ADVANCED MICROECONOMIC

THEORY
 LOTTERIES!

 Have previously considered preferences

over bundles of goods, will now expand
analysis to consider preferences over
lotteries (bundles of possible outcomes). 4 Outcomes:
1 – 10 million
 Begin with a Simple Lottery ‘L’:
2 – 10,000
 A list of ‘N’ outcomes with probabilities 3 – 1, 000
‘pn’.
4–0
 L=(p1,...,pn,...,pN)

 Each pn is Non-negative, the sum of all

probabilities =1.
National Lottery = LN
LN = (0.0001, 0, 0, 0.9999)
 If we have more than one simple lottery
and want to compare, then it is possible,
e.g., comparison of a National Lottery and a Tombola = LT
Tombola (Google it!).
LT = (0, 0.001, 0.001, 0.998)
 The national lottery offers a prize of
10million or nothing; the Tombola offers
10k, 1k or nothing; we can still compare
them!
 LOTTERIES

 Expanding on simple lotteries, we can National Lottery = LN

have situations where we have a lottery LN = (0.0001, 0, 0, 0.9999)
to decide what lottery to participate in!

 This is called a compound lottery:

Tombola = LT
LT = (0, 0.001, 0.001, 0.998)
 ‘K’ simple lotteries, L1,...,LK.
Flip a coin to decide what
 Each simple lottery has a probability lottery to participate in –
of participation of ‘αk’.
50% chance of either:
 Each αk is NON-NEGATIVE and the
sum of all αk = 1. Compound Lottery:

(LN, LT; 0.5, 0.5)
Can now calculate the probabilities of
getting a single outcome (in the
previous, for example, outcome 1
occurred w/ probability 0 if you chose Prob. of getting 1m:
the tombola and 0.0001 if you chose the (0.5x0.0001)+(0.5x0) =
National Lottery; here, it is (0.5 x 0) + 0.00005.
(0.5 x 0.0001).
So on and so forth for others.
 LOTTERIES

 To establish a preference relation between two lotteries (L and L’), we need

to impose three restrictions, although the last is arguably the most
important:
 Rationality, Continuity and Independence:
 Rationality – either one lottery is preferred to the other or both
(indifference). With three lotteries, if L > L’, and L’>L’’, then L> L’’ (see topic
1 slides).
 Continuity – A small change in probabilities does not affect the preference
ordering. This both rules out Lexicographic preferences (where one event
is always preferred or not preferred regardless of the probability
associated to it) and allows for the construction of a utility function
representing the preference relation.
 Independence – L is weakly preferred to L’ (L >L’) if and only if:
 aL + (1-a)L’ > aL’ + (1-a)L’’
 The preference relation is INDEPENDENT of the third lottery used. L’’ is
irrelevant because in a compound lottery, the consumer has either L/L’, or L’’.
Thus his preferences over L and L’ determine which of the two he would have
in a compound lottery.
 LOTTERIES

 We can represent preferences in lotteries using the von Neumann-

Morgenstern (v.N-M) expected utility function.

 The v.N-M expected utility function arises if every simple lottery ‘L’ has
a utility associated with each of its outcomes, such that:
 U(L) = p1.u(1)+...+pN.u(N)

 It is expected utility as the total utility of ‘L’ is simply the expected

value of the utilities of the N outcomes associated with it.

