Thinking and Deciding

510 RISK
times more likely than chronic liver disease as a cause of death. So a 5% should not
be a constant number for different causes. Still, subjects almost completely ignored
the distinction between number of lives and percent of the problem.
McDaniels (1988) compared subjects’ judgments to actual government expenditures
for the risks shown in the following table (in abbreviated form). The first
column of numbers shows the mean response of a group of U.S. respondents asked
how much they would be willing to pay from their household budget to reduce each
risk by 20%. The respondents had been told the number of deaths per year shown
in the second column, which were approximately true. The third column in the table
shows McDaniels’s calculation of the total national willingness to pay to prevent one
death from each risk, based on the first column and assuming 90 million households
at the time of the study. The fourth column shows McDaniels’s rough estimates of
what the U.S. was willing to spend to prevent one death from each risk, as revealed
by its government regulations.
Mean U.S. Implied Actual or
WTP in deaths national proposed
Hazard survey per year WTP expenditures
Workplace chemical $7.95 1 $715 mil. > $11 mil.
Explosives $7.68 2 $345 mil. $3 mil.
Aviation $46.07 40 $103 mil. $680,000
Power tools $15.05 80 $17 mil. $430,000
Automobiles $161.30 10,000 $1.3 mil. $95,000
This table shows several things. In the first column, we see that people are willing
to spend more to reduce larger risks. They are willing to spend about $161 to cut
the automobile death rate by 20%, a larger amount than other risks that kill far fewer
people. But we can also see that the adjustment in willingness to pay is not proportional
to the increased size of the risk. If people had a constant value per life saved,
they would be willing to pay 10,000 times as much to cut the automobile death rate
by 20% than to cut the death rate from the particular workplace chemical by 20%.
But the ratio of $161.30 to $7.95 is only about 20, not 10,000. People seem to compromise
between thinking about the proportional reduction in risk and the absolute
reduction. Or perhaps they are simply insensitive to both. In any case, the implied
willingness to pay, in the third column of number, is much higher for smaller risks.
The last column is based on rough estimates, but the differences among the risks
in actual expenditures are large. Most interesting here is the parallel between these
figures and the results of the survey. Although the implied willingness to pay per life
was much greater in the survey, the interesting thing is that the rank ordering is the
same. Both the people and the government seem willing to spend more to reduce
smaller risks. These results suggest that the government is responsive to what the
people want. And what the people want is inefficient. If the government wanted to
save more lives for less money, it could change the regulations so that less money
was spent on small risks and more money on larger risks. More generally, we can