Our reference: CHAOS7667 AUTHOR QUERY FORM Journal: CHAOS Article Number: 7667 Please e-mail your responses and any corrections to: E-mail: [email protected]Dear Author, Please check your proof carefully and mark all corrections at the appropriate place in the proof (e.g., by using on-screen annotation in the PDF file) or compile them in a separate list. Note: if you opt to annotate the file with software other than Adobe Reader then please also highlight the appropriate place in the PDF file. To ensure fast publication of your paper please return your corrections within 48 hours. Your article is registered as a regular item and is being processed for inclusion in a regular issue of the journal. If this is NOT correct and your article belongs to a Special Issue/Collection please contact [email protected]immediately prior to returning your corrections. For correction or revision of any artwork, please consult http://www.elsevier.com/artworkinstructions Any queries or remarks that have arisen during the processing of your manuscript are listed below and highlighted by flags in the proof. Click on the ‘Q ’ link to go to the location in the proof. Location Query / Remark: click on the Q link to go in article Please insert your reply or correction at the corresponding line in the proof Q1 AU: Please confirm that given names and surnames have been identified correctly. Q2 AU: Please verify the presentation of both the affiliations and corresponding author. Q3 AU: Please provide better quality artwork for all figures. Q4 AU: Figs. [3, 6, 10, 17, 18 and 20] have been submitted as color images; however, the captions have been reworded to ensure that they are meaningful when your article is reproduced both in color and in black and white. Please check and correct if necessary. Please check this box or indicate your approval if you have no corrections to make to the PDF file Thank you for your assistance.
24
Embed
Journal: CHAOS Please e-mail your responses and any ...€¦ · paper please return your corrections within 48 hours. Your article is registered as a regular item and is being processed
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Our reference: CHAOS7667
AUTHOR QUERY FORM
Journal: CHAOS
Article Number: 7667
Please e-mail your responses and any corrections to:
Please check your proof carefully and mark all corrections at the appropriate place in the proof (e.g., by using on-screenannotation in the PDF file) or compile them in a separate list. Note: if you opt to annotate the file with software otherthan Adobe Reader then please also highlight the appropriate place in the PDF file. To ensure fast publication of yourpaper please return your corrections within 48 hours.
Your article is registered as a regular item and is being processed for inclusion in a regular issue of the journal. Ifthis is NOT correct and your article belongs to a Special Issue/Collection please contact [email protected] prior to returning your corrections.
For correction or revision of any artwork, please consult http://www.elsevier.com/artworkinstructions
Any queries or remarks that have arisen during the processing of your manuscript are listed below and highlighted byflags in the proof. Click on the ‘Q’ link to go to the location in the proof.
Location Query / Remark: click on the Q link to goin article Please insert your reply or correction at the corresponding line in the proof
Q1 AU: Please confirm that given names and surnames have been identified correctly.
Q2 AU: Please verify the presentation of both the affiliations and corresponding author.
Q3 AU: Please provide better quality artwork for all figures.
Q4 AU: Figs. [3, 6, 10, 17, 18 and 20] have been submitted as color images; however, the captionshave been reworded to ensure that they are meaningful when your article is reproduced both incolor and in black and white. Please check and correct if necessary.
Please check this box or indicate your approval ifyou have no corrections to make to the PDF file
Figures 19 and 20 have been saved directly in eps format by Matlab. Figures 1--18 have been first saved in bitmap or PNG format and then converted in eps through a software called Gimp. If it may help, I can send you Figures 1--18 in the bitmap or PNG format. Please let me know if you wish so.
A. Naimzada, M. Pireddu / Chaos, Solitons and Fractals xxx (2015) xxx–xxx 5
ARTICLE IN PRESSJID: CHAOS [m3Gdc;May 21, 2015;15:42]
Fig. 2. The frequency distribution of US output gap from 1960 to 2009, having kurtosis = 3.61 and Jarque–Bera = 7.17 with p-value = 0.027. Source: US Department
of Commerce and Congressional Budget Office.
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
agents, but without financial sector, the author shows that
the obtained dynamics are in agreement with (some or all
of) the three stylized facts we consider, too.
3. The interaction degree approach
As we shall see in Section 4, when considering both fi-
nancial and real markets, we are led to analyze a nonlinear
high-dimensional system with many parameters. Such fea-
tures do not allow to easily handle that system from an ana-
lytical viewpoint when all the parameters vary, even if we are
able to analytically determine its steady state and the corre-
sponding stability conditions. For this reason, we propose an
approach that consists in studying, as a first step, the frame-
work with isolated markets, which are described by lower-
dimensional subsystems that are simpler to investigate and
whose different behaviors can be quickly classified. Then we
make the parameter representing the interaction degree in-
crease, keeping the other parameters fixed. In such way we
are able to analytically find (if it exists) the set of values of
the interaction parameter that implies stability. Moreover,
the use of numerical tools allows us to understand what hap-
pens also in the unstable regime.
This is the strategy we are going to employ in Section 6 to
classify and investigate the various scenarios for our system.
However, in order to make the exposition more fluent, we
start Section 5 with the analysis of the stability conditions
for the case of interacting markets and we derive next the
stability conditions for the framework with isolated markets.
In symbols, if we denote our integrated system by
Sω(x1, . . . , xN), where ω ∈ [0, 1] is the interaction de-
gree parameter and x1, . . . , xN are the endogenous variables
governing the system, when setting ω = 0 we are led to
Please cite this article as: A. Naimzada, M. Pireddu, Real and finan
Solitons and Fractals (2015), http://dx.doi.org/10.1016/j.chaos.2015
study two (or, in general, more) isolated subsystems, we de-
note by S10(x1, . . . , xm) and S2
0(xm+1, . . . , xN), for some m ∈{1, . . . , N − 1}. When instead ω = 1 the subsystems are fully
integrated. The case with ω ∈ (0, 1) represents a partial inter-
action between the subsystems.