 When does the v.N-M expected utility function occur? I.e., when does a
utility function have an expected utility form that makes it a v.N-M
expected utility function?
 When the utility function is LINEAR – when utility of the expected
value of a compound lottery is the same as the expected utility from
the compound lottery:

 K  K
U    k Lk     kU  Lk 
 k 1  k 1
 LOTTERIES

Proof of Linearity: Linearity  Expected utility Expected utility  Linearity

Need to show that BOTH
Linearity  Ex. U U  L  U  pn Ln  EU Implies that :
 
N
 k Lk   N pnun
Ex. U  Linearity
By linearity : U K
To do so, consider
DEGENERATE LOTTERIES (any
lottery can be written as a U  L    N pnU Ln   pn   K  k pnk
U  L    N pn u n
compound lottery over
degenerate lotteries):
A degenerate lottery Ln occurs   N  K  k pnk un
when outcome ‘n’ occurs with Which is expected utility!
certainty in that lottery.
  N  k  K pnk un
 
Ln = (0,...,1,...,0), Hence, the
overall lottery ‘L’ can be
expressed as the sum of these
The last line derives from   N  kU Lk
lotteries: the fact that U(Ln) is the

L   N pn L n
utility of state ‘n’ because  Linearity!
Ln is a degenerate lottery).
 LOTTERIES

Proof:
Preference relations in v.N-M
expected utility functions are
~
U  K

 k Lk  U  K

 k Lk   A linear
preserved only by LINEAR
transformations:   K

 kU  Lk   
transformation of
U is therefore still
 K  k  U  Lk     has an expected
U and U~ are both v.N-M utility form, as it is
expected utility functions
~
  ~
U K  k Lk  K  kU  Lk  linear.
representing the same

U  L   U  L '
preference ordering iff:
The preference
~
U  L   U  L    U  L   U  L  ordering is
therefore
 0 U  L     U  L    preserved during a
linear
~ ~
U  L   U  L ' transformation
 LOTTERIES

 The EXPECTED UTILITY THEOREM:

 Most important result in choice under uncertainty, according to Mas-

Colell et al. Proof on next slide!

 It states that:
 If individual’s preferences satisfy the continuity (such that they can
be represented by a utility function) and independence axioms, then
his preferences can be represented by a utility function with an
expected utility form.

 Before we were considering the cases where a utility function could

have an expected utility form, now we have a direct relationship
between preferences and the expected utility form.

 In an equation, the Expected Utility Theorem suggests that:

L  L '   N pn u n   N pn ' u n
~
 LOTTERIES

5 Steps to the proof:

Step 1: Need to use the
If L  L' and    0,1, then : Independence Axiom
L L  1    L'  L'
 L ~ L  1    L L  1    L'
Let  ,    0,1, and L be L  1    L' L'1    L' ~ L'
the best lottery and L the worst :
Step 2: Need to use step 1
 L  1    L  L  1    L Iff   

  L  1    L   L  1     L  1    L 
For any L, there is a unique  L where :    
1 ; If   0,   

L ~  L L  1   L  L   L  L  1    L  Step 1
The Utility Function U  L   α L represents 
  L  1     L  1    L   Can simplify
bottom line!

the Preference Relation.   

  L  1    L  1     L  1    L 
U L     L  is linear and
 
  L  1     L  1    L  L  1    L
  L  1    L  L  1    L
 has the Expected Utility Form
Iff   0,   
 LOTTERIES

Step 5: Use step 4

Step 3: Implied by the continuity Need to show :
of preferences. U   L  1    L'  U  L   1   U  L'

Step 4: Use steps 2 + 3 By Step 4 : L ~  L L  1   L  L

L ~ U  L  L  1  U  L   L
L  L'  Apply step 3
L ' ~ U  L '  L  1  U  L '   L
  L L  1   L  L  L ' L  1   L '  L
Apply step 2 :
•Construct a compound lottery of L and L’
 L   L '  U  L   U  L ' •Shown next slide due to complexity.
 LOTTERIES
  L  1    L ' ~
  
 U  L  L  1  U  L   L  1    U  L '  L   1  U  L '   L 
  L  1    L ' ~
 U  L   1   U  L'  L  1   U  L   1   U  L'  L
Remember Step 3!
 L'' ~  L '' L  1   L ''  L
 L ''   U  L   1   U  L'
By the way we construct the utility function (Step 4) :
U L' '   L ''
U  L  1    L'   L ''   U  L   1   U  L' Q.E.D