Nonlinear Dyn (2009) 58: 209–216

DOI 10.1007/s11071-009-9472-5

O R I G I N A L PA P E R

Chaos and Hopf bifurcation of a finance system

Qin Gao · Junhai Ma

Received: 28 September 2008 / Accepted: 24 January 2009 / Published online: 21 February 2009
© Springer Science+Business Media B.V. 2009

Abstract The complex dynamical behavior of a fi- factors. It is the developmental direction of economics
nance system is investigated in this paper. The Ruelle– to utilize the nonlinear dynamics, especially the bifur-
Takens route to chaos and strange nonchaotic attrac- cation and chaos theory to study the internal complex-
tors (SNA) are found through numerical simulations. ity of economy and finance systems [1, 2]. It has wide
Then the system with time-delayed feedback is con- foreground, important theoretical and practical mean-
sidered and the stability and Hopf bifurcation of the ings to do this research further.
controlled system are investigated. This research has Chaos is ubiquitous in nonlinear systems. The re-
important theoretical and practical meanings. search about the mode which the systems change from
simple state, such as equilibrium, periods and quasi-
Keywords Lyapunov exponent · Ruelle–Takens periods, to chaos is an important issue in nonlinear dy-
route · Strange nonchaotic attractor (SNA) · Time namics. For the actual results, researchers have found
delay · Hopf bifurcation four important routes to chaos through experiment,
theoretical analyses and numerical simulation: period
doubling bifurcation, intermittency, quasi-periodicity
1 Introduction (i.e., Ruelle–Takens route) and the broken of KAM
torus, of which the period doubling bifurcation route
is the most familiar and the typical one. The complex
With the great development of nonlinear science, es-
dynamical behavior of a finance system is investigated
pecially the chaotic theory, the investigation on the
in this paper; the saving amount is taken as bifurcation
complex nonlinear system is more and more deep and
parameter. With the decrease of the saving amount, a
many beautiful productions in nature and engineering
route to chaos of the system is found: Ruelle–Takens
field are yielded. In recent years, this research has been
route to chaos, and there are strange nonchaotic attrac-
extended to complicated economy and finance sys-
tors (SNA) during this process.
tems. It is well known that the economy and finance
The SNA is another important production after
systems are very complicated nonlinear systems which
strange attractor. Here, ‘strange’ refers to the nontriv-
are concerned with people and contain many complex ial, complicated geometry of the attractor, and ‘non-
chaotic’ indicates that the maximum Lyapunov expo-
nent of the attractor is nonpositive, and there is thus no
Q. Gao () · J. Ma
School of Management, Tianjin University,
sensitive dependence on initial conditions. Since Gre-
Tianjin 300072, People’s Republic of China bogi [3] found the SNA firstly, many useful produc-
e-mail: gq_qingao@yahoo.com.cn tions have been gained in this area [4–8]. So far, the
210 Q. Gao, J. Ma

