Policy decisions made today affect people in the future, who do not participate in making the decisions. Most policy decisions are made at the national level, so they also affect can people in other nations, who also do not participate. In this way, national decisions and decisions about the future are similar. They involve externalities, effects on nonparticipants. Examples of policies that affect future people are those that affect future biodiversity, such as forest management, trade (in wood), population control, zoning, and so on. Examples of policies that affect other countries are investments in research (e.g., on tropical diseases or agriculture), trade, debt relief, and foreign aid. The general issue addressed here is how we should make decisions of this sort.
Decisions that affect the future have been debated in terms of cost-benefit analysis.1 When cost-benefit analysis compares present costs with later costs, the latter must be discounted. A million dollars today will be worth quite a bit more in 25 years, if we invest it - or it will cost us more if we have to borrow it and pay interest. It is natural to discount benefits as well, or else, under reasonable assumptions, we could find ourselves postponing all beneficial projects while we wait for our stock of money to grow ever larger.2 In particular, this argument assumes that we can continue to trade off money and other benefits at roughly the same rate, i.e., that the benefits are fungible.
How much should we discount the future in cost-benefit analysis? The use of current interest rates seems to discount the future too much, and use of zero discounting implies that we ought to be spending huge amounts of money to prevent global warming or the loss of biodiversity.
It may help to see this problem as an example of a more general sort of tradeoff, that between us, the decision makers, and others. Both kind of policies - those that affect people in other nations and those that affect future people - involve such tradeoffs. If we were completely selfish, we would ignore these effects and make decisions for ourselves only. If we were perfect utilitarians, who believe that everyone counts the same, we would maximize utility for everyone involved. To judge from our actual behavior, we are somewhere between between perfectly selfish and perfectly utilitarian. We give others less weight then ourselves, but usually not zero weight. (When competition is involved, the weight may be negative, however.)
Even perfect utilitarians have some reason to discount effects on people in the future. First, the future is uncertain. As events get farther away in the future, we lose the ability to predict them accurately. A reasonable way to approximate this loss is to assume that it results from uncertain events, such as technological changes, which could occur with equal probability at any time. Under this assumption, the expected utility of a future event declines exponentially as a function of time.
Second, marginal utility of goods is declining, and this affects the optimal distribution of material goods. People in poor countries now derive greater utility from a given amount of goods than people in rich countries. People in the future are likely to be richer than people today, so they are likely to derive less benefit from a given additional amount of goods.3
Beyond these considerations, though, people who make policy - and that includes all of us in our roles as citizens - may weigh their own utilities more heavily than the utilities of others. The discounting of others' utilities may fall off as a function of distance. We may weigh family members more than others, and we may those who belong to the same nation, race, or ethnic group more than those who do not. Likewise, we may weigh our great grandchildren's generation more heavily then the generation of their great grandchildren.
Pure utilitarians do not find such differential weighing to be justified. But, given that it is inevitable for us, as decision makers, it is typically better for others if we do it consistently according to the effects on them. That is, we should look for the greatest benefit to others in the same category, given a fixed sacrifice of our own interest, or we should look for the smallest sacrifice required for a given benefit. Suppose two programs that will save lives of people who live 100 years from now, one at the cost of $1 million per life and one at the cost of $2 million. We should put all our money into the first and none into the second, up to whatever amount we are willing to spend. Yet, if we weight people inconsistently, we will sometimes put money into less effective programs, thus hurting ourselves, others, or both.
Somewhat in this spirit, some authors have suggested that the discounting of long-term future effects should be based on values elicited from current citizens, those who must make the sacrifice in question for the benefit of others.4 Surveys of citizens' values might be valuable to governments that want to satisfy public values without insisting that citizens get involved in all the nitty-gritty details of policy. At least, if we had a consistent number to indicate how much we should weigh outcomes as a function of their temporal distance, we would avoid the losses that result from inconsistent allocation to projects that have their effects at the same time.