In our framework, when ω = 0 we find two isolated sub-
systems, corresponding to the financial and real markets.
The former is described by two variables, i.e., the stock price
and the difference between the shares of optimistic and pes-
simistic agents, while the latter is described by a unique vari-
able, that is, the level of income. We stress that the influence
of the real market on the financial market is due to the fact
that the reference value used in the decisional mechanism
by the agents in the financial market is determined by the
level of income; on the other hand, the investments depend
also on the price of the financial asset. Such a double inter-
action is described by the parameter ω. Notice that we chose
a unique parameter, ω, to represent both the influence of the
real market on the financial sector and vice versa of the finan-
cial sector on the real market, not only to limit the already
large number of parameters in our model, but also in view
of applying the interaction degree approach precisely in the
formulation described above. Of course the framework with
two different parameters describing the mutual relationships
between the two markets could be analyzed in a similar
manner.
4. The model
4.1. The stock market
With respect to the stock market, we assume that agents
are not able to observe the true underlying fundamental. We
cial interacting markets: A behavioral macro-model, Chaos,
Figures 2, 3, 4 and 20 seem to be too large. Please make the size of the figures along the paper more uniform if possible.
6A
.Na
imza
da
,M.P
iredd
u/C
ha
os,So
liton
sa
nd
Fracta
lsxxx
(20
15)
xxx–xxx
ART
ICLE
INPRESS
JID:CH
AO
S[m
3Gdc;M
ay21,
2015;15:42]
Fig. 3. US output gap (in purple-red, with corresponding scaling on the right vertical axis) and Michigan sentiment index (in blue, with corresponding scaling on the left vertical axis) in period 1978–2008. Source: US
Department of Commerce, Bureau of Economic Analysis, and University of Michigan: Consumer Sentiment Index. (For interpretation of the references to color in this figure legend, the reader is referred to the web version
of this article.) Q4
Ple
ase
citeth
isa
rticlea
s:A
.N
aim
zad
a,
M.
Pire
dd
u,
Re
al
an
dfi
na
ncia
lin
tera
cting
ma
rke
ts:A
be
hav
iora
lm
acro
-mo
de
l,C
ha
os,
So
liton
sa
nd
Fracta
ls(2
01
5),h
ttp://d
x.d
oi.o
rg/1
0.1
01
6/j.ch
ao
s.20
15
.05
.00
7
Original text:
Original text:
AU: Figs. [3, 6, 10, 17, 18 and 20] have been submitted as color images; however, the captions have been reworded to ensure that they are meaningful when your article is reproduced both in color and in black and white. Please check and correct if necessary.
A. Naimzada, M. Pireddu / Chaos, Solitons and Fractals xxx (2015) xxx–xxx 11
ARTICLE IN PRESSJID: CHAOS [m3Gdc;May 21, 2015;15:42]
Fig. 5. The bifurcation diagrams with respect to ω ∈ [0, 1] for P in (A) and Y in (B), and the Lyapunov exponent in (C), respectively, for μ = 28 and the initial
conditions P0 = 5, x0 = 0.8 and Y0 = 25.
Fig. 6. The (P, Y)-phase portraits for μ = 28, and ω = 0.25 in (A) and ω = 0.70 in (B), respectively; in (C) the time series for P in red (below) and Y in blue (above)
when μ = 28 and ω = 0.95. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
the case, as the first chain of inequalities reads as −6.25 <
1 < 7.812 and the last inequality is 3.5 < 5 when μ = 5,
while the first chain of inequalities reads as 0.167 < 1 < 1.395
and the last inequality is again 3.5 < 5 when μ = 28. As it
is immediate to see from such calculations, although for the
considered parameter values both isolated markets are sta-
ble, when μ increases the stability region of the financial
subsystem decreases: this may explain the detected grow-
ing destabilization of the whole system in correspondence to
larger values of μ.
As concerns the stability conditions for ω varying in
[0, 1], when μ = 5 we have B1 = 1.2, B2 = 2.386, B3 =−2.4509, B4 = −6.778, and thus (5.2) reads as ω ∈ [0, 1],
that is, the system is stable for any ω, in agreement with
Fig. 4; when instead μ = 28 we have B1 = 1.2, B2 =0.274, B3 = −0.098, B4 = −0.793, and thus (5.2) reads as
ω ∈ [0,√
B2) = [0, 0.523), that is, the system is stable just for
small values of ω, in agreement with Fig. 5.
What we can then conclude in this scenario is that in-
creasing μ has a destabilizing effect. In fact, fixing all the
other parameters as above and letting just μ vary, we find
that∂B1∂μ
= 0,∂B2∂μ
< 0,∂B3∂μ
> 0 and∂B4∂μ
> 0. Hence, the sta-
bility region decreases when μ increases (as B1 does not
vary with μ and the upper bound B2 decreases, while the
lower bounds B3 and B4 increase), confirming the highlighted
destabilizing role of the parameter μ.