SNAs are almost gained in systems under the driven variable, [9–13] offer the simplified finance model as
of quasiperiodic force, but the system in this paper is ⎧
not under the drive of quasiperiodic force and it also ⎨ ẋ = z + (y − a)x,
brings SNAs: this is also a new result on SNA. ẏ = 1 − by − x 2 , (1)
⎩
Besides the research on the route to chaos and ż = −x − cz,
SNAs of the system, the time-delayed feedback is also
where a ≥ 0 is the saving amount, b ≥ 0 is the cost
considered and the stability and Hopf bifurcation of
per investment, and c ≥ 0 is the elasticity of demand
the controlled system are investigated. Time delays are
of commercial markets.
frequently encountered in many practical systems and
References [9–11] studied the dynamical character
very often are the main cause for poor performance
of this system under different parameter combinations.
and instability of system. Modern researches on fi-
Due to the change of interest rate and the appear-
nance systems seem to be generating a renewed inter-
ance of many investment channels, the saving amount
est in differential equations with time delay terms. In
has changed enormously and has far-reaching influ-
view of this, the issue of time delay system is a topic of
ence on the development of the economy, so it has im-
great practical importance, which has attracted a great
portant theoretical and practical meanings to study the
deal of interest for several decades. In the final of this effect of the change of the saving amount on the whole
paper, the features of the finance system with time de- finance system.
lay are investigated.
The organization of this paper is as follows. In
Sect. 2, the model of a finance system is given. In
3 Analysis and numerical simulations of the model
Sect. 3, the finance system is analyzed and a route
to chaos is found through numerical simulation. In
Sect. 4, the stability of the equilibrium and the exis- Consider the change of the saving amount a of the
tence of Hopf bifurcation of the time delayed finance system and keep other parameters fixed. From the
system are discussed. Section 5 gives the conclusions. analysis of [9–11] we can see that the unique equi-
librium (0, b1 , 0) of the system is a stable sink when
c − b − abc < 0 and c + a − b1 > 0. Here we take
b = 0.1, c = 1; it is clear that the above conditions are
2 Finance system model true when a > 9. So a = 9 is a critical value where the
system will lose its stability.
References [9–13] have reported a dynamic model of Consider the Jacobi matrix of system (1) at equilib-
finance which is composed of four sub-blocks: pro- rium (0, 10, 0):
duction, money, stock and labor force, and expressed ⎛ ⎞
as three first-order differential equations. The model −a + 10 0 1
describes the time variation of three state variables: J =⎝ 0 −0.1 0 ⎠
−1 0 −1
the interest rate, x, the investment demand, y, and the
price index, z. The factors that influence changes in the character equation of which is
x mainly come from two aspects: first, contradictions
from the investment market, i.e., the surplus between λ̄ + 0.1 λ̄2 + (a − 9)λ̄ + (a − 9) = 0.
investment and savings, and second, structural adjust-
ment from the prices. The changing rate of y is in pro- Obviously, the eigenvalues corresponding to equi-
portion to the rate of investment, and in proportion to librium (0, 10, 0) are λ̄1 = −0.1 < 0, λ̄2,3 = 0; when
an inversion with the cost of investment and interest a = 9, then (0, 10, 0) is a higher order equilibrium.
rates. Changes in z, on the one hand, are controlled by This is just the case which is not considered in [9–11].
a contradiction between supply and demand in com- When a < 9, the eigenvalues corresponding to equi-
mercial markets, and on the other hand, are influenced librium (0, 10, 0) are λ̄1 = −0.1 < 0, λ̄2 > 0, λ̄3 < 0,
by inflation rates. By choosing an appropriate coor- then (0, 10, 0) is a saddle point. Figure 1 shows the
dinates and setting proper dimensions for every state Lyapunov exponents’ spectrum of system (1) with the
Chaos and Hopf bifurcation of a finance system 211

Fig. 1 Lyapunov exponents’ spectrum of system (1)

Fig. 2 Bifurcation diagram of system (1)

increasing of parametera, where b = 0.1, c = 1. Fig- (2) For 6.41 < a ≤ 6.6 and 7.03 < a ≤ 7.09, λ1 = 0,
ure 2 shows the bifurcation diagram of system (1) in λ2 = 0, λ3 < 0, system (1) has a 2-torus as shown
the a ∼ x plane. in Fig. 3(b).
Assume that the Lyapunov exponents of system (1) (3) For 7.1 < a ≤ 8.94, λ1 = 0, λ2 < 0, λ3 < 0, sys-
are λi for i = 1, 2, 3, satisfying λ1 > λ2 > λ3 . Then tem (1) has a periodic orbit as shown in
the dynamical behavior of system (1) can be classified
Fig. 3(c)–(e).
as follows:
(4) For 8.95 < a ≤ 10, λ1 < 0, λ2 < 0, λ3 < 0, system
(1) For 0 < a ≤ 6.42 and 6.61 < a ≤ 7.02, λ1 > 0, (1) is stabilized to a equilibrium point.
λ2 = 0, λ3 < 0, system (1) has a chaotic attractor
as shown in Fig. 3(a). Figure 3(f) shows the SNA for a = 7.03.
212 Q. Gao, J. Ma

Fig. 3 Typical dynamical behavior of system (1): (a) a = 6, (b) a = 7.07, (c) a = 7.15, (d) a = 7.2, (e) a = 8.3, (f) a = 7.03