One trouble is that every attempt to measure discount rates by asking people questions, or even by observing individual decisions, has found dynamic inconsistency. Specifically, the rate of discounting declines as the effects are farther away in the future from the time the decision is made. This means that people would make different choices concerning the same outcomes, as function of when the decision is made.
From a utilitarian perspective, this makes no sense, so long as outcomes can be assigned to specific times. The goodness, or utility, of an outcome must depend on the extent to which it achieves people's goals as expressed in their fundamental objectives, the criteria by which they evaluate states of affairs. So utility cannot depend simply on when a decision is made. (Note that some utilities cannot be assigned to specific times in the way required: The utility of hearing a piece of music with a pleasing form cannot be assigned to a point in time. But this sort of example is largely irrelevant to policy decisions with effects over decades and centuries.)
Consider a simple case. It is 2:59 P.M. A child has a choice of one piece of candy at 3:00 or two pieces at 4:00. Many children will take the first option. Suppose that such a child is given the same choice at 2:00 instead of 2:59. That is, the choice is still between one piece at 3:00 or two pieces at 4:00. It is easy to imagine that the child will choose the second option. In this case, the child has made different choices as a function of the time alone. (Let us put aside the child's emotions of frustration and anticipation.) This pattern of choices is dynamically inconsistent, that is, inconsistent over time. This is nonnormative, because the time at which the decision is made does not affect the extent to which the options achieve the child's goals. If the child's main goal here is to get as much candy as possible, for example, then the second option should be chosen in both cases. Dynamic inconsistency violates the principle of delay independence.
To avoid dynamic inconsistenty, discounting must be exponential. Researchers have repeatedly found that discounting is more like a hyperbola than an exponential function, in both animals and people.5-10 Discounting is inconsistent in other ways.10-12 The discount rate is higher when rewards are smaller, and the discount rate for losses is lower than that for gains. These results are found both with hypothetical questions and with experiments in which decisions lead to real rewards (money for people, food pellets for hungry pigeons).
Hyperbolic discounting might reflect a tendency to see events closer in time as more distinct from each other than events farther in the future. This is a kind of perceptual distortion. If such distortion operates when people consider policies that affect the long-range future, then their evaluation of these policies need not reflect their attitude toward the utility or well-being of those affected. It certainly need not reflect the utilities of those affected, but I have already acknowledged that we must live with a certain amount of partiality toward self, family, and cohort, if not co-national and co-race.
On the other hand, hyperbolic discounting could reflect our true opinion about how we want to treat others. We could, for example, care very much about those close to us in time, and less about others after that, but nearly the same for others after that no matter when they live. This would lead to policies that future people, with different utilities, would want to overturn. They would make a larger distinction than we made between those directly after them and those in the more distant future. Of course, people in the future will have all sorts of reasons for wanting to overturn our decisions.
In sum, the question addressed here is whether people's judgments about how they want to weigh the utilities of others can be taken at face value. If these judgments are the results of distortions, then they could be internally inconsistent, hence useless. The problem is not so much that future generations will want to undo the decisions we make today; that is practically inevitable. Rather, it may turn out that our own judgments depend on how we are asked about them, on what perspective we are asked to take.
To demonstrate how judgments can be inconsistent, hence invalid as measures of our utility for outcomes, I conducted three experiments. I report only the last, since the results of the first two were the same, and the last is more complete in the variables it examines. I asked subjects about programs that saved lives or species, now or in the future. I asked:
I also asked the same questions multiplying the years by two, so that it was 50 and 100 instead of 25 and 50. I shall call this manipulation ``scale.''
These questions allow two consistency checks. First, the 0 vs. 50 years (for example) should be inferable from the 0 vs. 25 and 25 vs. 50. In particular, if we look at the ratio of the response to 100,000 and call that the compensating ratio, then (assuming that utility is a power function of the number of units) the compensating ratio for 0 vs. 50 should be the product of the other two. This is a test of delay independence. Unlike previous studies, though, the items were next to each other and subjects could try to avoid the effect.
