process. Assuming that
P(t) � exp􏼨􏽚 􏼒μ􏽥s − 􏽥 􏼩,
σ􏽥 s 􏼓ds + 􏽚 σ􏽥 s dW(s) (1) π :� 􏼈(π 0 (t), π1 (t), π2 (t))􏼉0 ≤ t ≤ T is an adaptive control var-
0 2 0
iable, the dynamics of asset process is governed by the
following equation:

⎪
⎧
⎪ dXπ (t) dB(t) dI(t) dS(t)
⎪
⎪ � π0 (t) + π1 (t) + π2 (t)
⎪
⎪ X π (t) B(t) I(t) S(t)
⎪
⎪
⎪
⎨
⎪ 􏽥 􏽢 (4)
⎪
⎪ � 􏼂􏽥r + π1 (t) μ􏽥t + r − 􏽥r􏼁 + π 2 (t) μ􏽢t − 􏽥r􏼁􏼃dt + π1 (t)􏽥σ t dW(t) + π2 (t)􏽢σ t dW(t),
⎪
⎪
⎪
⎪
⎪
⎪
⎩ π
X (0) � x0 .

􏽢
dL(t)
To avoid the influence of control variable on the liability � 􏼐μt − μ􏽥t + σ􏽥2t 􏼑dt − σ􏽥t dW(t)
􏽥 + σ t dW(t),
􏽢
L(t)
process L(t), we also assume the liability process is governed t t
by a geometric Brownian motion as follows: 1 1
􏽢
L(t) � L0 exp􏼨􏽚 􏼒μs − μ􏽥s + σ􏽥2s − σ 2s 􏼓ds − 􏽚 σ􏽥s dW(s)
􏽥
0 2 2 0
t 1 2 t
t
L(t) � L0 exp􏼨􏽚 􏼒μs − σ s 􏼓ds + 􏽚 σ s dW(s)􏼩, L0 > 0. + 􏽚 σ s dW(s)􏼩.
0 2 0
0
(5) (6)
􏽢
The real liability process L(t) after eliminating inflation The asset process after eliminating the influence of in-
􏽢
is defined as L(t) � L(t)/P(t). It is easy to get the following flation is defined as X􏽢 π (t) � Xπ (t)/P(t); then, we get the
form of expression using Itô’s formula: following asset process according to Itô’s formula:
4 Complexity

π practical experience and rigorous mathematical calculation,

dX􏽢 (t)
� 􏽨􏽥r − μ􏽥t + σ􏽥2t + π1 (t)􏼐μ􏽥t + r − 􏽥r − σ􏽥 2t 􏼑 and only can be completed by well-trained and experienced
􏽢 π (t)
X investment managers based on the real market quotation
􏽥 (7) and their practical experiences.
+ π2 (t) μ􏽢t − 􏽥r􏼁􏼃dt + π1 (t) − 1􏼁σ􏽥 t dW(t)
Let Π denote the set of all admissible strategies π. The
􏽢 investor aims to find an optimal portfolio π ∈ Π of the
+ π2 (t)􏽢σ t dW(t).
following original problem:
⎪
⎧
⎪ min Var[Zπ (T)] � E􏼂Zπ (T) − (α + ρ(β − α))􏼃2
⎪
⎪
2.3. Asset-Liability Problem with Regulation Constraints. ⎨ 􏽑
π∈

The asset-liability ratio is defined as Zπ (t) � X 􏽢 π (t)/L(t),

􏽢 ⎪
⎪
⎪ s.t. E[Zπ (T)] � α + ρ(β − α) > 0,
⎪
⎩
which describes the ratio between the size of asset and Zπ (t)satisfy(8), ρ ∈ (0, 1), for π ∈ 􏽑,
the size of liability at time t. According to Itô’s formula,
(10)
the dynamic process of the asset-liability ratio can be given
by where ρ is a constant. Only if ρ takes an unequivocal value,
dZπ (t) then the expectation of Zπ(T) can be fixed at a certain value
� 􏼐􏽥r − μt − σ 2t 􏼑dt − σ t dW(t) + π1 (t) μ􏽥t + r − 􏽥r􏼁dt between the top regulation bound β and the bottom regu-
Zπ (t)
lation bound α.
􏽥
+ σ􏽥t dW(t)􏼁 􏽢 􏼑.
+ π2 (t)􏼐 μ􏽢t − 􏽥r􏼁dt + σ􏽢 t dW(t) For a determined value of expectation α + ρ(β − α), if
there exists at least one admissible pair (Z(·), π(·)) satis-
(8) fying E[Zπ(T)] � α + ρ(β − α), then the problem (10) is
In practice, the regulation only occurs at certain fixed called feasible. Given α + ρ(β − α), the optimal strategy π∗
time which is called regulatory time. The regulatory time is of (10) is called an efficient strategy. The pair
∗ ∗

usually selected in advance, which provides a time period of (Zπ (T), Var[Zπ (T)]) under this optimal strategy π∗ is
appropriate length, but the length may decrease if the called an efficient point. The set of all efficient points is
regulatory frequency increased. In accordance with the called the efficient frontier. Obviously, problem (10) is a
need of practice, this paper considers the asset-liability dynamic quadratic convex optimization problem, and
problem with the constraints imposed at the regulatory time thus, it has an unique solution.
T to find the optimal investment strategy, in order to reduce
the risk of abnormal high-speed growth of asset price within 3. Solutions of the Problem
a short period of time or high investment leverage and also
to lessen the risk of too low return rate or a sharp fall. Their In this section, an unconstrained optimization problem is
mathematical descriptions of the regulation constraints at derived by the Lagrange multiplier method and then, a
the regulatory time T are formulated as the following Hamilton–Jacobi–Bellman equation is deduced to solve
inequality: the unconstrained problem. Next, solution of the original
problem is obtained according to Lagrange duality be-
α < Zπ (T) < β, (9)
tween the unconstrained problem and the original prob-
where the bottom regulation bound α ≥ 1 and the top reg- lem. At the end, solution of a special case is given for
ulation bound β are two appropriate constants satisfying the comparison.
requirement β > α. The bottom bound α should not be too
small; otherwise, it cannot cover the liability which means
the investment has failed. The top bound β should not be too 3.1. Solution of an Unconstrained Problem. For this convex
large; or else, a much larger value of β may lead to excessive optimization problem (10), the equality constraint E
leverage multiples and too much investment risks being [Zπ(T)] � α + ρ(β − α) can be eliminated by using the
pulled in the company. In practice, it is highly technical to Lagrange multiplier method, so we get a Lagrange dual
determine the values of α and β, which requires abundant problem:

2
⎪
⎧ min
⎨ π∈ E􏼂Zπ (T) − (α + ρ(β − α))􏼃 + 2λE􏼂Zπ (T) − (α + ρ(β − α))􏼃,
􏽑 (11)
⎪
⎩
s.t. Zπ (T)satisfy(8), ρ ∈ (0, 1), for π ∈ 􏽑,
2
E􏼂Zπ (T) − (α + ρ(β − α))􏼃 + 2λE􏼂Zπ (T) − (α + ρ(β − α))􏼃
where λ ∈ R is a Lagrange multiplier.
2
After a simple calculation, it yields � E􏼂Zπ (T) − ((α + ρ(β − α)) − λ)􏼃 − λ2 .
(12)
Complexity 5

Remark 1. The link between problems (10) and (11) is To solve problem (14), a truncated form beginning at
provided by the Lagrange duality theorem, which can be time t is considered here, and the corresponding value
referred in the study by Luenberger [17] that function is defined as
2􏼌
􏼌
min Var􏼂Zπ (T)􏼃 V(t, z) � inf E􏼔 Zπ (T) − ((α + ρ(β − α)) − λ)􏼁 􏼌􏼌􏼌 Zπ (t)�z 􏼕,
π∈ 􏽑
π∈ 􏽑
2
� max min 􏽮E􏼂Zπ (T) − ((α + ρ(β − α)) − λ)􏼃 − λ2 􏽯. (15)
λ∈R π∈ 􏽑
with the boundary condition V(T, z) � (z − ((α+
(13)
ρ(β − α)) − λ))2 .
Thus, for the fixed constants λ and u, problem (11) is The controlled infinitesimal generator for any test
equivalent to function ϕ(t, x) and any admissible control π is deduced as
2 follows:
⎪
⎧ min
⎨ π∈ E􏼂Zπ (T) − ((α + ρ(β − α)) − λ)􏼃
􏽑 (14) Aπ ϕ(t, x) � ϕt + xϕx 􏽨􏽥r − μt − σ 2t + π1 (t) μ􏽥t + r − 􏽥r􏼁
⎪
⎩
s.t. Zπ (t) satisfy(8), ρ ∈ (0, 1), for π ∈ 􏽑 .
x2
+ π2 (t) μ􏽢t − 􏽥r􏼁􏼃 + ϕ 􏽨σ 2 + π21 (t)􏽥σ 2t + π22 (t)􏽢σ 2t 􏽩,
Therefore, in order to solve problem (11), we only need 2 xx t
to solve problem (14). Fortunately, problem (14) can be (16)
solved by using the stochastic optimal control technology.
for any real valued function ϕ(t, x) ∈ C1,2([0, T], R) and
admissible strategy π, where

C1,2 ([0, T], R) � 􏼈ψ(t, x)|ψ(t, ·) is once continuously differentiable on [0, T], and ψ
(17)
(·, x) is twice continuously differentiable almost surely on R}.

If V(t, z) ∈ C1,2([0, T] × R), then V(t, z) satisfies the which can be given more specifically as follows:
following HJB equation:
inf Aπ V(t, z) � 0, (18)
π∈ 􏽑

z2 z2
inf 􏼨Vt + zVz 􏼐􏽥r − μt − σ 2t 􏼑 + Vzz σ 2t + zVz 􏼂π1 (t) μ􏽥t + r − 􏽥r􏼁 + π2 (t) μ􏽢t − 􏽥r􏼁􏼃 + Vzz 􏼐π21 (t)􏽥σ 2t + π22 (t)􏽢σ 2t 􏼑 � 0􏼩. (19)
π∈ 􏽑 2 2

For convenience, some simple notations are introduced Solving equation (19) gives the following important
as follows: theorem.
2 2
μ􏽥t + r − 􏽥r􏼁 μ􏽢t − 􏽥r􏼁
δ≜ + ,
σ􏽥2t σ􏽢 2t Theorem 1. For the asset-liability management with in-
vestment regulation and eliminating inflation, the efficient
􏼐e(t− T)(δ+σ 2t )
δ + σ 2t 􏼑 investment strategy π∗ � (π∗0 , π∗1 , π∗2 ) corresponding to
ϱ(t) ≜ , problem (11) and problem (14) is given by
δ+ σ 2t
⎪
⎧ 1
⎪
⎪
⎪ π∗0 � 1 + δ􏼒1 − ξ βα (t)􏼓,
(α + ρ(β − α)) ⎪
⎪ z
ω≜ , (20) ⎪
⎪
(1 − ϱ􏽢) ⎪
⎪
⎪
⎪
⎪
⎪ μ􏽥 + r − 􏽥r􏼁 ξ βα (t)
⎨ ∗
π1 � t 2 􏼠 − 1􏼡,
􏼐e− T(δ+σ t ) δ + σ 2t 􏼑
2
⎪ σ􏽥 t z (21)
ϱ􏽢 ≜ ϱ(0) � , ⎪
⎪
⎪
⎪
δ + σ 2t ⎪
⎪
⎪
⎪
⎪
⎪ μ􏽢t − 􏽥r􏼁 ξ βα (t)
⎪
e− T(− 􏽥r+δ+μt +σ t ) z0 ⎪
2 ∗
⎪ π
⎩ 2 � 􏼠 − 1􏼡,
ε≜ . σ􏽢 2 t
z
(1 − ϱ􏽢)
and the value function is
6 Complexity

T)(− 2􏽥
r+δ+2μt +σ 2t ) Substituting these derivatives of W(t, z) into (24), the
V(t, z) � z2 e(t−
HJB equation becomes
− 2ze(t− T)(− 􏽥
r+δ+μt +σ 2t ) (22)
+((α + ρ(β − α)) − λ) ϱ(t), 2 f′ (t)z2 + g′ (t)z + h′ (t) + 􏼐2f(t)z2 + g(t)z􏼑􏼐􏽥r − μt − σ 2t 􏼑