Figures 1 and 2 show that with the decreasing of From the above theoretical analysis and numerical
the saving amount a, the economic fluctuation is more simulation we can see that the system will stabilize
and more acute, and the system is more and more un- to equilibrium (0, 10, 0) ultimately, when a = 9, b =
stable to be chaotic ultimately. Whereas if the saving 0.1, c = 1, but at this time the interest rate and the
amount a is too plentiful, the system will lose en- price index are all zero; this is unpractical. When a
ergy, which is coincident with the practice instance. decreases a tiny amount, the system state will soon
So if the saving amount a will be kept at a proper change from a stable equilibrium to a 1-periodic move-
level, then the economy may develop quickly and ment. This is not Hopf bifurcation because the char-
healthily. acter equation has not a pair of pure imaginary solu-
Chaos and Hopf bifurcation of a finance system 213

tions but a duplicate zero solutions. The 2-periodic, Lemma 3 The equilibrium P0 (0, b1 , 0) is stable when
4-periodic movements and 2-torus will appear in the 1 + ac − bc > 0 and c + a − b1 > 0.
system with the even decrease of a, the system will
enter into chaotic state until a decreases to a proper In this section, we investigate the Hopf bifurcation
value, and that is to say, it is the Ruelle–Takens route to near the unique equilibrium P0 (0, b1 , 0) only. In what
chaos. There are SNAs during this process, for exam- follows, we assume that the coefficients in system (2)
ple, when a = 7.03, the Lyapunov exponent of system satisfy the following condition:
(1) is −0.0012 < 0 but the system is fractal. All these
c 1
complex changes can be seen clearly in the numerical (H1 ) 1 + ac − > 0 and c + a − > 0.
b b
simulation above.
In the following, we focus on the existence of local
Hopf bifurcation at equilibrium P0 (0, b1 , 0) of the sys-
4 Hopf bifurcation of the model with time delay tem (2). Let x = x, y = y − b1 , z = z; the linearization
of system (2) at the equilibrium (0, 0, 0) can be writ-
The effects of the time delay on nonlinear systems ten as:
⎧
have long been investigated [12, 14]. The main pur- ⎨ ẋ(t) = ( b1 − a)x(t) + z(t),
pose of this section is to investigate the effects of the ẏ(t) = −by(t) + K[y(t) − y(t − τ )], (3)
⎩
time delay on the solutions of (2). The time delay is ż(t) = −x(t) − cz(t),
absolute in realistic finance systems. For predigesting
the investigation, here we only put time delay on in- the characteristic equation of which is given by
vestment demand y. By adding time delay feedback to  
1 c
the second equation of system (1), one can obtain the λ + c+a−
2
λ + 1 + ac −
b b
following new system:  
−λτ
× λ − −b + K − Ke = 0. (4)
⎧
⎪ ẋ(t) = z(t) + [y(t) − a]x(t),
⎨ Let P = c + a − b1 , Q = 1 + ac − bc ; apparently
ẏ(t) = 1 − by(t) − x 2 (t) + K[y(t) − y(t − τ )], then we have the following result.
⎪
⎩
ż(t) = −x(t) − cz(t), 
Lemma 4 The two roots λ1,2 = 12 (−P ± P 2 − 4Q)
(2) of (4) have always negative real parts when τ ≥ 0 with
the condition (H1 ).
where τ ≥ 0 is time delay and K indicates the strength
of the feedback. It is easy to see that λ1,2 is real, provided that the
System (2) is equivalent to system (1) when τ = 0. following condition is true:
For system (1), we have the following result [9–11].
(H2 ) P 2 − 4Q > 0.
Lemma 1 When c − b − abc ≤ 0, i.e.,1 + ac − bc > 0,
system (1) has a unique equilibrium Next we only need to consider the distribution of the
root of the following transcendental equation:
1
P0 0, , 0 . λ − −b + K − Ke−λτ = 0. (5)
b
Obviously, iω (ω > 0) is a root of (5) if and only
Lemma 2 When c − b − abc > 0, i.e.,1 + ac − bc < 0, if ω satisfies
system (1) has three equilibria:
iω + b − K + K(cos ωτ − i sin ωτ ) = 0.
1
P0 0, , 0 , and
b Separating the real and imaginary parts, we have
  
c − b − abc 1 + ac 1 c − b − abc b − K + K cos ωτ = 0,
P± ± , ,∓ . (6)
c c c c ω − K sin ωτ = 0.
214 Q. Gao, J. Ma

From (6) we have Lemma 6 Consider the exponential polynomial

ω2 = 2Kb − b2 . (7) P λ, e−λτ1 , . . . , e−λτm

(0) (0)
Obviously, (7) has not real solution when K ≤ b2 , = λn + p1 λn−1 + · · · + pn−1 λ + pn(0)
 (1) 
and when K > b2 , (7) has a unique solution + p1 λn−1 + · · · + pn−1 λ + pn(1) e−λτ1
(1)