Please cite this article as: A. Naimzada, M. Pireddu, Real and finan
Solitons and Fractals (2015), http://dx.doi.org/10.1016/j.chaos.2015
Summarizing, for the above parameter configurations, the
interaction between the financial and real markets either
maintains the stability of the system, or it has a destabilizing
effect, through a flip bifurcation. We stress that such bifurca-
tion differs from the flip bifurcation detected in [26] in two
aspects. The first one is that, as already stressed in the Intro-
duction, those authors deal just with the isolated financial
market, while our flip bifurcation concerns the interaction
between the real and financial markets. The second aspect
is that after the flip bifurcation in [26] some numerical sim-
ulations we performed suggest that the system diverges and
thus such bifurcation would not lead to complex behaviors:
in our framework, the flip bifurcation is followed instead by
a stable period-two cycle which, increasing further the inter-
action parameter, undergoes a secondary double Neimark–
Sacker bifurcation, giving rise to quasiperiodic motions.
6.2. Unstable financial subsystem – stable real subsystem
In the framework we are going to consider, when iso-
lated, the financial subsystem is unstable and characterized
by quasiperiodic motions, while the real subsystem is sta-
ble. For not too large values of the parameter γ , when ω in-
creases, the fixed point becomes stable through a reversed
Neimark–Sacker bifurcation. According to the value of γ , that
fixed point can either persist until ω = 1 or can undergo a flip
cial interacting markets: A behavioral macro-model, Chaos,
12 A. Naimzada, M. Pireddu / Chaos, Solitons and Fractals xxx (2015) xxx–xxx
ARTICLE IN PRESSJID: CHAOS [m3Gdc;May 21, 2015;15:42]
Fig. 7. The bifurcation diagrams with respect to ω ∈ [0, 1] for P in (A) and Y in (B), and the Lyapunov exponent in (C), respectively, for γ = 5 and the initial
conditions P0 = 12, x0 = −0.3 and Y0 = 61.
Fig. 8. The bifurcation diagrams with respect to ω ∈ [0, 1] for P in (A) and Y in (B), and the Lyapunov exponent in (C), respectively, for γ = 8 and the initial
conditions P0 = 12, x0 = −0.3 and Y0 = 61.
Fig. 9. The bifurcation diagrams with respect to ω ∈ [0, 1] for P in (A) and Y in (B), and the Lyapunov exponent in (C), respectively, for γ = 8.8 and the initial
conditions P0 = 12, x0 = −0.3 and Y0 = 61.
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
bifurcation and then a secondary double Neimark–Sacker bi-
furcation. For even larger values of γ , we just obtain a reduc-
tion of the complexity of the system for suitable intermediate
values of ω, but the system is never stabilized.
More precisely, in Figs. 7–9 below, ω varies in [0, 1] and
the other parameters are: F∗ = 2, k = 0.1, α = 0.08, β =1, c = 1, a = 2.4, μ = 28, a1 = 3, a2 = 1, A = 12, b = 0.7,
and γ = 5 in Fig. 7, γ = 8 in Fig. 8, and γ = 8.8 in Fig. 9. In
Fig. 7 the fixed point becomes stable for ω � 0.2 and remains
stable until ω = 1. In Fig. 8, instead of remaining stable, it
undergoes a flip bifurcation for ω � 0.5 and then a secondary
double Neimark–Sacker bifurcation for ω � 0.96. In Fig. 9 the
Please cite this article as: A. Naimzada, M. Pireddu, Real and finan
Solitons and Fractals (2015), http://dx.doi.org/10.1016/j.chaos.2015
fixed point is never stable: we just observe a reduction of the
complexity of the system for ω ∈ (0.2, 0.8), where we have
a stable period-two cycle. In more detail, in Figs. 7 (A)–9 (A)
we show the bifurcation diagrams with respect to ω ∈ [0,
1] for P, while in Figs. 7 (B)–9 (B) we draw the bifurcation
diagrams for Y; in Figs. 7 (C)–9 (C) we depict the Lyapunov
exponents when ω varies in [0, 1].
Let us now check whether the theoretical results in
Section 5 are in agreement with the numerical achievements
above.
First of all let us verify that, for the chosen parameter
sets, when ω = 0 just the real subsystem is stable, i.e., let
cial interacting markets: A behavioral macro-model, Chaos,
A. Naimzada, M. Pireddu / Chaos, Solitons and Fractals xxx (2015) xxx–xxx 13
ARTICLE IN PRESSJID: CHAOS [m3Gdc;May 21, 2015;15:42]
Fig. 10. The bifurcation diagram with respect to ω ∈ [0, 1] for Y in (A), for μ = 20 and the initial conditions P0 = 8, x0 = 0.3 and Y0 = 75 for the blue (darker)
points, and P0 = 4, x0 = 0.25 and Y0 = 60 for the green (lighter) points; the (P, Y)-phase portrait showing the coexistence of the period-six cycle (in blue) with a
period-two cycle (in green) when ω = 0.7948 in (B), for μ = 20 and the initial conditions P0 = 5, x0 = 0.3 and Y0 = 65 for the period-two cycle and P0 = 4, x0 =0.3 and Y0 = 65 for the period-six cycle; the (P, Y)-phase portrait showing the coexistence of the period-six cycle (in blue) with two closed invariant curves (in
green) when ω = 0.826 in (C), for μ = 20 and the initial conditions P0 = 4, x0 = 0.3 and Y0 = 65 for the period-six cycle and P0 = 4, x0 = 0.1 and Y0 = 64 for the
invariant curves. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
us check that the last inequality in (5.6) is fulfilled, but not
the first chain of inequalities therein. A straightforward com-
putation shows that this is the case, as the first chain of in-
equalities reads as 0.116 < 1 < 0.968 and the last inequality
is 5 < 8.888 when γ = 5; the first chain of inequalities reads
again as 0.116 < 1 < 0.968 and the last inequality becomes
8 < 8.888 when γ = 8; finally, the first chain of inequalities
reads once again as 0.116 < 1 < 0.968 and the last inequal-
ity becomes 8.8 < 8.888 when γ = 8.8. As it is immediate
to see from such calculations, although for the considered
parameter values the isolated real market is stable, when γincreases it progressively approaches the instability region:
this may explain the detected growing destabilization of the
whole system when ω increases, in correspondence to larger
values of γ .