T)(􏽥
where ξ βα (t) � e(t− r− μt )
((α + ρ(β − α)) − λ). g2 (t)
+ z2 f(t)σ 2t − δ􏼠f(t)z2 + g(t)z + 􏼡 � 0.
4f(t)
Proof. According to the first-order necessary condition of
extremum, equation (19) yields (29)
⎧
⎪
⎪ μ􏽥 + r − 􏽥r􏼁 Vz
⎪
⎪ π ∗1 (t) � − t 2 , Equation (29) can be split into three ordinary differential
⎪
⎪ z􏽥σ t Vzz equations as follows:
⎨
⎪ (23)
⎪
⎪ f′ (t) + 2f(t) 􏽥r − μt 􏼁 − f(t)σ 2t − δf(t) � 0,
⎪
⎪ μ􏽢 − 􏽥r􏼁 Vz ⎪
⎧
⎪
⎪ π∗2 (t) � − t 2
⎩ . ⎪
⎪
z􏽢σ t Vzz ⎪
⎪
⎪
⎪
⎪
⎪
⎪
⎪ f(T) � 1,
Plugging the above expressions of π∗1 (t) and π∗2 (t) into ⎪
⎪
⎪
⎪
(19), the HJB equation becomes ⎪
⎪
⎪
⎪
z2 V2 ⎪
⎪ g′ (t) + g(t) 􏽥r − μt − σ 2t 􏼁 − δg(t) � 0,
Vt + zVz 􏼐􏽥r − μt − σ 2t 􏼑 + Vzz σ 2t − δ z � 0. (24) ⎪
⎪
⎪
⎪
2 2Vzz ⎨
⎪ (30)
Based on the boundary condition V(T, z) � ⎪ g(T) � − 2((α + ρ(β − α)) − λ),
⎪
⎪
⎪
(z − ((α + ρ(β − α)) − λ))2 , we try a conjecture ⎪
⎪
⎪
⎪
⎪
⎪
W(t, z) � f(t)z2 + g(t)z + h(t), ⎪
⎪
(25) ⎪
⎪ δ g(t)2
⎪
⎪ h ′ (t) − � 0,
⎪
⎪ 4 f(t)
for equation (24), which satisfies ⎪
⎪
⎪
⎪
⎪
⎪
f(T) � 1, ⎪
⎩
g(T) � − 2((α + ρ(β − α)) − λ), (26) h(T) � ((α + ρ(β − α)) − λ)2 .
2
h(T) � ((α + ρ(β − α)) − λ) .
Solving these ordinary differential equations in (30)
The derivatives of this conjecture W(t, z) are given as yields
follows:
f(t) � e(t− T)(− 2􏽥r+δ+2μt +σ t ) ,
2
⎪
⎧
⎪
Wt � f′ (t)z2 + g′ (t)z + h′ (t), ⎪
⎪
⎨
(t− T)(− 􏽥
r+δ+μt +σ 2t ) (31)
Wz � 2f(t)z + g(t), (27) ⎪ g(t) � − 2e
⎪
⎪
((α + ρ(β − α)) − λ),
Wzz � 2f(t). ⎪
⎩
h(t) � ((α + ρ(β − α)) − λ)2 ϱ(t).
Plugging the expressions of Wz and Wzz into (23), the
optimal strategy becomes Plugging the explicit expression of f(t) and g(t) into (28),
and using the equality π∗0 (t) + π∗1 (t) + π∗2 (t) � 1, the opti-
⎪
⎧
⎪ μ􏽥 + r − 􏽥r􏼁 g(t)
⎪
⎪ π∗1 (t) � − t 2 􏼠z + 􏼡, mal strategy is obtained as follows:
⎪
⎪ z􏽥 σ 2f(t)
⎨ t
⎪ (28)
⎪
⎪
⎪
⎪ μ􏽢t − 􏽥r􏼁 g(t)
⎪
⎩ π ∗2 (t) � − 􏼠z + 􏼡.
z􏽢σ 2t 2f(t)

⎪
⎧
⎪ ((α + ρ(β − α)) − λ)
⎪
⎪
⎪ π∗0 (t) � 1 + δ􏼠1 − e(t− T)(􏽥r− μt ) 􏼡,
⎪
⎪ z
⎪
⎪
⎪
⎪
⎪
⎨ ∗ μ􏽥 + r − 􏽥r􏼁 μ􏽥 + r − 􏽥r􏼁 (t− T)(􏽥r− μt )
⎪ π1 (t) � − t 2 + t 2 e ((α + ρ(β − α)) − λ), (32)
⎪
⎪ σ􏽥t z􏽥σ t
⎪
⎪
⎪
⎪
⎪
⎪
⎪ ∗
⎪ μ􏽢t − 􏽥r􏼁 μ􏽢t − 􏽥r􏼁 (t− T)(􏽥r− μt )
⎪
⎩ π2 (t) � − 􏽢2 + e ((α + ρ(β − α)) − λ).
σt z􏽢σ 2t
Complexity 7

Denoting ⎧
⎪
⎪
⎪ ∗ ζ βα (t)
⎪ π � 1 − δ 􏼠 − 1􏼡,
ξ βα (t)≜e(t− T)(􏽥r− μt ) ((α + ρ(β − α)) − λ), (33) ⎪
⎪
⎪
0
z
⎪
⎪
⎪
⎪
the optimal strategy can be simplified into ⎪
⎪
⎪
⎪
⎨ μ􏽥 + r − 􏽥r􏼁 ζ βα (t)
⎪
⎧ 1 ⎪ π∗1 � t 2 􏼠 − 1􏼡, (38)
⎪
⎪
⎪ π∗0 (t) � 1 + δ􏼒1 − ξ βα (t)􏼓, ⎪
⎪
⎪ σ􏽥t z
⎪
⎪ z ⎪
⎪
⎪
⎪ ⎪
⎪
⎪
⎪ ⎪
⎪
⎪ ⎪
⎪
⎪
⎨ ∗ μ􏽥 + r − 􏽥r􏼁 μ􏽥 + r − 􏽥r􏼁 β
⎪
⎪
⎪ ∗ μ􏽢t − 􏽥r􏼁 ζ βα (t)
π 1 (t) � − t 2 + t 2 ξ α (t), ⎪
⎩ π2 � 􏽢2 􏼠 z − 1􏼡,
⎪ 􏽥
σ z􏽥σ t (34) σt
⎪
⎪ t
⎪
⎪
⎪
⎪ and the optimal value function is given by
⎪
⎪
⎪
⎪ μ􏽢t − 􏽥r􏼁 μ􏽢 − 􏽥r􏼁
⎪
Var􏽨Zπ (T)􏽩 � e− T(− 2􏽥r+δ+2μt +σ t ) z20 +(ω − ε)2 ϱ􏽢
∗ 2
⎪ ∗
⎩ π2 (t) � − 2 + t 2 ξ βα (t).
σ􏽢 t z􏽢σ t
− 2e− T(− 􏽥r+δ+μt +σ t ) (ω − ε)z0 − (ε − ϱ􏽢ω)2 ,
2

