Thinking and Deciding

CONCLUSION 339
win the $100, if all get the gamble. We can often describe tradeoffs across people as
gambles for individuals.
Another way of stating the third problem is that decisions made according to
one measure can sometimes make everyone worse off according to another measure
(Kaplow and Shavell, 2002). According to the PTO, in the last example, it would
be better to give everyone a gamble with a .7 chance of winning $100 than to give
everyone $50. Yet the $50 is better for each person according to the standard gamble
method (which we could assume to be a measure of individual utility). So, by
following the results of the PTO, we make everyone worse off.
A fourth problem is that the suggested solution does not work. This is a mere
fact, however. The other reasons would apply even if it worked. The solution does
not work because methods of utility assessment are internally inconsistent. The internal
inconsistencies also seem to have a lot to do with the disagreements between
methods.
For example, we have seen (in Chapter 11) that the distortion of utilities in SG
can be explained in terms of the π function of prospect theory. This produced a
certainty effect, which makes the utility of money more sharply declining than it
would otherwise be. The same function can explain internal inconsistencies within
gambles.
The PTO shows another kind of internal inconsistency, ratio inconsistency. Subjects
are unresponsive to changes in the ratios of conditions being compared. For
example, a subject will say that 40 people becoming blind and deaf (BBDD) is just
as bad as 100 people becoming blind (BB). The subject will also say that 20 people
becoming BB is just as bad as 100 people becoming blind in one eye (B). These judgments
would imply that BB is .4 as bad as BBDD, and B is .2 as bad as BB. Thus, B
should be .08 as bad as BBDD. If we ask how many people becoming BBDD is as
bad as 100 people becoming B, we should get an answer of 8. In this kind of task, the
answers are usually much higher, for example, 16. It is impossible to know which
judgment is best from this information alone, but we know the three judgments, 40,
20, and 16, are inconsistent with each other. Consistency could be restored either
by moving the 40 up to 80 or the 16 down to 8, or other ways. The problem is that
the judgments are not far enough apart (Ubel, De Kay, Baron, and Asch, 1996a).
Depending on which conditions are used, this effect could make the PTO disagree
in various directions with the results of other methods. This effect could result from
the tendency to spread out judgments evenly, described on p. 320.
Conclusion
It may seem that utility measurement is beset with so many problems that it is hopeless.
The various methods of measurement are internally inconsistent and inconsistent
with each other. There are two responses to this negative assessment. First,
“compared to what?” Alternative methods of allocating resources, which rely on direct
intuitive judgments about the decisions in question, are very likely even worse