As concerns the stability conditions for ω varying in
[0, 1], when γ = 5 we have B1 = 3, B2 = 1.9004, B3 =0.0379, B4 = −2.3117, and thus (5.2) reads as ω ∈ (
√B3, 1] =
(0.194, 1], that is, the system is stable for large values of ω,
in agreement with Fig. 7; when γ = 8 we have B1 = 3, B2 =0.271, B3 = 0.035, B4 = −1.34, and thus (5.2) reads as ω ∈(√
B3,√
B2) = (0.189, 0.521), that is, the system is stable
just for intermediate values of ω, neither too small, nor too
large, in agreement with Fig. 8; finally, when γ = 8.8 we
have B1 = 3, B2 = 0.024, B3 = 0.038, B4 = −1.193, and thus,
since B2 < B3, (5.2) implies that there exists no ω for which
the system is stable, in agreement with Fig. 9.
What we can then conclude is that increasing γ has a
destabilizing effect. In fact, fixing all the other parameters as
above and letting just γ vary, we find that∂B1∂γ
= 0,∂B2∂γ
< 0,
∂B3∂γ
changes sign (in particular, it is negative for γ = 5 and
positive for γ = 8 and γ = 8.8), but B3 always lies in (0, 0.1)
for γ ∈ [5, 8.8] and thus it does not restrict the stability region
too much, and finally∂B4∂γ
> 0. Hence, the stability region de-
creases when γ increases (as B1 does not vary with γ and the
upper bound B2 decreases, while B3 is small and it does not
vary a lot, and the lower bound B4 increases), confirming the
highlighted destabilizing role of γ .
Summarizing, for the above parameter configurations, for
small values of ω, the instability of the financial market gets
Please cite this article as: A. Naimzada, M. Pireddu, Real and finan
Solitons and Fractals (2015), http://dx.doi.org/10.1016/j.chaos.2015
transmitted to the real market. However, increasing ω de-
creases the complexity of the whole system. This effect may
either persist until ω = 1 or it may vanish for larger values of
ω, where we can find instead quasiperiodic motions, follow-
ing a secondary double Neimark–Sacker bifurcation of the
period-two cycle. Pursuing further the comparison with [26],
we stress that when the markets are isolated we are led to
consider instability regimes characterized by quasiperiodic
motions, like those in [26]. On the other hand, our analysis
additionally shows that increasing the interaction degree it is
possible to obtain a stabilization of the whole system, maybe
interrupted by a flip bifurcation.
We remark that in the existing literature on nonlinearities
it is possible to find some examples of stabilization phenom-
ena for interacting oscillating subsystems via an increase of
the interaction degree between the subsystems (see, for in-
stance, [45]). However, differently from our model, in those
examples the subsystems are symmetric and described by
the same functional relation: moreover, to the best of our
knowledge, they do not belong to the economic literature,
but they rather concern biological or physical systems.
6.3. Stable financial subsystem – unstable real subsystem
In this framework, when isolated, the financial sub-
system is stable, while the real subsystem is unsta-
ble. When ω increases, different possible behaviors
may arise. We depict some of them in Figs. 10–13 be-
low, where we have fixed the parameters as follows:
F∗ = 2, k = 0.1, α = 0.08, β = 0.5, c = 1, a = 2.4, γ =11.5, a1 = 3, a2 = 1, A = 12, b = 0.7, while we have μ = 20
in Fig. 10, μ = 28 in Fig. 11 and μ = 31 in Figs. 12 and 13.
More precisely, in Fig. 10 (A) we show the bifurcation di-
agram with respect to ω ∈ [0, 1] for Y, when μ = 20. For
such parameter configuration, when ω ∈ [0, 0.68] there is
just a stable period-two cycle. For ω � 0.68 a period-six cy-
cle emerges, which coexists with the period-two cycle un-
til ω � 0.823 (see the (P, Y)-phase portrait in Fig. 10 (B)
for ω = 0.7948). For ω � 0.823, through a double Neimark–
Sacker bifurcation of the period-two cycle, quasiperiodic mo-
tions emerge, which coexist with the period-six cycle (see
the (P, Y)-phase portrait in Fig. 10 (C) for ω = 0.826). For ω
cial interacting markets: A behavioral macro-model, Chaos,
14 A. Naimzada, M. Pireddu / Chaos, Solitons and Fractals xxx (2015) xxx–xxx
ARTICLE IN PRESSJID: CHAOS [m3Gdc;May 21, 2015;15:42]
Fig. 11. The bifurcation diagrams with respect to ω ∈ [0, 1] for P in (A) and Y in (B), and the Lyapunov exponent in (C), respectively, for μ = 28 and the initial
conditions P0 = 10, x0 = 0.5 and Y0 = 61.
Fig. 12. The bifurcation diagrams with respect to ω ∈ [0, 1] for P in (A) and Y in (B), and the Lyapunov exponent in (C), respectively, for μ = 31 and the initial
conditions P0 = 10, x0 = 0.5 and Y0 = 61.
Fig. 13. The (P, Y)-phase portrait for μ = 31 and ω = 0.1 in (A), ω = 0.4 in (B), ω = 0.7 in (C) and ω = 0.9 in (D), respectively.