Meanwhile, substituting the expressions of f(t), g(t), and

h(t) into W(t, z), the explicit solution of HJB equation is (39)
obtained as follows:
where
W(t, z) � z2 e(t− T)(− 2􏽥r+δ+2μt +σ t ) +((α + ρ(β − α)) − λ)2
2
T)( 􏽥
r− μt )
ζ βα (t) � e(t− (ω − ε). (40)
(t− T)(− 􏽥
r+δ+μt +σ 2t )
ϱ(t) − 2ze ((α + ρ(β − α)) − λ).
(35)
Proof. The optimal value function V(0, 􏽢 z0 ) can be ob-
□ tained with the equivalence between problem (11) and
problem (14). By taking z0 � Z(0) and setting t � 0 in V(t, z), it
The following verification theorem shows the obtained yields
results are exactly the optimal strategy and the optimal value
function as required. 􏽢 0, z0 􏼁 � V 0, z0 􏼁 − λ2 � f(0)z20 + g(0)z0 + h(0) − λ2
V

� e− T(− 2􏽥r+δ+2μt +σ t ) z20 +((α + ρ(β − α)) − λ)2

2

Theorem 2 (Verification Theorem). If W(t, z) is a solution

ϱ􏽢 − 2e− T(− 􏽥r+δ+μt +σ t ) ((α + ρ(β − α)) − λ)z0 − λ2 .
2
of HJB equation (19) and satisfies the boundary condition
W(T, z) � (z − ((α + ρ(β − α)) − λ))2 , then for all admissi- (41)
ble strategies π ∈ Π, V(t, z) ≥ W(t, z). If π∗ satisfies
T(δ+σ 2t )
2 Since (δ + σ 2t ) > 0, then e− < 1, and it yields
π∗∈ arg inf E􏽨 Zπ (T) − ((α + ρ(β − α)) − λ)􏼁 |Zπ (t) � z􏽩,
π∈ 􏽑 − T(δ+σ 2t )
􏼒e δ+ σ 2t 􏼓 < δ + σ 2t , (42)
(36)
then V(t, z) � W(t, z) and π∗ is the optimal investment and then
􏼐e− T(δ+σ t ) δ + σ 2t 􏼑
2
strategy of problem (14).
ϱ􏽢 � < 1. (43)
δ + σ 2t
Proof. Proof is similar to that given by Pham [18]; Chang
Note that the optimal value function V(0, 􏽢 z0 ) is a
[19]; and so on, so the detail is omitted. □
quadratic function with respect to λ and ϱ􏽢 − 1 is the coef-
ficient of quadratic term:
3.2. Solution of the Original Problem. The optimal value
􏼐e− T(δ+σ t ) δ + σ 2t 􏼑
2
function of problem (11) is defined as (44)
ϱ􏽢 − 1 � − 1 < 0.
􏽢 0, z0 􏼁 � inf E􏽨 Zπ (T) − ((α + ρ(β − α)) − λ)􏼁2 􏽩 − λ2 ,
V δ + σ 2t
π∈ 􏽑
􏽢
Thus, the finite maximum value of V(0, z0 ) can be
(37) obtained at a specific value λ∗ , and
where z0 � Z(0).
e− T(− 􏽥r+δ+μt +σ t ) z0 ϱ􏽢(α + ρ(β − α))
2
∗
In this section, the Lagrange duality method is used to λ � − . (45)
find a solution of original optimization problem (10) based (1 − ϱ􏽢) (1 − ϱ􏽢)
on the obtained results of the unconstrained problem. Applying the Lagrangian duality, substituting λ∗ into
􏽢
V(0, z0 ), the minimum variance corresponding to an ar-
Theorem 3. For original problem (10), the efficient strategy bitrarily given expectation (α + ρ(β − α)) is obtained as
π∗ � (π∗0 , π∗1 , π∗2 ) is given by follows:
8 Complexity

2
(α + ρ(β − α)) e− T(− 􏽥r+δ+μt +σ t ) z0 ⎞
2
π∗
e− T(− 2􏽥r+δ+2μt +σ t ) z20
2
Var􏽨Z (T)􏽩 � + ⎝
⎛ − ⎠
(1 − ϱ􏽢) (1 − ϱ􏽢)

2
+ ρ(β − α)) e− T(− 􏽥r+δ+μt +σ t ) z0 ⎞ e− T(− 􏽥r+δ+μt +σ t ) z0 ϱ􏽢(α + ρ(β − α))⎠
2 2

· ϱ􏽢 − 2e − T(− 􏽥 ⎝(α
r+δ+μt +σ 2t ) ⎛
− ⎠ z0 − ⎝
⎛ − ⎞ .
(1 − ϱ􏽢) (1 − ϱ􏽢) (1 − ϱ􏽢) (1 − ϱ􏽢)
(46)