  
ω+ = 2Kb − b2 . (8) + · · · + p1(m) λn−1 + · · · + pn−1
(m)
λ + pn(m) e−λτm

Substitute (8) into (6) , then we have where τi ≥ 0 (i = 1, 2, . . . , m) and pj(i) (i = 0, 1, 2,

. . . , m; j = 1, 2, . . . , n) are constants. As (τ1 , τ2 ,
1 K − b 2j π . . . , τm ) vary, the sum of the order of the zeros of
τj = arccos + , j = 0, 1, 2, . . . . (9)
ω+ K ω+ P (λ, e−λτ1 , . . . , e−λτm ) on the open right half-plane
can change only if a zero appears on or crosses the
Suppose λ(τ ) = α(τ ) + iω(τ ) denotes the root of imaginary axis.
(5) near (0, 0, 0) satisfying α(τj ) = 0 and ω(τj ) =
ω+ , j = 0, 1, 2, . . . ; it is not difficult to verify the fol- From Lemmas 1–6, it is easy to obtain the follow-
lowing result. ing theorem:
dλ
Lemma 5 The transversality conditions Re{ dτ }τ =τj Theorem 1 Suppose that the conditions (H1 ) and
> 0 (j = 0, 1, 2, . . .) hold. (H2 ) are satisfied.
(i) If τ ∈ [0, τ0 ), then the equilibrium P0 of system
Proof Substitute λ(τ ) into (5) and differentiating it
(2) is asymptotically stable and unstable when
with respect to τ , it is easy to obtain:
τ > τ0 .
dλ dλ (ii) System (2) can undergo a Hopf bifurcation at
+ Ke−λτ − τ − λ = 0. (10) the equilibrium P0 when τ = τj (j = 0, 1, 2, . . .),
dτ dτ
where τj is defined by (9).
It follows that
Proof (i) It is easy to see that for (5), whenτ = 0, the
dλ −Kλe−λτ characteristic equation is
= . (11)
dτ 1 + Kτ e−λτ
λ + b = 0.
From (6) we have
  Then the root of (5) had negative real part, that is to
dλ say, the equilibrium P0 is locally stable when τ = 0.
Re
dτ τ =τj While τ0 is the minimum τj at which the real parts of
  these roots are zero. So P0 is locally asymptotically
iKω+
= Re iω τ stable if τ ∈ [0, τ0 ); and unstable if τ > τ0 .
e + j − Kτj
(ii) From Lemmas 1–6 and the conditions (H1 ) and
Kω+ sin ω+ τj
= (H2 ), it is easy to see that the conditions for the Hopf
(cos ω+ τj − Kτj )2 + (sin ω+ τj )2 bifurcation to occur in [16] are hold, then system (2)
ω+2 will undergo a Hopf bifurcation at the equilibrium P0
= > 0. (12) when τ = τj (j = 0, 1, 2, . . .), where τj is defined
(cos ω+ τj − Kτj )2 + (sin ω+ τj )2
by (9). 
Then we obtain the result of the lemma. 
Now we perform some numerical simulations of
Now, we can employ a result from [15] to analyze system (2) to verify the result of Theorem 1. Take
(5), which is, for the convenience of the reader, stated a = 15, b = 0.1, c = 1, k = 1, which satisfy the con-
as follows. ditions in Theorem 1, and τ0 = 1.03473 at this time.
Chaos and Hopf bifurcation of a finance system 215

Fig. 4 The time history phases of system (2) with different τ

Take different τ in system (2) and give the correspond- a satisfying conclusion, so we omit it. The direction
ing time history phases in the following. of the Hopf bifurcation and the stability of bifurcating
From the simulations we can see clearly that: periodic solutions are important issues which we shall
consider in the future.
(1) The system (2) will stabilize to the equilib-
rium (0, 0, 0) when τ ∈ [0, τ0 ), i.e. system (2)
is asymptotically stable, as shown in Fig. 4(a),
5 Conclusion
τ = 0.8 < τ0 .
(2) Hopf bifurcation will occur when τ = τ0 , i.e. the
Based on the results of many researchers, the intrinsic
final state of the system (2) will change from an
complexity of a finance system is investigated in this
equilibrium to a cycle, as shown in Fig. 4(b),
paper. A route that leads system to chaos is found by
τ = τ0 = 1.03473, the variables x and z are still
numerical simulation: Ruelle–Takens route to chaos,
stabilized to zero but the variable y is periodic.
and there are SNA during this process. From the result
(3) The system (2) is unstable when τ > τ0 , as shown
of this paper we can see that it is a complex trans-
in Fig. 4(c), τ = 1.1 > τ0 , the variables x and z are
formation process for the system changes from stable
still stabilized to zero, but the amplitude of vari-
equilibrium to chaos, different states such as periodic
able y is larger and larger.
states, quasiperiodic states and SNAs will appear in
In fact, all the three state variables are influenced this process. For the quick and healthy development
by time delays; we have put time delay on the other of the economy, not too active to be chaotic which is
two variables but the calculation and proof about the difficult to forecast and control, nor too faded to lose
existence of Hopf bifurcation are too complex to gain the energy. The saving amount must be kept at a proper
216 Q. Gao, J. Ma