The second consistency check involves the question about what the judgment would be if the decision were made in 25 years (or 50 years when this was the step size). This should be the same as the compensating ratio for 25 vs. 50 years. The hypothesis is that it is smaller.
The manipulation of scale also tests another hypothesis. When we ask people about discounting the future for social programs, we need to consider two different kinds of effects, those that occur within the lifetime of individuals and those that affect different people who exist at different times. We might think that it is possible to discount at different rates. We might, for example, care equally about people who exist 50 and 100 years from now. But we might care more about the immediate present than 25 years hence in people who are alive at both times. This would lead to a high discount rate for the next 25 years and a zero rate once everyone alive has died. Although such an argument runs into many technical problems13, people may think this way, and this may affect their judgments in tasks that involve discounting. They will have a higher discount rate for smaller scale.
I also manipulated three other variables. The first was whether the effects of the program were in poor countries or rich countries. The hypothesis is that people in rich countries would see the decisions involved as between ``helping ourselves now'' and ``helping others.'' If so, the ``others'' would be discounted largely because they are not ``us,'' and, once this discounting occurred, further discounting for delay into the future would not occur. In this case, the discount rate for helping people in poor countries would be lower than that for helping people in rich countries.
The second was lives vs. species. Although I had no particular hypothesis about whether the discount rate would be higher for one or the other, people might think of species as more of a world resource, so that where the species existed (poor vs. non-poor) would not matter.
The third was fungibility, the possibility of saving money now and spending it later. As I noted earlier, one of the arguments for discounting benefits as well as costs is that, if we save money now, we can invest it and use the proceeds later on similar projects. If we value future benefits as much as present benefits, and if we can usefully spend the proceeds from investment in the future, we will always do better by investing money and spending it in the future than by spending it now (and this will hold for the future too). If we must spend the money now, though, we are free to consider future benefits and current benefits as equal. We might thus discount future benefits at a lower rate when expenditures are not fungible. In each case, I either reminded subjects of the possibility of delaying expenditures or explicitly ruled it out. (Previous research has found that the correlation of discount rates for health and money increases when health and money are fungible in this way.14)
Sixty subjects completed a questionnaire on the World Wide Web, for $3 each. (To receive the $3, the subject had to provide a name, address, email address, and Social Security Number if in the U.S.) The questionnaire was written in JavaScript, which randomized the order of 16 conditions separately for each subject and which carried out various checks after each answer, to insure that the subjects were answering with sufficient care. Four subjects were eliminated for answers that seemed to reflect misunderstanding (and which were missed by the checks), leaving 56.
The introduction read (with some minor editing):
This is about two problems, species and health. The species
problem is that human activity has increased the rate at which
species of plants, animals (including insects) become extinct.
Some estimates say that as many as 500,000 species disappear each
year (out of the world's 100,000,000 species) (Ehrlich and
Wilson, Science magazine, Aug. 16, 1991). This results
from pollution, fishing, and using land for agriculture, roads,
buildings, etc.
The health problem is that about 25,000,000 people around the
world die young each year from health conditions that can be
prevented or cured (out of 6,000,000,000 people and 50,000,000
total deaths). These include respiratory infections, diarrhea,
lung cancer, malaria, accidents, and AIDS (Murray and Lopez,
Science, Nov. 1, 1996).
We, that is, people throughout the world, might spend money to
slow down the loss of species or prevent early deaths. This
study is about how we evaluate programs that do these things. In
all cases, imagine that the programs are paid for by
contributions from many nations based on ability to pay. Nations
classified as ``poor'' pay nothing. These are,
according to the U.N.: Afghanistan, ... [countries were
listed], and Zambia. Here I call other nations
``non-poor''.