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
� 0.830, the quasiperiodic motions end and only the period-
six cycle survives until ω � 0.88, where new quasiperiodic
motions emerge, lasting until ω = 1.
We recall that the coexistence of different kinds of attrac-
tors is also known as multistability. This feature may be con-
sidered as a source of richness for the framework under anal-
ysis because, other parameters being equal, i.e., under the
same institutional and economic conditions, it allows to ex-
plain different trajectories and evolutionary paths. The ini-
tial conditions, leading to the various attractors, represent
indeed a summary of the past history, which in the presence
of multistability phenomena does matter in determining the
Please cite this article as: A. Naimzada, M. Pireddu, Real and finan
Solitons and Fractals (2015), http://dx.doi.org/10.1016/j.chaos.2015
evolution of the system. Such property, in the literature on
complex systems, is also called “path-dependence” [46].
In Figs. 11 (A) and 12 (A), we show the bifurcation dia-
grams with respect to ω ∈ [0, 1] for P, while in Figs. 11 (B) and
12 (B), we show the bifurcation diagrams with respect to ω∈ [0, 1] for Y; in Figs. 11 (C) and 12 (C), we represent the Lya-
punov exponents when ω varies in [0, 1]. Finally, in regard to
the parameter configuration considered in Fig. 12, we show
in Fig. 13 (A) the (P, Y)-phase portrait for ω = 0.1, where we
have a period-two cycle, in Fig. 13 (B) the (P, Y)-phase portrait
for ω = 0.4, where we have a period-eight cycle, in Fig. 13 (C)
the (P, Y)-phase portrait for ω = 0.7, where we have again a
cial interacting markets: A behavioral macro-model, Chaos,
A. Naimzada, M. Pireddu / Chaos, Solitons and Fractals xxx (2015) xxx–xxx 15
ARTICLE IN PRESSJID: CHAOS [m3Gdc;May 21, 2015;15:42]
Fig. 14. The bifurcation diagrams with respect to ω ∈ [0, 1] for P in (A) and Y in (B), and the Lyapunov exponent in (C), respectively, for the initial conditions
P0 = 6, x0 = 0.25 and Y0 = 0.63.
Fig. 15. The (P, Y)-phase portrait for ω = 0 in (A), ω = 0.5 in (B), ω = 0.8 in (C) and ω = 1 in (D), respectively.
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
period-two cycle, and in Fig. 13 (D) the (P, Y)-phase portrait
for ω = 0.9, where we have a chaotic attractor.
Straightforward computations, analogous to those per-
formed in Sections 6.1 and 6.2, show that the theoretical
results in Section 5 are in agreement with the numerical
achievements above.
Hence, for the selected parameter values, the system is
never stable. From the pictures we also notice that increas-
ing μ has a further destabilizing effect, as the complexity of
the system grows when μ moves from 20 to 31. In particular,
in Figs. 11 and 12 we highlight the presence of the so-called
“bubbles” (see [47,48]) and, as shown in Figs. 12 and 13, for
μ = 31 we have even chaotic dynamics for values of ω close
to 1.
Summarizing, in the case of isolated stable financial mar-
ket and unstable real market, for the above parameter config-
urations we have been not able to find stabilizing values for
ω, even if intermediate values of ω may lead to a reduction
of the complexity of the system. In this sense we could argue
that the instability of the real market seems to have stronger
destabilizing effects than the instability of the financial mar-
ket: in fact, the former gets transmitted and possibly ampli-
fied by the connection with the financial market, while, as
we saw in Section 6.2, the latter gets dampened and possibly
eliminated by the connection with the real market.
6.4. Unstable financial and real subsystems
In this last framework, when isolated, the financial
and the real subsystems are unstable. In particular, when
Please cite this article as: A. Naimzada, M. Pireddu, Real and finan
Solitons and Fractals (2015), http://dx.doi.org/10.1016/j.chaos.2015
ω = 0 both of them may be chaotic. When ω increases,
we do not find a complete stabilization of the system,
but we observe some periodicity windows in Figs. 14
and 15 below, where we have fixed the parameters as fol-
lows: F∗ = 2, k = 0.1, α = 0.08, β = 1, c = 1, a = 2.4, μ =28, γ = 20, a1 = 3, a2 = 1, A = 12, b = 0.7. Notice that
these are the same parameter values considered in the
second scenario, except for a larger value of γ . We already
observed in Section 6.2 that increasing γ has a destabilizing
effect for the above parameter configuration and this is
confirmed by Figs. 14 and 15. More precisely, in Fig. 14 (A)
and (B) we show the bifurcation diagrams with respect to
ω ∈ [0, 1] for P and Y, respectively; in Fig. 14 (C) we show
the Lyapunov exponent when ω varies in [0, 1]. In Fig. 15
(A)–(D) we depict, in the (P, Y)-phase plane, the chaotic
regime when ω = 0, a period-eleven cycle when ω = 0.5, a
chaotic attractor when ω = 0.8 and a period-fourteen cycle
when ω = 1, respectively.
Again, it is trivial to check that the theoretical results in
Section 5 are in agreement with the numerical achievements
above, so that, for the above parameter configuration (as well
as for many other ones we investigated), when both the iso-
lated financial and real markets are unstable, we may have a
reduction of the complexity of the whole system until peri-
odic motions, but not a complete stabilization.
