Transcript Trust, News and the Efficient Markets Hypothesis

BEHAVIOURAL RESEARCH
Trust, News and the Efficient
Markets Hypothesis
Fairness, Trust and Emotions
1 July 2010
Behavioural Finance Working Group
Cass Business School
Efficient markets hypothesis
• Bachelier (1900), Samuelson (1965), Fama
(1970)
• Prices in an efficient market reflect all publicly
available information
• In strong form, prices reflect all information
• So what is information?
What is information?
• Is this information?
– “IBM profits in 2011 will be $15.1 billion”
• How about this:
– “General Electric’s profits in 2006 were $20.9
billion”
• That was “information” until profits were
restated
• It depends on the source...
An extended model of information
• We model beliefs instead of information
• A belief:
– is held by an agent
– has a source
– expresses the relative value of two goods
(typically money and a financial instrument)
– has a confidence level
For instance
• Jim Cramer told me that Intel
stock is worth $28
• I place a confidence level of
5% in this belief
On the other hand...
• Google Finance tells me that
Intel stock is worth $21.22
• I place a confidence level of
70% in this belief
How do I integrate these beliefs?
• I could just weight the confidence levels and
produce an average valuation
• Or I could go with the most trusted source
• Other integration functions are available
Confidence-weighted integration
• Leads to smooth behaviour
• Small changes in trust or value result in small
changes in price
Most-trusted integration
• Leads to volatility
• A small change in trust can result in a big jump
in valuation
Simulation 1
• Integration by weighted probability
Price - weighted confidence
100
90
80
70
60
50
40
30
20
10
0
1
5
9 13 17 21 25 29 33 37 41 45 49 53 57 61 65 69 73 77 81 85 89 93 97
Simulation 2
• Integration by most trusted source
Price - most trusted source
100
90
80
70
60
50
40
30
20
10
0
1
5
9 13 17 21 25 29 33 37 41 45 49 53 57 61 65 69 73 77 81 85 89 93 97
Manipulation test
• We specified one agent as a “promoter” of a
higher price
• We give them a one-off boost of 0.05 to trust,
representing an investment in their reputation
• Average price of stock is increased by 16%
• In many plausible scenarios it is more
worthwhile to invest in reputation than in
fundamentals
Other variations in model
• Trust in some agents depends on trust in other
agents
– Leads to extreme volatility
• Merged integration functions
– Leads to more realistic medium volatility
So what?
• Regulators ought to be aware of the value of
manipulating trust
• Model implies that agents will overinvest in
the structures of the market itself compared
with investment in productive assets
• Future research directions:
– Multi-good markets
– Inference of beliefs from market prices