Rubinstein2005-page118

Rubinstein2005-page118 - inFnite space Z requires...

This preview shows page 1. Sign up to view the full content.

October 21, 2005 12:18 master Sheet number 116 Page number 100 LECTURE 9 Risk Aversion Lotteries with Monetary Prizes We proceed to a discussion of a decision maker satisfying vNM as- sumptions where the space of prizes Z is a set of real numbers and a Z is interpreted as “receiving \$ a .” Note that in Lecture 8 we as- sumed the set Z is Fnite; here, in contrast, we apply the expected utility approach to a set that is inFnite. ±or simplicity we will still only consider lotteries with Fnite support. In other words, in this lecture, a lottery p is a real function on Z such that p ( z ) 0 for all z Z , and there is a Fnite set Y such that z Y p ( z ) = 1. We will follow our general methodology and make special assump- tions that Ft the interpretation of the members of Z as sums of money. Let [ x ] be the lottery that yields the prize x with certainty. We will say that % satisFes monotonicity if a > b implies [ a ]Â[ b ] . Thus, if u is a vNM utility function representing a monotonic pref- erence relation, then u is a strictly increasing function. An axiomatization (not presented here) of vNM preferences on an
This is the end of the preview. Sign up to access the rest of the document.

Unformatted text preview: inFnite space Z requires strengthening of the continuity assump-tion so that if p Â q , then small changes in the prizes, and not just in probabilities, leave the preferences unchanged. ±rom here on we focus the discussion on preference relations over the space of lotter-ies for which there is a continuous function u (referred to as a vNM utility function), such that the preference relation is represented by the function Eu ( p ) = ∑ z ∈ Z p ( z ) u ( z ) . This function assigns to the lot-tery p the expectation of the random variable that receives the value u ( x ) with a probability p ( x ) . The following argument, called the St. Petersburg Paradox , is some-times presented as a justiFcation for assuming that vNM utility func-tions are bounded. Assume that a decision maker has an unbounded vNM utility function u . Consider playing the following “trick” on him:...
View Full Document

{[ snackBarMessage ]}

Ask a homework question - tutors are online