6.5. The role of an increasing belief bias
We already stressed at the end of Section 5 the desta-
bilizing role of the belief bias for the financial side of the
cial interacting markets: A behavioral macro-model, Chaos,
16 A. Naimzada, M. Pireddu / Chaos, Solitons and Fractals xxx (2015) xxx–xxx
ARTICLE IN PRESSJID: CHAOS [m3Gdc;May 21, 2015;15:42]
Fig. 16. The bifurcation diagrams for P with respect to a ∈ [2, 7] with μ = 5 in (A) and with respect to a ∈ [2, 3.9] with μ = 28 in (B), respectively, both obtained
for the initial conditions P0 = 10, x0 = 0.25 and Y0 = 50.
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
1001
1002
1003
1004
1005
1006
1007
1008
1009
1010
1011
1012
1013
1014
1015
1016
1017
1018
1019
economy when the two markets are isolated (see (5.5)
and the subsequent comments). We now investigate which
are the effects of an increasing bias on the stability of
the whole system when the markets are interconnected.
Since the results we got are uniform enough across the
various scenarios considered so far, we illustrate our find-
ings just for the scenario in Section 6.1, in which, when
isolated, both the financial and the real subsystems are
stable. Indeed, in Fig. 16 below we have fixed the pa-
rameters as follows: F∗ = 5, k = 0.25, α = 0.08, β = 1, c =1, γ = 3.5, a1 = 2, a2 = 4, A = 5, b = 0.7, ω = 0.9, and μ =5 in Fig. 16 (A), where we show the bifurcation diagram for P
with respect to a ∈ [2, 7], while μ = 28 in Fig. 16 (B), where
we depict the bifurcation diagram for P with respect to a ∈[2, 3.9]. In Fig. 16 (A) the steady state gets destabilized for a �5.9 through a Neimark–Sacker bifurcation; in Fig. 16 (B) the
steady state is instead unstable for values of a close to the ex-
treme values of the considered interval, while it is stable for
intermediate values of a.
Hence, even if the destabilizing role of the bias is clear
when markets are isolated, Fig. 16 suggests that its role
becomes more ambiguous when the markets are intercon-
nected. Indeed, increasing a may have either a destabilizing
or a stabilizing role, according to the value of the other pa-
rameters. However, taking into account also the conclusions
we got for the scenarios in Sections 6.2–6.4 (we do not report
here for the sake of brevity), it seems that increasing a has
generally a destabilizing effect, as usually we do not reach a
complete stabilization of the system, or we achieve it just in
small intervals for the belief bias.
7. Interpretation of the results
It is commonly recognized that agents in making their
choices in the real market are influenced also by the condi-
tions in the asset market, and vice versa. Hence, we wish to
analyze the role of an increase in the intensity of the interac-
tion between the two sectors, starting from a framework in
which both isolated markets are stable, in order to show the
arising economic fluctuations, i.e., booms and busts. To such
aim, we will fix the parameters like in the destabilizing sce-
nario considered in Section 6.1, with μ = 28.5 and ω = 0.99,
Please cite this article as: A. Naimzada, M. Pireddu, Real and finan
Solitons and Fractals (2015), http://dx.doi.org/10.1016/j.chaos.2015
and we will interpret the dynamics of the time series for the
main economic variables of our dynamical system, for t ∈[1030, 1050] in Fig. 17 and for t ∈ [1045, 1075] in Fig. 18. The
variables we will consider are output Y, stock price P and the
profit differential �π = πopt − πpes from (4.10). Moreover,
in analogy with the Michigan consumer confidence indica-
tor considered in Section 2, we define the long-period opti-
mism index as X10,t+1 = xt +xt−1+···+xt−910 . Such variable mea-
sures the average share of optimistic agents in the last ten
periods and its behavior is meant to indicate the dynamics of
animal spirits in the medium run, avoiding considering sud-
den and transient fluctuations, which may make the overall
dynamics more difficult to interpret.
Let us start our analysis, by considering a boom, i.e., a
phase of persistent economic growth. In particular, we focus
our attention on the time series in Fig. 17, where we put in ev-
idence t ′ = 1036. Starting from t = t ′ we observe a predom-
inance of long-period optimism among the agents in the fi-
nancial market, as X10 is positive, and increasing output in the
real sector. Since the excess demand in the stock market, i.e.,
α{[(1 − ω)F∗ + ωkYt ] − Pt + axt} is positive, recalling (4.7), it
follows that the stock price grows. By (4.14) this implies that
output increases due to the wealth effect, as the output dif-
ferential is proportional to aggregate demand. This in turns
makes the stock price increase because, recalling (4.1) and
(4.2), the perceived reference values depend on income and
the same holds for the demands of optimists and pessimists
in (4.3) and (4.4). By (4.5) this makes the stock price increase
further and consequently also the relative profits of optimists
grow by (4.10). The latter phenomenon maintains the over-
all optimism level, so that the share of optimists increases,
together with the long-period optimism index. Due to the
presence of our sigmoidal function in the real sector, such
a virtuous cycle persists until an upper turning point, that in
Figs. 17 and 18 corresponds to t ′′ = 1050, after which a bust
phase starts. Such turning point exists because, due to physi-
cal physical and material constraints, the system is not able to
maintain further positive output variations in the short run,
as capital and labor forces are fixed in the short run. See also
[49], where the authors identify the combination of bound-
edly rational managerial behavior with rigidities and delays
in capacity adjustments as crucial for the occurrence and the
cial interacting markets: A behavioral macro-model, Chaos,
A. Naimzada, M. Pireddu / Chaos, Solitons and Fractals xxx (2015) xxx–xxx 17
ARTICLE IN PRESSJID: CHAOS [m3Gdc;May 21, 2015;15:42]
Fig. 17. The time series for Y (in blue), P (in red), �π (in green) and X10 (in black) for the boom phase, corresponding to the time periods t ∈ [1030, 1050], in
which we put in evidence t ′ = 1036. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)
Fig. 18. The time series for Y (in blue), P (in red), �π (in green) and X10 (in black) for the bust phase, corresponding to the time periods t ∈ [1045, 1075], in which
we put in evidence t ′′ = 1050. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)
1020
1021
1022
1023
1024
1025
1026
1027
1028
1029
1030
1031
1032
1033
1034
1035
nature of boom and bust cycles, and [50] where ample ev-
idence from experiments and case studies is provided that
boundedly rational behavior are quite persistent in boom and
bust cycles. We recall that the role of capacity constraints
is also discussed in the literature on Hicksian business cycle
model [51].