Substitute (45) into (21), the optimal investment strategy and the expected value of Zπ(t) corresponding to the optimal
of problem (10) is given by parameter value ρ∗ is given by
⎪ π∗ � 1 − δ􏼒1ζ β (t) − 1􏼓, z
α + ρ∗ (β − α) � 0 e− T(− 􏽥r+δ+μt +σ t ) .
2
⎧
⎪ (50)
⎪
⎪
⎪ 0
⎪ z α ϱ􏽢
⎪
⎪
⎪
⎪
⎪
⎪ Substituting ρ∗ into (47), we can get the optimal strategy
⎪
⎨ ∗ μ􏽥 + r − 􏽥r􏼁 μ􏽥 + r − 􏽥r􏼁 β
π1 � − t 2 + t 2 ζ α (t), corresponding to the optimal value α + ρ∗ − α) of expecta-
⎪ (47)
⎪
⎪ 􏽥
σ t σt
z􏽥 tion. Moreover, plugging the expression of ρ∗ into the
⎪
⎪ ∗
⎪
⎪ formula of Var[Zπ (T)], the optimal minimum variance
⎪
⎪ Var[Z π∗ ,ρ∗
(T)] can also be determined corresponding to the
⎪
⎪ μ􏽢t − 􏽥r􏼁 μ􏽢 − 􏽥r􏼁
⎪
⎪ ∗
+ t 2 ζ βα (t), uniquely optimal value (z0 /􏽢ϱ)e− T(− 􏽥r+δ+μt +σ t ) of the
2
⎩ π2 � − 2
σ􏽢 t σt
z􏽢 expectation.
where
+ ρ(β − α)) e− T(− 􏽥r+δ+μt +σ t ) z0 ⎞
2

ζ βα (t) � e(t− T)(􏽥 ⎝(α

r− μt ) ⎛
− ⎠. 3.3. Special Case. The investment regulation is also im-
1 − ϱ􏽢 1 − ϱ􏽢 portant for an investment process without liabilities. Let
μt � σ t � 0, the original problem degenerates to the case of no
(48) liability, and the process Zπ(t) becomes the dynamics process
X􏽢 π (t), as follows:
□
π
􏽢 (t) � 􏽨􏽥r + π1 (t) μ􏽥t + r − 􏽥r􏼁 + π2 (t) μ􏽢t − 􏽥r􏼁 − π1 (t)􏽥σ 2t − μ􏽥t 􏽩
dX
Remark 2. However, in the obtained results (47) and (46) 􏽢 π (t)dt + π1 (t) − 1􏼁σ􏽥t X
􏽢 π (t)dW(t)
􏽥
·X + π2 (t)􏽢σ t
above, the expected value (α + ρ(β − α)) of Zπ(t) is not fixed
at a determined position but can take any value between α 􏽢 π (t)dW(t).
·X 􏽢
and β depending on the value of ρ ∈ (0, 1). Therefore, the (51)
obtained minimum variance depends on the parameter
value ρ ∈ (0, 1) in the expression of expectation Meanwhile, the regulation bound constraints also de-
(α + ρ(β − α)). Next, let us turn our attention to the pa- generate into a restriction on the investment dynamics
rameter ρ and determine the optimal parameter value ρ∗ for process after eliminating inflation:
minimizing the variance with respect to ρ as well. The 􏽢 π (T) < β. (52)
∗ α<X
optimal ρ∗ to achieve the minimum of Var[Zπ (T)] is given
as follows:
􏼐􏼐e− T(− 􏽥r+δ+μt +σ t ) z0 􏼑/􏽢ϱ􏼑 − α
2
Proposition 1. For the special case μt � σ t � 0, the minimum
ρ∗ � , (49)
(β − α) variance of Zπ(t) is

2
􏽥 2 􏽥 2
∗ δ⎜
− T􏽥⎛
⎜
⎜ (α + ρ(β − α)) e− T − r+δ+􏽥σ t 􏼁 x
􏽢0⎞
⎟
⎟ − T − r+􏽥
δ+􏽥
2
⎜
⎜(α
σt 􏼁 ⎛
⎜ + ρ(β − α)) e− T − r+δ+􏽥σ t 􏼁 x
􏽢0⎞
⎟
⎟
Var􏼂Z (T)􏼃 � e ⎜ ⎝ − ⎟
⎠ − 2e
⎟ ⎜
⎝ − ⎟
⎠x
⎟ 􏽢0
1 − e− T􏽥
δ 1 − e− T􏽥δ 1 − e− T􏽥δ 1 − e− T􏽥δ
􏼒 􏼓 􏼒 􏼓 􏼒 􏼓 􏼒 􏼓
(53)
2
􏽥􏽥􏼁 2
r+δ+σ t 􏽥
⎜e−
⎛ T −
􏽢 0 − e− Tδ (α + ρ(β − α))⎞
x ⎟
⎟ T − 2r+􏽥
δ+2􏽥
2
σt 􏼁 2
− ⎜
⎜
⎜
⎝ ⎟
⎟
⎠ +e
−
􏽢0,
x
− T􏽥
δ
􏼒1 − e 􏼓

and the optimal strategy of the special case is given as follows:

Complexity 9

T) − r+􏽥 􏽥 2
2
δ+􏽥
μ􏽥t + r − 􏽥r − σ􏽥 2t 􏼁 ⎛
⎜
⎜ e(t− σt 􏼁
(α + ρ(β − α)) − e− T − r+δ+􏽥σ t 􏼁 x
􏽢0⎞
⎟
⎟
π∗1 (t) �1− ⎜
⎜
⎝1 − ⎟
⎟
⎠,
(t− T) − 2r+􏽥
2 2
􏽥
σ􏽥 t 􏽢e
x δ+2􏽥
σt 􏼁
􏼒1 − e− Tδ 􏼓

T) − r+􏽥 􏽥 2
2
δ+􏽥
μ􏽢t − 􏽥r􏼁 ⎛
⎜
⎜ e(t− σt 􏼁
(α + ρ(β − α)) − e− T − r+δ+􏽥σ t 􏼁 x
􏽢0⎞
⎟
⎟
π∗2 (t) � − ⎜
⎝1 −
⎜ ⎟
⎟
⎠, (54)
(t− T) − 2r+􏽥
2 2
􏽥
σ􏽢 t δ+2􏽥
σt 􏼁 − T δ
􏽢e
x 􏼒1 − e 􏼓

T) − r+􏽥 􏽥 2
2
δ+􏽥
⎜
⎛ e(t− σt 􏼁
(α + ρ(β − α)) − e− T − r+δ+􏽥σ t 􏼁 x
􏽢0⎞
⎟
π∗0 (t) � δ􏽥⎜
⎜
⎜
⎝1 −
⎟
⎟
⎟
⎠,
(t− T) − 2r+􏽥
2
􏽢e δ+2􏽥
σt 􏼁 − T􏽥
δ
x 􏼒1 − e 􏼓