The programs are efficiently run, and as effective as they are
claimed to be. The life-saving programs save, on the average, 20
years of life for each person ``saved,'' and the people saved
range from children to older people.
Some of the programs affect only the poor nations, and others
affect only the non-poor nations. For example, some health
problems, like cholera, occur mostly in poor nations, and others,
like automobile crashes, occur mostly in non-poor nations. And
each nation has different species of plants and animals. Please
assume that the poor nations will continue to be poor in the
future - just as poor as they are now.
For each choice, you get two options, for example:
Option A: Save 100,000 species now.
You have to enter a number that makes options A and B equally
attractive. If the number of species in B were higher than the
number you enter, you should prefer B. If the number of species
in B were lower than what you enter, you should prefer A. You
should not write LESS than 100. If it is better to save species
now than later, you would have to save more later to make up for
it. (But 100 is OK.)
The items also differ in when the species or lives are saved.
Programs of this sort can have delayed effects. It may seem
unrealistic to say that some program will do something in exactly
50 years. But it is not unrealistic to think that some part of a
program will do that. And the purpose of this study is to see
how you think about timing of these benefits. So try to imagine
that the statements are true.
The cost of the programs is fixed at a certain amount of money
per life or per species.
The items differ in whether money saved [by not spending it now]
can be invested for later use in the same kind of program. When
money can be invested for future use, the interest rate on the
money is 5% in real interest, and the cost of the future program
is the same in current dollars. This means that you can suppose
there is no inflation. When money cannot be invested, it is
returned to the taxpayers in the form of reduced taxes.
In some items, you put yourself in the position of someone
making the decision in the future.
There are 16 screens, each with four items. The screens may
look alike, but they are all different. So please pay attention
to the differences. (Some differences may not matter to you, but
pay attention anyway.)
This decision affects only poor [non-poor] nations.
Option A: Save 100,000 lives [species] now
Option A: Save 100,000 lives [species] then
All responses were converted to logarithms before analysis; hence
all means are based on geometric means, both those computed
within subjects and those computed across subjects. All
statistical tests were based on these geometric means, and back
transformation was done only to enter numbers in Table 1.
Species vs. lives, poor vs. non-poor, and fungible vs.
non-fungible had no significant effects on discount rate and did
not interact with each other or with scale (25 vs. 50 year
steps). Further analysis collapsed across these variables.
Table 1 shows the results, collapsing over the variables that had
no effect. Several results are apparent:
Sensitivity to scale.
Subjects were somewhat sensitive to scale. The ratios in the
bottom half of the table, where the step size was 50, are larger
than those in the top half (t54 = 6.72, p = .0000). But the
adjustment is insufficient, because the implied discount rates
are lower when the step size is larger (t54 = 7.62,
p = .0000). The clearest comparison is that between the 0-50
period when it is the longest interval on the screen (second row
of table) and when it is half of the longest (fifth row). The
ratio is larger when it is the longest (t54 = 4.11,
p = .0001). In sum, subjects seem to evaluate time intervals by
comparison to what is immediately available (in this case, on the
same screen). This result is consistent with this
psychological-distortion explanation, rather than the idea that
people prefer the current and next generation and treat other
generations nearly alike.
Discount rate and delay.
As found before by many others, discount rates (as percent
interest) are not constant as a function of delay. The discount
rates for the longer intervals are all significantly lower than
those for shorter intervals. However, this result is not a
function of a general judgment that the discount rate should
decline in the future. In particular, the discount ratio for
0-25 is the same as (not significantly different from) that for
25-50 (based on the choice between spending money in 25 years or
50 years), and likewise for 0-50 vs. 50-100.
The rate estimated from a decision made at a future time,
however, is higher than same decision made ``now'' (in an
analysis of variance that included both scale sizes,
F1,54 = 7.58, p = .0080; the interaction was also significant,
the effect being larger for the larger scale, F1,54 = 4.75,
p = .0337). Subjects seemed to think that they would discount
more if they were transported into the future, but they did not
in fact discount more for time intervals beginning in the future
than for intervals beginning now, when they made the decision
``now.''
