25.06.2013 Views

statistique, théorie et gestion de portefeuille - Docs at ISFA

statistique, théorie et gestion de portefeuille - Docs at ISFA

statistique, théorie et gestion de portefeuille - Docs at ISFA

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

A Corcos <strong>et</strong> al Q UANTITATIVE F INANCE<br />

p t<br />

0.9<br />

0.8<br />

0.7<br />

0.6<br />

0.5<br />

0.4<br />

0.3<br />

0.2<br />

0.1<br />

0 1000 2000 3000 4000 5000 6000 7000 8000 9000 10000<br />

t<br />

Figure 4. Evolution of the system over 10 000 time steps for<br />

N =∞, m = 60 polled agents and the param<strong>et</strong>ers<br />

ρhb = ρbh = 0.72 and ρhh = ρbb = 0.85. Note the existence of<br />

acceler<strong>at</strong>ions ending in crashes and anti-crashes (fast jumps).<br />

and ∂pH >c>0, where λ is the Lyapunov exponent for pt,as<br />

follows from δπt+1 = ∂πH · δπt + ∂pH · δpt. This condition is,<br />

in particular, s<strong>at</strong>isfied for a law of the form πt+1 = πt + G(pt),<br />

where G is strictly monotonic. Thus, chaotic behaviour of<br />

bullish agents leads to chaotic behaviour of prices.<br />

In the following, we shall take the simplest form of a<br />

log-difference of the price linearly proportional to the or<strong>de</strong>r<br />

unbalance (Farmer 1998), leading to<br />

ln πt+1 − ln πt ≡ rt+1 = γ(pt − 1<br />

), (8)<br />

2<br />

showing th<strong>at</strong> the r<strong>et</strong>urn rt calcul<strong>at</strong>ed over one period is<br />

proportional to the imbalance pt − 1<br />

. Thus, the properties<br />

2<br />

of the r<strong>et</strong>urn time series can be <strong>de</strong>rived directly from those of<br />

pt as we document below. This linear function of (pt − 1<br />

) can 2<br />

be improved for instance by the hyperbolic tangent function<br />

tanh(pt − 1<br />

) documented in Plerou <strong>et</strong> al (2001).<br />

2<br />

To summarize this qualit<strong>at</strong>ive analysis of the case of an<br />

infinite number N of agents, we observe a time evolution<br />

which, while s<strong>at</strong>isfying certain criteria of randomness (such<br />

as possessing an absolutely continuous invariant measure<br />

and exhibiting a positive Lyapunov exponent (compare with<br />

Eckmann and Ruelle 1985)) <strong>at</strong> the same time exhibits some<br />

regularities on short time scales, since it is d<strong>et</strong>erministic.<br />

Our mo<strong>de</strong>l thus establishes th<strong>at</strong> straightforward fundamental<br />

conditions may suffice to gener<strong>at</strong>e chaotic stock mark<strong>et</strong><br />

behaviour, <strong>de</strong>pending on the param<strong>et</strong>er values. If the mark<strong>et</strong><br />

adjusts present mark<strong>et</strong> price on the basis of expect<strong>at</strong>ions and<br />

mimicry—self-referred behaviour—then chaotic evolution of<br />

the popul<strong>at</strong>ion will also imply chaotic evolution of prices.<br />

4. Quantit<strong>at</strong>ive analysis of the<br />

specul<strong>at</strong>ive bubbles within the chaotic<br />

regime in the symm<strong>et</strong>ric case<br />

For an infinite number N of agents and in the symm<strong>et</strong>ric case<br />

ρhb = ρbh ≡ ρ1 and ρhh = ρbb ≡ ρ2, l<strong>et</strong> us rewrite the<br />

270<br />

p t –1/2<br />

p t –1/2<br />

0.25<br />

167<br />

0.20<br />

0.15<br />

0.10<br />

0.05<br />

0.00<br />

0 100 200 300 400 500 600<br />

10 0<br />

t<br />

10 –1<br />

10 –2<br />

10 –3<br />

10 –4<br />

0 100 200 300<br />

t<br />

400 500 600<br />

Figure 5. The first bubble of figure 4 for N =∞agents with<br />

m = 60 polled agents and param<strong>et</strong>ers ρhb = ρbh = 0.72 and<br />

ρhh = ρbb = 0.85.<br />

dynamical evolution (2) of the system as<br />

p ′ m<br />

<br />

m<br />

= p − p p<br />

j j=0<br />

m−j (1 − p) j <br />

j<br />

f<br />

m<br />

m<br />

<br />

m<br />

+ (1 − p) (1 − p)<br />

j j=0<br />

m−j p j <br />

j<br />

f ,<br />

m<br />

where<br />

(9)<br />

<br />

1,<br />

f(x)=<br />

0,<br />

if x ρ1 or x

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!