μ􏽥t + r − 􏽥r􏼁 ζ βα (t)

where π∗1 � 􏼠 − 1􏼡,
σ􏽥2t z
2 2
μ􏽥 + r − 􏽥r − σ􏽥2t 􏼁 μ􏽢t − 􏽥r􏼁 (58)
δ􏽥 � t + . (55)
σ􏽥 2t σ􏽢 2t μ􏽢 − 􏽥r􏼁 ζ β (t)
π∗2 � t 2 􏼠 α − 1􏼡,
σ􏽢t z

Proof. The results can be obtained by the same calculation where ζ βα (t) � e(t− T)(􏽥r− μt ) (ω − ε), ω � (α + ρ(β − α))/
(1 − ϱ􏽢), and ε � (e− T(− 􏽥r+δ+μt +σ t ) z0 )/(1 − ϱ􏽢), it is easy to find
2

process as the original problem, so the details are omitted

here. □ that understanding implications of these simplified nota-
tions δ, ϱ􏽢, ω, ε, and ζ βα (t) is the key point. Subsequently, the
interpretations will start from descriptions of these sim-
Remark 3. The expected value (α + ρ(β − α)) is not fixed at plified notations.
a determined position in the above proposition, but it can In notation δ � ((􏽥μt + r − 􏽥r)2 /􏽥σ 2t ) + ((􏽢μt − 􏽥r)2 /􏽢σ 2t ), the
take any value between α and β depending on the value of one part (􏽥μt + r − 􏽥r)2 /􏽥σ 2t is the risk compensation rate of
ρ ∈ (0, 1). Therefore, the obtained minimum variance inflation-linked index bond and the other part (􏽢μt − 􏽥r)2 /􏽢σ 2t is
depends on the parameter value ρ ∈ (0, 1). Using the first- the risk compensation rate of risky stock. Thus, δ synthet-
order necessary condition of extremum value for Var ically reflects the risk compensation of two risky assets.
2
[Z∗ (T)] with respect to the parameter ρ, the optimal value Obviously, e− T(δ+σ t ) < 1, so the notation
2
of parameter ρ is given by ϱ􏽢 � (e− T(δ+σ t ) δ + σ 2t )/(δ + σ 2t ) is less than 1. Notation ϱ􏽢 gives
2 a ratio of one sum with the risk compensation rate δ being
T − r+􏽥
σt 􏼁
∗ e− 􏽢0 − α
x (56)
2
converted by e− T(δ+σ t ) divided by the other sum in which the
ρ � ,
(β − α) risk compensation rate δ has no conversion.
The notation ω � (α + ρ(β − α))/(1 − ϱ􏽢) is an expression
and the corresponding optimal expectation of top regulation bound β and bottom regulation bound α,
T − r+􏽥
2
σt 􏼁
where α + ρ(β − α) is a value of the expectation E[Zπ(T)].
α + ρ∗ (β − α)􏼁 � e− 􏽢0.
x (57) Because ρ takes a determined value, the changes of regu-
lation bounds directly result in the changes of ω. Therefore,
Substituting ρ∗ into the optimal strategy (π ∗0 (t),
ω is the key point reflecting the influences of both regulation
π∗1 (t), π∗2 (t)), the optimal strategy corresponding to the
bounds on investment strategy.
optimal value α + ρ∗ (β − α) of expectation can be obtained.
∗ In notation ε, the parameter z0 has the most obvious
Meanwhile, plugging the expression of ρ∗ into Var[Zπ (t)],
∗ ∗ economic connotation, since it represents the initial value of
the optimal minimum variance Var[Zπ ,ρ (T)] can also be
asset-liability ratio at time 0. The expression
determined. It is worth mentioning that the optimal
(e− T(− 􏽥r+δ+μt +σ t ) z0 )/(1 − ϱ􏽢) gives a value resulted from the
2

minimum variance for this special case with no liability

initial asset-liability ratio z0 influenced by several different
is zero under the uniquely optimal expectation
factors (return rates and volatilities) and then converted at
(z0 /􏽢ϱ)e− T(− 􏽥r+δ+μt +σ t ) .
2

the constant relative rate ϱ􏽢.

The following expression
4. Interpretations of the Main Results − T −􏽥
ζ βα (t) e(t− T)(􏽥r− μt ) 􏼐 (α + ρ(β − α)) − e (
r+δ+μt +σ 2t )
z0 􏼑
� ·
The expressions of the obtained results are so abstract that it z (1 − ϱ􏽢) z
is necessary to give some interpretations for the main results (59)
to clarify their specific meaning, especially for the invest-
ment share π1(t) of the inflation-linked index bond and the is key to understand the optimal strategy (π∗0 , π∗1 , π∗2 ). The
investment share π2(t) of the risky stock. denominator z of this fraction ζ βα (t)/z represents the current
Scrutinizing features of the expressions π1(t) and π2(t), level of asset-liability ratio at time t.
10 Complexity

But the numerator ζ βα (t) of this fraction is more Table 1: Values of the parameters.
complicated. 2 The one multiplier ((α + ρ(β − α))−
μ􏽥 � 0.08 r � 0.01 σ􏽥 � 0.2 α�1 β�7 z0 � 3 z�5
z0 e− T(− 􏽥r+δ+μ+􏽥σ t ) ) is a difference between the expected value
(α + ρ(β − α)) of Z(T) and the converted value 􏽥r � 0.05 T � 6 μ􏽢 � 0.15 σ􏽢 � 0.6 μ � 0.03 σ � 0.1 ρ � 0.5
e− T(− 􏽥r+δ+μt +σ t ) z0 of the initial value z0, which characterizes
2

the distance between initial value affected by various eco-

nomic factors and the expectation of Z(T). The other π1∗ changes with α when top bound β fixed
multiplier e(t− T)(􏽥r− μt ) /(1 − ϱ􏽢) is a converted value compre- 1.0
hensively influenced by several parameters embodying the
states of financial market. 0.8
In addition, it is easy to see that if the relative rate ζ βα (t)/z
is larger than 1, π∗1 (t) � ((􏽥μt + r − 􏽥r)/􏽥σ 2t )((ζ βα (t)/z) − 1) and
0.6
π∗2 (t) � ((􏽢μt − 􏽥r)/􏽢σ 2t )(ζ βα (t)/z − 1) are positive; otherwise,

