Expected Utility : Chapter 2

William Forbes
Wiley (2009)

1 Expected Utility: Behavioural finance

 Expected utility
The workhorse model of standard finance evaluates
utility by asking how much would an agent/investor
pay to enter a certain lottery. In an uncertain world
alternative courses of actions can be seen as
purchasing tickets in different lotteries.
This suggests the utility of alterative
actions, x, which I denote U(x)) can be evaluated by
an investor’s willingness to pay to enter that lottery
with specified outcomes. For example, how much
are you willing to pay to be allowed to receive a
pound from me if I toss a coin and it lands on heads
2 Expected Utility: Behavioural finance
 Cardinal versus ordinal
outcomes
This schema allows us to build up a cardinal
ordering of anticipated outcomes by expressing a
preference for outcomes as weights in the reference
lottery. This little trick moves us on from considering
choices with lots of different outcomes occurring with
different probabilities to a simple comparison of two
outcomes which occur with differing probabilities.
A cardinal ordering allows us to say not just which
outcome the investor would prefer but also how much
he would be willing to pay to have his preferred
outcome rather than his next best outcome. In
contrast an ordinal ranking can only tell us which of
3 theExpected
outcomes he prefers
Utility: Behavioural finance the first, second, etc,
 Illustration of EU
Assume a floatation price of 30 pence. Suppose the
initial one day price “pop” (or mark-up of the trading
price over the floatation price) can vary over three
increasing levels.
If the initial floatation price is set at 30 pence per-
share suppose that the price at the end of the day
can close at either 30, 60, or 90 pence per-share.
Suppose, further, each outcome is equally likely. If
the price closes the day at 30 pence the investor get
nothing.
 If the price closes the day at 60 pence the investor
get 30
4
pence. Finally, if the price closes the day at 90
Expected Utility: Behavioural finance
pence the investor get a 60 pence pay-off.
 Illustration of EU (cont)
The outcome of this lottery can be expressed as
follows
L=[pi,oi] i=1 to 3 and Σipi=1
which in this example implies
L=[0.33, 0.33, 0.33.0,30,60]
This lottery has an expected value of
0.33x0+0.33x30+0.33x60=0+10+20=30
Note there in no increase in expected wealth as a
consequence of undertaking the investment in the IPO.
The lottery is a “fair bet” in the sense it offers an
expected value equal to its cost.
5 Expected Utility: Behavioural finance
 An investor’s concave (risk-averse)
utility function
Expected Value

0.5

Wealth

0
o 0.5xO O

6 Expected Utility: Behavioural finance

 EU as a normative model
But in some cases the reason why our standard
Finance models seem to do us a disservice is that
the reality of decision- making by investors does
not conform the assumptions under which John von-
Neumann and Oscar Morgenstern formulated the
theory (von Neumann and Morgenstern, 1947).
For these pioneering authors expected utility
theory it constituted a normative model of how a
rational person should make decisions about
alternative courses of action not a
positive/descriptive about how decisions are really
made.
7 Expected Utility: Behavioural finance
 St petersberg paradox
Nicholas asked what is the value of a gamble that
offers two pounds if you toss a coin and it comes up
heads once, or four pounds if it lands on heads
twice in a row, eight pounds if it lands on heads
three times in a row etc? The expected value is
clearly
1/2x2+1/4x4+1/8x8+…….=1+1+1…….=∞

8 Expected Utility: Behavioural finance

 Allais paradox
Gamble A: Receiving a million for sure.
Gamble B: A 10% chance of getting £ 5 million,
A 89% chance of getting £ 1 million,
And 1% chance of getting nothing.
Then choose between
Gamble C: A 11% chance of receiving £ 1 million,
A 89% chance of getting nothing.
Gamble D: A 10% chance of receiving £ 5 million.
A 90% chance of getting nothing.
in the second set of gambles.

9 Expected Utility: Behavioural finance

 Allais paradox 2
Most people choose A rather than B in the first set of
gambles. Later the same group of people mostly appear to
choose D over C. Consider the expected value of the
gambles
 A 1x£1m=£1m
 B 0.1x£5m+0.89x£1m+0.01x£0=£1.39
This suggests the group making the choice are very risk-
averse, leaving £390,000 of expected value unclaimed
rather than face a 1% chance of getting zero. Later the
same group of people make the calculation

 C: 0.11x£1+0.89x0=£0.11
 D: 0.10x£5+0.90x0=£0.50
And choose D. Here the increase in expected utility
seems enough to compensate for the 1% increase in the
Expected Utility: Behavioural finance
10probability of getting nothing
 Allais paradox 3
But note if £1 million for sure is preferred to
0.1x5+0.89x1+0.01x1 we can deduct an 89%
chance of winning a million to in Gamble A
above imply for this group 0.11x1 + 0.89x0 is
preferred to 0.1x£5m+0.90x0. But this is
exactly the reverse of the choice people
actually make in the second choice!

11 Expected Utility: Behavioural finance

 Frames for actions

To address this apparent failing of expected utility

theory Kahneman and Tversky suggest investor’s
evaluate prospects in two consecutive steps
An editing-framing stage when the (gamble or
investment) is initially structured for detailed
consideration,
A later more detailed evaluation of the prospect.

12 Expected Utility: Behavioural finance

 Even Professors make
mistakes
Kahneman and Tversky document similar biases
as being present amongst participants at
conferences they attended as part of their
academic life. They found academics with similar
training (if not prestige) to themselves
Overestimated the power of their tests despite
the fact they often relied on small samples (of 10-
50 in psychological studies)
Displayed undue confidence in early trends in
their data and the stability of the patterns they
observed,
13 Expected Utility: Behavioural finance
 Even Professors make
mistakes
Bore unreasonably high expectations about the
ease of replicating “true” results; i.e. they
underestimated the impact of sample variability on
retrieving statistically significant results,
Found spurious “explanations” of later refutations
of previous results without recognising sample
variation alone could explain the failure to verify
prior results

14 Expected Utility: Behavioural finance