Consistency of discounting.
Discounting from 25 to 50 years should be the ratio of
discounting from 0 to 25 years to that from 0 to 50 years. I
tested this by looking at the difference of the predicted vs.
obtained log compensation ratios. The ratio for 25-50 predicted
from 0-25 and 0-50 was 1.64, which was significantly lower than
the obtained 25-50 ratio of 2.31 (t55 = 5.05, p = .0000).
Likewise, the ratio for 50-100 predicted from 0-50 and 0-100
was 1.73, which was lower than the obtained 50-100 ratio of 3.07
(t55 = 6.42, p = .0000). This effect is easily explained in
terms of insufficient sensitivity to delay. The difference
between a 25-year interval (0-25) and a 50-year interval (0-50)
is too small.
These results suggest that the finding of declining discount rates
in questionnaire studies may be an artifact of judgment. As found
in many other studies15, people are insensitive to
quantity. Here, they are insensitive to the amount of delay.
They distinguish between present and future, but they are
insufficiently sensitive to the amount of future delay. They are
also insensitive to scale when this is manipulated across items.
They seem to make judgments primarily by comparing the cases they
see. These effects lead to inconsistent judgments.
Moreover, in this study, subjects were not sensitive to
fungibility, even though their implicit discount rates were
usually below the 5% interest rate described.
Nor did they have different discount rates for ``us'' and
''them.'' In this study, at least, the preference for the
immediate present applies even though the effects are entirely on
other people, strangers in foreign countries. Such a judgment
may be challenged. If we do not know the people our programs
effect, why should we care when they live?
The general conclusion is that these judgments cannot be taken to
be judgments about the differential weight of consequences that
occur at different times.
Such inconsistency might also affect judgments about the
allocation of resources among people existing now. When people
think about allocation, they apply various heuristics, such as
equality or deservingness. These heuristics can lead to
inconsistency. For example, when given a fixed budget for cancer
screening, people would rather offer a less effective screening
test to 100% of a population, thereby saving 1,000 lives, than a
more effective test to 50% of the population, thereby saving
1,100. However, when both percentages are reduced by a constant
proportion or a constant amount, the number of people preferring
the less effective screening test was reduced.16 This effect
can be induced as a framing effect, simply by doubling the size
of the ``population'' at issue, which makes 100% vs. 50% into
50% vs. 25%.17 Even though the people affected are the
same, the judgment changes. When it is impossible to treat
``everyone'' equally by giving them the test, people are no
longer willing to sacrifice lives for the sake of equality. Yet
``everyone'' depends on which group is being discussed.
Of course, the analogy between this sort of result and the kind
of inconsistencies shown in the present experiment is loose.
When we make judgments about the future, we tend to think in
terms of distance, with no sharp boundary. When we think about
fairness to groups, though, we tend to have sharper boundaries,
such as that between co-nationals and foreigners. What is common
to the two cases, though, is that the judgments we make from
different perspectives disagree.
If we concerned ourselves only with outcomes for individuals,
ignoring when or where they live (or their race or gender, for
that matter), we would make more consistent judgments. We could
still do this and weigh ourselves and our families more than
others.
Kopp and Portney4 recently proposed that intergenerational
decision making be done through a mock referendum, in which a
random sample of citizens were given all the information
available about the costs and benefits of alternative policies,
such as those concerning global warming. Then the respondents
would vote on various proposals, such as carbon taxes of various
amounts. One argument for this proposal was that the issues were
so complex and uncertain as to make cost-benefit analysis, with a
fixed discount rate, less useful as a tool. The second main
argument was that, if the people in such a referendum rejected
some proposal, under such maximum-information conditions, it
would not pass the legislature anyway.