(π1)∗
both will be negative. Since π ∗0 (t) � 1 − π∗1 (t) − π∗2 (t), it
indicates that π∗0 (t) � 1 + δ(1 − (1/z)ζ βα (t)) can be deter- 0.4
mined as long as the investment shares of risky assets have
been determined, so the adjustment of risk-free asset is 0.2
usually passive. If the relative rate ζ βα (t)/z is greater than 1,
the investment shares on risk-free asset can be ensured
within a much better range between 0 and 1. These ob- 1 2 3 4 5 6
servations can serve as an important reference for deter- t
mining both the top regulation bound β and the bottom α=1
regulation bound α. α = 1.8
α = 2.6
Figure 1: Values of π∗1 change with t for different bottom bound α.
5. Numerical Examples
This section discusses the variation tendency of optimal
strategies π∗1 and π∗2 varying with some important parame- π∗2 changes with α when top bound β fixed
ters, mainly focusing on the current value z of asset-liability
ratio, the top regulation bound β, the bottom regulation 0.20
bound α, and the parameter ρ. The values of parameters are
given in Table 1; unless otherwise specified, other values for
0.15
the same variable can be consulted from the corresponding
graphic legend.
(π2)∗

First, let top regulation bound β be fixed; if the bottom 0.10

regulation bound takes much higher values, then the in-
vestment shares of both inflation-linked index bond and risky
stock increase significantly. These variety trends can be ob- 0.05
served intuitively from Figures 1 and 2, respectively. Mean-
while, Figures 1 and 2 also show that the appropriate bottom
regulation bound should be taken at one value between 1 and 1 2 3 4 5 6
2 for an economic environment same as this example. t
Otherwise, the investment shares on risky assets may increase α=1
too much, but the investment share on risk-free asset becomes α=2
negative, which may result in leverage risk increases rapidly. α=3
Next, let bottom regulation bound α be fixed; if the top Figure 2: Values of π∗2 change with t for different bottom bound α.
regulation bound β takes much higher value, then the invest-
ment shares of both inflation-linked index bond and risky stock
should also increase significantly. The variety trends can be bound means allowing a wider volatility range of investment
perceived from Figures 3 and 4, respectively. Meanwhile, returns, which may lead to the result that assets with greater
Figures 3 and 4 also show that the appropriate top regulation investment risk are allowed to invest, or greater market risks
bound should be taken at one value between 6 and 7 for an are incorporated into the wealth process. In addition, the
economic environment same as this example. Otherwise, the above numerical analysis roughly gives a valuable reference
investment shares on risky assets increase too much, but the for selecting an appropriate range of top regulation bound
investment share on risk-free asset becomes negative, which and bottom regulation bound according to the relationship
also result in leverage risk increases rapidly. between the reasonable values of investment strategy and
As can be seen from Figures 3–6, taking the larger value regulation bounds for an economic environment set by
of either the top regulation bound or the bottom regulation values of parameters.
Complexity 11

π∗1 changes with β when bottom bound α fixed π∗1 changes with t for different z
0.8 0.8

0.6 0.6

(π1)∗
(π1)∗

0.4 0.4

0.2 0.2

1 2 3 4 5 6 1 2 3 4 5 6
t t

β=6 z=4
β=7 z=5
β=8 z=6

Figure 3: Values of π∗1 change with t for different top bound β. Figure 5: Values of π∗1 change with t for different current asset-
liability ratio z.

π∗2 changes with β when bottom bound α fixed noted that advanced statistical analysis and maturely
0.4 practical experience of personnel are essential and of great
advantage for determining an exact value of α or β.
Assuming that both top regulation bound β and bottom
0.3 regulation bound α are determined, the impact of asset-li-
ability ratio z at time t on investment strategy can be ob-
served in Figures 5 and 6. If the current asset-liability ratio z
(π2)∗

0.2 at time t takes much greater value, then the investment

shares of both inflation-linked index bond and risky stock
should be much lower, which shows that increasing shares of
0.1 investment on risk-free asset is an important measure to
reduce investment risk. For the numerical examples shown
in Figures 5 and 6, the regulation bounds being fixed at α � 1
and β � 7, the most appropriate asset-liability ratio should
1 2 3 4 5 6
take values around 5 which is much more advantageous for
t
the company.
β=6
β=8
The impact of parameter ρ on investment strategy can be
β = 10 observed in Figures 7 and 8. If the parameter ρ takes a much
greater value, then the investment shares of risky assets must
Figure 4: Values of π∗2 change with t for different top bound β. increase correspondingly. Since the top regulation bound β
and the bottom regulation bound α are predetermined, the
In fact, the regulatory bound cannot be determined expected value α + ρ(β − α) of asset-liability ratio entirely
arbitrarily. Both the top regulatory bound and bottom depends on the value of parameter ρ. Therefore, the variation
regulatory bound determine the fluctuation range that de- tendency of investment shares π∗1 (t) and π∗2 (t) with pa-
cision-making needs to control. Too narrow fluctuation rameter ρ actually illustrates the trend of investment shares
range of investment returns results in loss of great invest- changing with the expectation of asset-liability ratio Z(T). In
ment opportunities and makes investment in the financial one word, the greater the expectation of Z(T), the much
market meaningless. If the regulatory bound is too large, the higher the shares of the company’s wealth to be invested on
potential huge risks cannot be dealt with effectively when the risky assets.
investment return soars fast, which is almost equivalent to
situations without regulatory bounds and also lose the 6. Conclusion
significance of using regulatory bounds. It is hard to de-
termine the regulatory bounds only by quantitative calcu- Some empirical studies in the banking literature deal with
lation. Once there is a model error, it may result in serious the effects of prolonged periods of low interest rates in the
deviation between theory and practice in actual manage- economy on risk-taking, real effects on the real sector and
ment; thus, the determination of upper and lower limits also on financial stability and so on as given by Chaudron
should not be purely theoretical. Nevertheless, it should be [20]; Bikker; and Vervliet [21]. However, this paper uses
12 Complexity

π∗2 changes with t for different z π∗2 changes with t for different ρ
0.25

0.4
0.20

0.3
0.15

(π2)∗
(π2)∗

0.2
0.10

0.05 0.1

1 2 3 4 5 6 1 2 3 4 5 6
t t
z=3 ρ = 0.4
z=4 ρ = 0.6
z=5 ρ = 0.8

Figure 6: Values of π∗2 change with t for different current asset- Figure 8: Values of π∗2 change with t for different ρ.
liability ratio z.