Hence, after the turning point income increases less and
less, and the same holds for the perceived reference values,
Please cite this article as: A. Naimzada, M. Pireddu, Real and finan
Solitons and Fractals (2015), http://dx.doi.org/10.1016/j.chaos.2015
so that the optimists’ profits are reduced and their share in
the population decreases. Thus, also the demand in the finan-
cial market decreases and the stock price starts falling. Due
to the wealth effect expressed in (4.14), this makes the out-
put differential become negative because aggregate demand
decreases and this in turns drives down the perceived refer-
ence values, and consequently, also the excess demand in the
financial market diminishes, making the stock price go down.
cial interacting markets: A behavioral macro-model, Chaos,
A. Naimzada, M. Pireddu / Chaos, Solitons and Fractals xxx (2015) xxx–xxx 19
ARTICLE IN PRESSJID: CHAOS [m3Gdc;May 21, 2015;15:42]
t910 920 930 940 950 960 970 980 990 1000
X10
,t
-0.2
-0.1
0
0.1
0.2
0.3
0.4
Yt
25
30
35
40
45
50
55
60
65
Fig. 20. The time series for output Y (in purple-red, with corresponding scaling on the right vertical axis) and the long-period optimism index X10 (in blue, with
corresponding scaling on the left vertical axis) for the last nearly 100 periods considered in Fig. 19. (For interpretation of the references to color in this figure
legend, the reader is referred to the web version of this article.)
1036
1037
1038
1039
1040
1041
1042
1043
1044
1045
1046
1047
1048
1049
1050
1051
1052
1053
1054
1055
1056
1057
1058
1059
1060
1061
1062
1063
1064
1065
1066
1067
1068
1069
1070
1071
1072
1073
1074
1075
1076
1077
1078
1079
1080
t t t t
Hence, by (4.10) the relative pessimists’ profits increase and
thus also their share, so that the long-period optimism in-
dex decreases. Moreover, by the wealth effect, falling stock
prices make the output differential become more and more
negative. Such a vicious cycle corresponds to a bust phase
which, again due to the presence of our sigmoidal function
in the real sector, persists until a lower turning point, that in
Fig. 18 corresponds to t ′′′ = 1070 (non-explicitly indicated in
that picture) after which a new growth phase starts. Such
turning point exists because, due to physical and material
constraints, the system is not able to maintain further neg-
ative output variations in the short run, as capital and la-
bor forces are fixed in the short run. Hence, after the turn-
ing point income decreases less and less, and the same holds
for the perceived reference values, so that the pessimists’
profits are reduced and their share in the population de-
creases. Thus, the demand in the financial market increases
and the stock price starts raising. Due to the wealth effect,
this makes the output differential become positive because
aggregate demand increases. This in turns raises the per-
ceived reference values, and consequently, also the excess
demand in the financial market grows, making the stock
price goes up. Hence, by (4.10) the relative optimists’ prof-
its increase and thus also their share, so that the long-period
optimism index increases. In such way a new boom gets
triggered.
Please cite this article as: A. Naimzada, M. Pireddu, Real and finan
Solitons and Fractals (2015), http://dx.doi.org/10.1016/j.chaos.2015
We stress that the whole mechanism works for suffi-
ciently large values of the interaction parameter ω. Indeed,
for ω = 0 no relationships between the real and the financial
sectors exist: namely, on the one hand, by (4.14) no wealth
effect is present and, on the other hand, by (4.1) and (4.2) the
reference values are exogenous and do not depend on the real
market performance.
8. Introducing stochastic shocks
Following the approach in [36], we add a noise term to
each of the demand components in (4.3) and (4.4). Such
terms are meant to reflect a certain within-group hetero-
geneity, describing the accidental fluctuations of the com-
position of the many individual digressions from the simple
rules they are supposed to follow. The heterogeneity is rep-
resented by two independent and normally distributed ran-
dom variables εoptt and εpes
t , for the optimists and pessimists,
respectively. Combining the deterministic and stochastic ele-
(npest σ pes)2), and consequently a stochastic term will enter
also the dynamic equation (4.11) governing the switching
mechanism, due to the presence of the profits.
In the next simulations, reported in Fig. 19, we con-
sider the following parameter set: F∗ = 5, k = 0.25, α =0.08, β = 1, c = 1, a = 2, γ = 3.5, a1 = 2, a2 = 4, A =5, b = 0.7,μ = 21, ω = 0.9 and σopt = 0.1 = σpes, where we
denote by σopt and σpes the standard deviation for optimists
and pessimists, respectively. For simplicity, we assumed that
σopt and σpes coincide. Hence, we add the stochastic noises
and observe which are the generated dynamics for variables
Y, P, x, X10 in the most interesting of the previously consid-
ered scenarios, i.e., that in Section 6.1. More precisely, for
Y, P, x, X10 in Fig. 19 we plot the time series (in the first row),
the histograms (in the second row), the autocorrelations (in
the third row) and the Q–Q test plots (in the fourth row). The
initial conditions are Y0 = 0.2, P0 = 0.3, x0 = 0.25, x−1 =0.1, x−2 = · · · = x−9 = 0, and we report 300 values after a
transient of 700 iterations.
