NetLogo banner

 Home
 Download
 Resources
 Extensions
 FAQ
 References
 Contact Us
 Donate

 Models:
 Library
 Community
 Modeling Commons

 User Manuals:
 Web
 Printable
 Chinese
 Czech
 Japanese

  Donate

NetLogo User Community Models

(back to the NetLogo User Community Models)

Herding

by Carlos Pedro Gonçalves (Submitted: 11/04/2003)

[screen shot]

Download Herding
If clicking does not initiate a download, try right clicking or control clicking and choosing "Save" or "Download".

(You can also run this model in your browser, but we don't recommend it; details here.)

WHAT IS IT?

This is a model of herding behaviour in a financial market. The market goes from disorganized states with
great diversity of opinions to organized behaviour where there are clusters of opinions regarding the buying
or selling of a financial asset. If the selling clusters dominate there is a high probability of a crash, if, on the other hand, the buying clusters dominate, there is a high probability of a bubble.
Market clearing is acheived through a mechanism used that was partly inspired in Sornette (2003) and partly inspired in Farmer (1998).

HOW IT WORKS

Patches are financial agents that at each time can either buy or sell an asset. The decision of each agent
to buy depends on information that the agent gathers on his own (which is considered to be random and on the
opinions of the neighbouring agent's. The model of decision is taken from Sornette () and is close to a spin glass model.

HOW TO USE IT

There is one critical parameter that controls the herding behaviour, this is the degree of importance an agent gives to his neighbours (the coupling parameter). If the attention is high then we observe herding behaviour if it is small then there is a disorganized state.
The critical parameter cannot be controled directly it has a more or less periodic behaviour with a random component, the equation for this parameter is given by:

A = (abs (sin (f)) * 0.03)
f(t)=f(t-1)+z(t), where z(t) is a normally distributed random variable.

We can control the parameters of the random component, namely, the mean and standard deviation of z. These two parameters lead to different dynamics in terms of herding behaviour, for instance the higher the standard deviation the less the periodic component is felt. The higher the mean, the more the market tends towards an organized behaviour.

The user can also control the parameters of the random component of the signal that each agent receives specifically the standard deviation and the mean. This expresses the agents own opinion. If one increases the mean towards a positive value, there is a bias towards buying, if one increases the standard deviation the market stays in the disorganized state longer. The standard deviation of the own signal also affects stability of opinions.

THINGS TO NOTICE

Notice how the market passes from a disorganized state towards an organized state, and see what happens to the price.

THINGS TO TRY

Run the model once as is presented then explore the parameter space by changing first the parameters that control the herding behaviour, then the own signal and finally the noise traders. This order of exploration of parameter space provides important information concerning the influence of the various elements on the model.

EXTENDING THE MODEL

Possible extensions could be:
- Other dynamics for the coupling parameter.
- Different topology for the agents network connections;
- A model with random contacts, instead of a fixed neighbourhood of contacts;
- Alter the market making mechanism.

CREDITS AND REFERENCES

Sornette, Didier (2003), Critical Market Crashes, http://arXiv:cond-mat/0301543
Farmer (1998), Market, Force, Ecology, Evolution, Santa Fe Working Papers

(back to the NetLogo User Community Models)