level through adjusting the interest rate properly; it is 6. Yalcınkaya, T., Lai, Y.-C.: Blowout bifurcation route to
also the foundation to stabilize the price and keep the strange nonchaotic attractors. Phys. Rev. Lett. 77(25),
society balanced. 5039–5042 (1996)
7. Khovanov, I.A., Khovanova, N.A., McClintock, P.V.E., An-
The effect of the time delay on the finance system
ishchenko, V.S.: The effect of noise on strange nonchaotic
is investigated also. The time delay τ is taken as the attractors. Phys. Lett. A 268(46), 315–322 (2000)
bifurcation parameter and we show that when τ passes 8. Wenping, H., Guolin, F., Xinquan, G., Jifan, C.: Dynam-
through some certain critical values, the equilibrium ics of the Lorenz system under quasiperiodic driving. Acta
will lose its stability and Hopf bifurcation will take Phys. Sin. 55(06), 3175–3179 (2006) (in Chinese)
place. 9. Junhai, M., Yushu, C.: Study for the bifurcation topological
The research of this paper is a new breakthrough structure and the global complicated character of a kind of
non-linear finance system (I). Appl. Math. Mech. 11(22),
about the research of the routes which lead the com- 1119–1128 (2001) (in Chinese)
plex system to chaos and a new investigation about the 10. Junhai, M., Yushu, C.: Study for the bifurcation topological
stability and Hopf bifurcation on this finance system; structure and the global complicated character of a kind of
it has important theoretical and practical value. non-linear finance system (II). Appl. Math. Mech. 12(22),
1236–1242 (2001) (in Chinese)
11. Junhai, M.: The Reconstruction Technology of Complex
References Nonlinear System. Tianjin University Press, Tianjin (2005)
12. Chen, W.-C.: Dynamics and control of a financial sys-
1. Chian, A.C.-L., Rempel, E.L., Rogers, C.: Complex eco- tem with time-delayed feedbacks. Chaos Solitons Fractals
nomic dynamics: Chaotic saddle, crisis and intermittency. 37(4), 1198–1207 (2008)
Chaos Solitons Fractals 29(5), 1194–1218 (2006) 13. Chen, W.-C.: Nonlinear dynamics and chaos in a fractional-
2. Fanti, L., Manfredi, P.: Chaotic business cycles and fis- order financial system. Chaos Solitons Fractals 36(5),
cal policy: An IS-LM model with distributed tax collection 1305–1314 (2008)
lags. Chaos Solitons Fractals 32(2), 736–744 (2007) 14. Yan, X.-P., Zhang, C.-H.: Hopf bifurcation in a delayed
3. Grebogi, C., Ott, E., Pelikan, S., Yorke, J.A.: Strange at- Lokta—Volterra predator—prey system. Nonlinear Anal.
tractors that are not chaotic. Physica D Nonlinear Phenom.
Real World Appl. 9(1), 114–127 (2008)
13(1–2), 261–268 (1984)
4. Heagy, J.F., Hammel, S.M.: The birth of strange nonchaotic 15. Ruan, S., Wei, J.: On the zeros of transcendental functions
attractors. Physica D Nonlinear Phenom. 70(1–2), 140–153 with applications to stability of delay differential equations
(1994) with two delays. Dyn. Contin. Discrete Impuls. Syst. Ser. A
5. Prasad, A., Biswal, B., Ramaswamy, R.: Strange non- Math. Anal. 10, 863–874 (2003)
chaotic attractors in driven excitable systems. Phys. Rev. 16. Yushu, C.: Bifurcation and Chaos Theory of Nonlinear Vi-
E 68, 037201 (2003) bration Systems. Higher Education Press, Beijing (1995)