Seminar with Mark Bowden
Title: " Herd Behaviour as a Source of
Volatility in Financial Markets"
Speaker: Mark Bowden
School of Economics
University of Queensland
Place: 78-621/622 General Purpose South
Time: 10:30am 15th of September, Morning Tea Provided.
Abstract:
Herd behaviour is often sighted as one of the forces behind excess
volatility of stock prices as well as speculative bubbles. The key question
of this paper is whether herd behaviour, as documented in the behavioural
finance literature, can explain volatility. A multi-agent model of herd
behaviour is constructed based on the social learning framework developed in
Bikhchandani et al (1992). The social learning takes place in a network
consistent with the work on small worlds by Watts (1999) and more recently
in an economic context by Cowan and Jonard (2004).
A model consisting entirely of herd agents is developed (the herd agent
compares a signal they receive with the expectations other agents with whom
they are connected, the majority signal becomes the expectation of that
agent). It is shown that herd agents very quickly enter an information
cascade. The model is then augmented with the adding of expert agents and
sticky agents. Expert agents follow their own signal in preference to
engaging in social learning while the sticky agents will not change their
view from the previous round unless faced with overwhelming evidence to the
contrary (this attempts to capture the effects of the behavioural bias known
as cognitive dissonance).
There are a number of key outcomes. Adding a number of expert and sticky
agents to the network will break the cascade and add noise to the aggregate
outcome (with and a contrarian view forming that filters though to some of
the remaining agents). Another important outcome is the application of the
small world network introduces an asymmetric response to the different
states of the world. It also shown that if the number of connections between
agents is increased beyond a threshold then agents enter an information
cascade. This may explain the situation where the majority of agents
continue to hold a view on the market (for example a view that the market
remains in a bull run) despite evidence to the contrary.
Finally, the strength and the number of connections between expert and herd
agents are increased. Increasing the strength of the connections from expert
agents appears to have minimal effect. However, when the number of
connections between each expert agent and herd agents is increased beyond a
certain threshold the system becomes very volatile and the small world
effect is no longer present.
About the Author
Mark Bowden is currently enrolled as a PhD student in the School of
Economics, where he holds a PhD fellowship with the ARC Centre for Complex
Systems. Mark Bowden holds a Maters of Social Science (majoring in
Economics) and undergraduate degrees in economics and science, where he
majored in applied mathematics.
Speaker: Mark Bowden
School of Economics
University of Queensland
Place: 78-621/622 General Purpose South
Time: 10:30am 15th of September, Morning Tea Provided.
Abstract:
Herd behaviour is often sighted as one of the forces behind excess
volatility of stock prices as well as speculative bubbles. The key question
of this paper is whether herd behaviour, as documented in the behavioural
finance literature, can explain volatility. A multi-agent model of herd
behaviour is constructed based on the social learning framework developed in
Bikhchandani et al (1992). The social learning takes place in a network
consistent with the work on small worlds by Watts (1999) and more recently
in an economic context by Cowan and Jonard (2004).
A model consisting entirely of herd agents is developed (the herd agent
compares a signal they receive with the expectations other agents with whom
they are connected, the majority signal becomes the expectation of that
agent). It is shown that herd agents very quickly enter an information
cascade. The model is then augmented with the adding of expert agents and
sticky agents. Expert agents follow their own signal in preference to
engaging in social learning while the sticky agents will not change their
view from the previous round unless faced with overwhelming evidence to the
contrary (this attempts to capture the effects of the behavioural bias known
as cognitive dissonance).
There are a number of key outcomes. Adding a number of expert and sticky
agents to the network will break the cascade and add noise to the aggregate
outcome (with and a contrarian view forming that filters though to some of
the remaining agents). Another important outcome is the application of the
small world network introduces an asymmetric response to the different
states of the world. It also shown that if the number of connections between
agents is increased beyond a threshold then agents enter an information
cascade. This may explain the situation where the majority of agents
continue to hold a view on the market (for example a view that the market
remains in a bull run) despite evidence to the contrary.
Finally, the strength and the number of connections between expert and herd
agents are increased. Increasing the strength of the connections from expert
agents appears to have minimal effect. However, when the number of
connections between each expert agent and herd agents is increased beyond a
certain threshold the system becomes very volatile and the small world
effect is no longer present.
About the Author
Mark Bowden is currently enrolled as a PhD student in the School of
Economics, where he holds a PhD fellowship with the ARC Centre for Complex
Systems. Mark Bowden holds a Maters of Social Science (majoring in
Economics) and undergraduate degrees in economics and science, where he
majored in applied mathematics.