One difficulty with this proposal is that it does not really
solve the problems of complexity and uncertainty. It just pushes
off this problem onto the respondents, as if they were some sort
of black box that could deal with complexity better than any
policy analyst. Some respondents however get frustrated wading
through lots of detailed information when what they really want
is the results of a thorough policy analysis, even just the best
guess, as inaccurate as it may be. For example, if I am asked my
opinion of the Kyoto protocol, I do not want to read it, plus all
the supporting technical documentation. Even less do I want to
read a watered-down version of it meant for public consumption.
What I would like is the summary opinion of experts about whether
the benefits exceed the costs. The referendum idea does not
circumvent the need for policy analysis, for those respondents
who might care about such analysis.
Another argument for a referendum is that any democratic
government that ignores the final expression of public
preferences, whatever their basis, will not stay in power for
long. But this is surely an exaggeration. The public's
preferences may be transitory.18 Or citizens may come to
recognize that they lack sufficient expertise to make the final
decisions. The U.S. Federal Reserve, for example, does not do
surveys to decide what interest rates and inflation rates people
would like. People can, up to a point, understand their own
government ought to ignore their immediate preferences.
Breyer19 argues that this can happen even in the
controversial area of risk regulation, although his proposals
have not attracted much interest. An enlightened government could
try to maximize true utility, and explain what it is
doing, so that it can keep public acceptance. This is fully
consistent with democracy.
What should government do, and what should citizens expect of it?
The general type of decision at issue is one in which the current
citizens must reduce their current well-being for the sake of
other people, either people in the future (within the same nation
or not) or people in other nations.
One thing it can do is try to improve the measurement of public
preferences. One strategy for doing this is to present
respondents with their own inconsistencies and then ask the
respondents to resolve the inconsistencies. People seem not to
object to this procedure, and, when it is done, inconsistencies
are reduced beyond the ones that are resolved.20,21
In principle, such improved methods of value measurement could
determine how much altruism people have, that is, how much they
are willing to weigh the interests of non-nationals and future
people, relative to their own. Extensive application of such
methods might involve a kind of education in which people are
challenged not only to be consistent over time but also in their
treatment of future nationals and non-nationals. (No such
inconsistency was observed here, but it might be observed
elsewhere.) Indeed, in the limit, people might be challenged on
their preference for current co-nationals (other than themselves
and their families) over others, a preference that surely exists
yet is arguably as arbitrary is preference for one's own race or
sex.22,23
In the meantime, however, while we are waiting for value
measurement, and people's values, to become more consistent and
less arbitrary, government might at least try - with the
consent and knowledge of citizens - to bring some internal
consistency into its policies that affect non-nationals and
future people. We need not assign a monetary value on lives in
order to do this. We can, instead, apply simple
cost-effectiveness analysis. We can look for cases in which we
are now spending a lot of money for little benefit to others, and
other cases where we could spend much less money and do more
good, and we could transfer money from the former to the latter.
Nobody can make these judgments except experts who can figure the
costs and benefits. If government would adopt the policy of
relying on such experts, it might be able to earn the trust of
citizens, just it has done in the matter of setting interest
rates.19
2. Keeler, E., & Cretin, S. (1983). Discounting of life-saving
and other non-monetary effects. Management Science, 29,
300-306.
3. Schelling, T. C. (1996). Global decisions for the very long
term: Intergenerational and international discounting. In
R. J. Zeckhauser, R. L. Keeney, & J. K. Sebenius (Eds.),
Wise choices: Decisions, games, and negotiations, pp. 152-166.
Boston: Harvard Business School Press.
4. Kopp, R. J., & Portney, P. R. (1999). Mock referenda for
intergenerational decisionmaking. In P. R. Portney and J. P.
Weyant (Eds.), Discounting and intergenerational equity
(pp. 87-98). Washington: Resources for the Future.
5. Ainslie, G., & Haendel, V. (1983). The motives of the
will. In E. Gottheil, K. A. Druley, T. E. Skoloda, & H. M.