The values corresponding to Fig. 19 for the mean, stan-
dard deviation (SD), skewness, kurtosis and Jarque–Bera test
for normality (abbreviated in J–B), with its precise value,
can be found in Table 8.4. We stress that J–B = 0 means
normality, while J–B = 1 means non-normality. Hence, we
find that Y (as well as P) is not normally distributed. Such
conclusion is in agreement with the histograms in Fig. 19,
showing that Y (and P, too) has an higher kurtosis than the
normal distribution (which is equal to 3) and fatter tails. A
further confirmation for those findings is given by Q–Q test
plots in Fig. 19, which shows that Y (and P) are not nor-
mally distributed and that their distributions are fat-tailed.
We recall that the Q–Q plots (Quantile–Quantile plots) plot
the quantiles of one distribution against those of the nor-
mal, contrasting the two cumulative distribution functions.
If the variable under analysis is normally distributed, then
its plot lies on the 45-degree line, which corresponds to
the normal distribution. Moreover, if the considered vari-
able is not normally distributed and its tails lay below
(above) the 45-degree line, then its distribution is fat-tailed
(thin-tailed).
The non-normality of the distribution, and in particular
the presence of fat tails, implies that in our model there are
larger movements in output than is compatible with the nor-
mal distribution, and thus, as desired, booms and busts are
more likely to happen.
Moreover, looking at the time series in Fig. 19, we observe
strong cyclical movements for Y and P, implying that Y and
P are highly autocorrelated, as confirmed by the autocorrela-
tion plots reported in the same figure. We stress that in our
model output Y is the variable to be compared with the out-
Please cite this article as: A. Naimzada, M. Pireddu, Real and finan
Solitons and Fractals (2015), http://dx.doi.org/10.1016/j.chaos.2015
put gap considered in the stylized facts in Section 2.
Mean SD Skewness Kurtosis J − B J − Bvalue
Y 48.1370 7.3197 −0.7195 3.8505 1 34.9293
P 11.3703 2.5983 −0.8192 4.5586 1 63.9219
x 0.0059 0.4208 −0.0061 2.6425 0 1.5990
X10 0.0039 0.0683 −0.0295 2.7397 0 0.8904
(8.4)
Hence, in analogy with the first two stylized facts in
Section 2, we find a strong autocorrelation and a non-normal
distribution for output.
As regards the third stylized fact described in Section 2,
for the same parameter configuration used for Fig. 19, we
show in Fig. 20 the high correlation between the long-period
optimism index X10 (introduced in Section 7) and the move-
ments of output Y. Since X10 plays the role of the Michigan
sentiment index in Fig. 3 and Y that of the output gap therein,
we then find that our model is able to reproduce the third
stylized fact, too.
We stress that we performed the empirical verification
of our model in its stochastic version as adding shocks ac-
counts for taking into account all those aspects that our sim-
ple deterministic model cannot explicitly consider. Of course,
it would be interesting to analyze the consequences of adding
further stochastic noises to the stock price and to the real side
equations, as well as to the switching mechanism. Nonethe-
less we showed that, even adding stochastic terms just to the
demand functions, we have been able to replicate the desired
stylized facts.
9. Conclusion and future directions
In this paper we proposed a model belonging both to the
strand of literature on the interactions between real and fi-
nancial markets, e.g. [24], as well as to the literature about
heterogeneous fundamentalists, as for instance [26]. In fact,
in the model we presented the real economy, described via a
Keynesian good market approach, interacts with a stock mar-
ket with heterogeneous fundamentalists. Agents may switch
between optimism and pessimism according to which be-
havior is more profitable.
More precisely, our main difference with respect to [26] is
that we also considered the real sector of the economy, while
with respect to [24] is that we introduced the interaction de-
gree parameter and the switching mechanism, in order to
describe the changes in the share of agents in the financial
market. To the best of our knowledge, this was the first con-
tribution considering both real and financial interacting mar-
kets and an evolutionary selection process for the popula-
tion for which an analytical study is performed. Indeed, we
employed analytical and numerical tools to investigate the
role of the parameter describing the degree of interaction, in
order to detect the mechanisms and the channels through
which the stability of the isolated real and financial sectors
leads to instability for the two interacting markets. The main
contribution of the present paper to the existing literature
lies in fact in focusing on the role of real and financial feed-
back mechanisms, not only in relation to the dynamics and
stability of a single market, but for those of the economy as
a whole. In order to perform our analysis, we introduced the
cial interacting markets: A behavioral macro-model, Chaos,
The style used for the book in reference [8] is different, for instance, from that of reference [2]. Please change the style in reference [8] if needed.
Marina
Evidenziato
Marina
Nota
"a" should become "A" and "2006b" should become "2006"
Marina
Evidenziato
Marina
Evidenziato
Marina
Nota
"eighties: the" should become "Eighties: The"
Marina
Evidenziato
Marina
Nota
"ScheffknechtL., GeigerF.." should become "Scheffknecht L, Geiger F."
22 A. Naimzada, M. Pireddu / Chaos, Solitons and Fractals xxx (2015) xxx–xxx
ARTICLE IN PRESSJID: CHAOS [m3Gdc;May 21, 2015;15:42]