1 Mixture Models of Mixture Models of Choice under Risk Choice under Risk Peter G. MOFFATT Peter G. MOFFATT University of East Anglia, University of East Anglia, Norwich, United Kingdom. Norwich, United Kingdom. (with Anna CONTE and John D. HEY, both LUISS, Rome)
Mixture Models of Choice under Risk. Peter G. MOFFATT University of East Anglia, Norwich, United Kingdom. (with Anna CONTE and John D. HEY, both LUISS, Rome). Estimating preference functionals for choice under risk from the choice behaviour of individuals. Data: Hey (2001, Exp. Ec. ). - PowerPoint PPT Presentation
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
11
Mixture Models of Mixture Models of Choice under RiskChoice under Risk
Peter G. MOFFATTPeter G. MOFFATTUniversity of East Anglia, University of East Anglia, Norwich, United Kingdom.Norwich, United Kingdom.
(with Anna CONTE and John D. HEY, both LUISS, Rome)
22
Estimating preference Estimating preference functionals for choice under functionals for choice under
risk from the choice behaviour risk from the choice behaviour of individuals.of individuals.
Data: Hey (2001, Data: Hey (2001, Exp. Ec.Exp. Ec.).). 53 subjects53 subjects a sequence of 500 tasksa sequence of 500 tasks spread over 5 daysspread over 5 days they choose between risky they choose between risky
prospects of the following kind:prospects of the following kind:
33
€ 100
Random Lottery Incentive system used (one chosen lottery selected at random and played for real).
Lottery Lottery pp Lottery Lottery qq
€ 0
€50
€ 100
€50
€ 0
44
All 500 problems involved All 500 problems involved three three of of the four outcomes £0, £50, £100 the four outcomes £0, £50, £100
and £150.and £150. Let us denote them by Let us denote them by
xxii ( (i= 1, 2, 3, 4i= 1, 2, 3, 4) )
and the corresponding utility values byand the corresponding utility values by
uuii ( (i = 1, 2, 3, 4i = 1, 2, 3, 4)).. The probabilities of the four outcomes The probabilities of the four outcomes
in these two lotteries in pairwise-in these two lotteries in pairwise-choice problem choice problem tt ( (t = 1, … , 500t = 1, … , 500))
are pare p1t1t p p2t2t p p3t3t p p4t4t
and and qq1t1t q q2t2t q q3t3t q q4t4t
55
Subjects Subjects compute:compute:
where the where the P’s P’s and and Q’s Q’s are not are not necessarily the correct probabilities, necessarily the correct probabilities, but they are derived from the correct but they are derived from the correct probabilities in the following manner:probabilities in the following manner:
1 1 2 2 3 3 4 4t t t tP u P u P u P u
1 1 2 2 3 3 4 4t t t tQ u Q u Q u Q u
(p-lottery)
(q-lottery)
66
Here the function Here the function w(.) w(.) is a is a probability weighting function.probability weighting function.
1 2 3 4
2 2 3 4 3 4
3 3 4 4
4 4
1t t t t
t t t t t t
t t t t
t t
P w p p p
P w p p p w p p
P w p p w p
P w p
1 2 3 4
2 2 3 4 3 4
3 3 4 4
4 4
1t t t t
t t t t t t
t t t t
t t
Q w q q q
Q w q q q w q q
Q w q q w q
Q w q
77
THE MIXTURE ASSUMPTIONTHE MIXTURE ASSUMPTION EITHER the subject is EU ( ), EITHER the subject is EU ( ), OR the subject is Rank Dependent EU (RDEU).OR the subject is Rank Dependent EU (RDEU).
RDEU: Two versions:RDEU: Two versions:– Power weighting function: Power weighting function:
0w p p – Quiggin weighting function: Quiggin weighting function:
1/ 0.2791
pw p
p p
w p p
88
Subjects choose the lottery Subjects choose the lottery pp ( (qq) if and ) if and
only ifonly if
where where
1 1 2 2 3 3 4 4 0t t t t tD u D u D u D u
1,2,3,4jt jt jtD P Q j
here here t is a is a FechnerianFechnerian error term, with error term, with
represents extent ofrepresents extent of 2~ 0,t N . .
computational error by subjects.computational error by subjects.
99
Assume CRRA utility:Assume CRRA utility:
r r represents risk attitude. represents risk attitude.
Distribution of RD Distribution of RD parameterparameter
1919
95% bounds for weighting 95% bounds for weighting function (Quiggin)function (Quiggin)
2020
Summary (1)Summary (1)
We have taken into account We have taken into account heterogeneity in behaviour between heterogeneity in behaviour between individuals and within individuals:individuals and within individuals:
2121
by ‘heterogeneity between by ‘heterogeneity between individuals’ we mean that people are individuals’ we mean that people are different, not only in terms of which different, not only in terms of which type of preference functional that type of preference functional that they have, but also in terms of their they have, but also in terms of their parameters for these functionals.parameters for these functionals.
This means that trying to estimate a This means that trying to estimate a ‘representative agent’ preference ‘representative agent’ preference functional to represent the functional to represent the preference functional of all the preference functional of all the individuals may well lead to biased individuals may well lead to biased estimates.estimates.
Summary (2)Summary (2)
2222
by ‘heterogeneity within by ‘heterogeneity within individuals’ we mean that individuals’ we mean that behaviour can differ for a given behaviour can differ for a given individual solving the same individual solving the same choice problem.choice problem.
Summary (3)Summary (3)
2323
solutions to both these problems, solutions to both these problems, concentrating particularly on using a concentrating particularly on using a mixture model and introducing mixture model and introducing unobserved heterogeneity terms to unobserved heterogeneity terms to capture the heterogeneity of capture the heterogeneity of preference functionals across preference functionals across individuals.individuals.
an econometric approach that deals an econometric approach that deals with technical difficulties associated with technical difficulties associated with these issues.with these issues.
We have We have proposed:proposed:
2424
80% of the population are RDEU 80% of the population are RDEU with their own (Quiggin) with their own (Quiggin) weighting function; 20% are EU.weighting function; 20% are EU.
Mixture model tells us (with high Mixture model tells us (with high probability) which subjects are of probability) which subjects are of which type.which type.
Results:Results:
2525
Mixture Models of Mixture Models of Choice under RiskChoice under Risk
Peter MOFFATTPeter MOFFATT
University of East Anglia, University of East Anglia, NorwichNorwich(with Anna CONTE and John D. HEY, both LUISS, Rome)