Data Availability
π∗1 changes with t for different ρ The data used to support the findings of this study are
available from the corresponding author upon request.
0.8

Conflicts of Interest
0.6
The authors declare that they have no conflicts of interest.
(π1)∗

0.4
Acknowledgments
This research was supported by China Postdoctoral Science
0.2
Foundation funded project (Grant no. 2017M611192),
Youth Science Fund of Shanxi University of Finance and
Economics (Grant no. QN-2017019), and Youth Science
1 2 3 4 5 6 Foundation of National Natural Science Fund (Grant no.
t 11801179).
ρ = 0.4
ρ = 0.5
ρ = 0.6 References
Figure 7: Values of π∗1 change with t for different ρ. [1] W. F. Sharpe and L. G. Tint, “Liabilities- a new approach,” The
Journal of Portfolio Management, vol. 16, no. 2, pp. 5–10, 1990.
[2] A. Kell and H. Müller, “Efficient portfolios in the asset liability
mathematical models and methods to study a problem of context,” ASTIN Bulletin, vol. 25, no. 1, pp. 33–48, 1995.
asset-liability management with financial market risks: risks [3] M. Leippold, F. Trojani, and P. Vanini, “A geometric approach
of too low return rate, risks of the abnormal rapid growth of to multiperiod mean variance optimization of assets and li-
return rate, and risks of investment leverage. A model of abilities,” Journal of Economic Dynamics and Control, vol. 28,
quantitative regulation (both the top regulation bound and no. 6, pp. 1079–1113, 2004.
the bottom regulation bound being imposed on the asset- [4] D. Li and W.-L. Ng, “Optimal dynamic portfolio selection:
liability ratio) is put forward. The efficient strategy and ef- multiperiod mean-variance formulation,” Mathematical Fi-
ficient frontier are obtained under the objective of variance nance, vol. 10, no. 3, pp. 387–406, 2000.
[5] M. C. Chiu and D. Li, “Asset and liability management under
minimization with regulation constraints at the regulatory
a continuous-time mean-variance optimization framework,”
time T. But the most important value of a theoretical re- Insurance: Mathematics and Economics, vol. 39, no. 3,
search lies in its guidance and reference for practice. pp. 330–355, 2006.
Through numerical examples, it is found that the obtained [6] C. Fu, A. Lari-Lavassani, and X. Li, “Dynamic mean-variance
explicit optimal strategy can also provide reference for portfolio selection with borrowing constraint,” European
determination of regulation bounds, which reinforces the Journal of Operational Research, vol. 200, no. 1, pp. 312–319,
theoretical significance of this research work. 2010.
Complexity 13

[7] J. Pan and Q. Xiao, “Optimal mean-variance asset-liability

management with stochastic interest rates and inflation risks,”
Mathematical Methods of Operations Research, vol. 85, no. 3,
pp. 491–519, 2017.
[8] D. Li, Y. Shen, and Y. Zeng, “Dynamic derivative-based in-
vestment strategy for mean-variance asset-liability manage-
ment with stochastic volatility,” Insurance: Mathematics and
Economics, vol. 78, pp. 72–86, 2018.
[9] J. Wei and T. Wang, “Time-consistent mean-variance asset-
liability management with random coefficients,” Insurance:
Mathematics and Economics, vol. 77, pp. 84–96, 2017.
[10] Y. Zhang, Y. Wu, S. Li, and B. Wiwatanapataphee, “Mean-
variance asset liability management with state-dependent risk
aversion,” North American Actuarial Journal, vol. 21, no. 1,
pp. 87–106, 2017.
[11] T. B. Duarte, D. M. Valladão, and Á. Veiga, “Asset liability
management for open pension schemes using multistage
stochastic programming under solvency-II-based regulatory
constraints,” Insurance: Mathematics and Economics, vol. 77,
pp. 177–188, 2017.
[12] S. M. Guerra, T. C. Silva, B. M. Tabak, R. A. S. Penaloza, and
R. C. C. Miranda, “Systemic risk measures,” Physica A,
vol. 442, pp. 329–342, 2015.
[13] S. R. S. D. Souza, T. C. Silva, B. M. Tabak, and S. M. Guerra,
“Evaluating systemic risk using bank default probabilities in
financial networks,” Journal of Economic Dynamics and
Control, vol. 66, pp. 54–75, 2016.
[14] F. Black and M. Scholes, “The pricing of options and corporate
liabilities,” Journal of Political Economy, vol. 81, no. 3,
pp. 637–654, 1973.
[15] R. C. Merton, “On the pricing of corporate debt: the risk
structure of interest rates,” The Journal of Finance, vol. 29,
no. 2, pp. 449–470, 1974.
[16] T. C. Silva, S. R. S. Souza, and B. M. Tabak, “Monitoring
vulnerability and impact diffusion in financial networks,”
Journal of Economic Dynamics and Control, vol. 76,
pp. 109–135, 2017.
[17] D. G. Luenberger, Optimization by Vector Space Methods,
Wiley, New York, NY, USA, 1968.
[18] H. Pham, Continuous-time Stochastic Control and Optimi-
zation with Financial Applications, Springer, Berlin, Germany,
2009.
[19] H. Chang, “Dynamic mean-variance portfolio selection with
liability and stochastic interest rate,” Economic Modelling,
vol. 51, pp. 172–182, 2015.
[20] R. Chaudron, Bank Profitability and Risk Taking in a Pro-
longed Environment of Low Interest Rates: A Study of Interest
Rate Risk in The Banking Book of Dutch Banks, DNB Working
Papers, Amsterdam, Netherlands, 2016.
[21] J. A. Bikker and T. M. Vervliet, “Bank profitability and risk-
taking under low interest rates,” International Journal of
Finance & Economics, vol. 23, no. 1, pp. 3–18, 2018.