Waxman (Eds.), Etiologic aspects of alcohol and drug
abuse. Springfield, IL: Thomas.
6. Chung, S. H., & Herrnstein, R. J. (1967). Choice and delay of
reinforcement. Journal of the Experimental Analysis of
Behavior, 10, 67-74.
7. Solnick, J. V., Kanneberg, C. H., Eckerman, D. A., & Waller,
M. B. (1980). An experimental analysis of impulsivity and impulse
control in humans. Learning and Motivation, 11, 61-77.
8. Cropper, M. L., Aydede, S. K., & Portney, P. R. (1992).
Rates of time preference for saving lives. American
Economic Review, 82, 469-472.
9. Atherton, E., & French, S. (1999). Valuing the future: an
application of hyperbolic discounting with time weights.
Risk Decision and Policy, 4, 17-29.
10. Kirby, K. N. (1997). Bidding on the future: Evidence against
normative discounting of delayed rewards. Journal of
Experimental Psychology: General,126, 54-70.
11. Chapman, G. C. (1998). Sooner or later: The psychology of
intertemporal choice. In The Psychology of Learning and
Motivation, Vol. 38, pp. 83-113. New York: Academic Press.
12. Loewenstein, G., & Prelec, D. (1993). Preferences for
sequences of outcomes. Psychological Review, 100, 91-108.
13. Portney, P. R. & Weyant, J. P. (Eds.) (1999).
Discounting and intergenerational equity. Washington:
Resources for the Future.
14. Chapman, G. C. (1998). Sooner or later: The psychology of
intertemporal choice. In The Psychology of Learning and
Motivation, Vol. 38, pp. 83-113. New York: Academic Press.
15. Baron, J. (1997). Biases in the quantitative measurement of
values for public decisions. Psychological Bulletin, 122,
72-88.
16. Ubel, P. A., Baron, J., Nash, B., & Asch, D. A. (2000).
Are preferences for equity over efficiency in health care
allocation 'all or nothing'? Medical Care, 38,
366-373.
17. Ubel, P. A., Baron, J., & Asch, D. A. (2000). Preference
for equity as a framing effect. Unpublished manuscript,
University of Pennsylvania.
18. Loewenstein, G., & Mather, J. (1990). Dynamic processes in
risk perception. Journal of Risk and Uncertainty, 3,
155-175.
19. Breyer, S. (1993). Breaking the vicious circle: Toward
effective risk regulation. Cambridge, MA: Harvard University
Press.
20. Baron, J., Wu, Z., Brennan, D. J., Weeks C., & Ubel, P. A.,
(in press). Analog scale, ratio judgment and person trade-off as
utility measures: biases and their correction. Journal of
Behavioral Decision Making.
21. Baron, J. (in press). Measuring value tradeoffs: problems and
some solutions. In E. U. Weber, J. Baron, & G. Loomes (Eds.)
Conflict and tradeoffs in decision making: Essays in
honor of Jane Beattie. New York: Cambridge University Press.
22. Baron, J. (1996). Do no harm. In D. M. Messick & A. E.
Tenbrunsel (Eds.), Codes of conduct: Behavioral research
into business ethics, pp. 197-213. New York: Russell Sage
Foundation.
23. Baron, J. (1998). Judgment misguided: Intuition and
error in public decision making. New York: Oxford University
Press.
Table 1.
Species and lives
Option B: Save ,000 species 25 years from now.
Then each screen was of the following form (with alternatives in
brackets):
Any money you save by spending less can [cannot] be
invested at 5% interest and used later to save
lives [species] for the same cost [is returned to the
taxpayers and cannot be used later to save
lives [species].
The 16 conditions, presented in a random order, were all
combination of fungible vs. non-fungible (the top item on the
screen), poor vs. non-poor, species vs. lives, and 25 vs. 50
year steps. For the 25 year steps, the four items on each screen
were now vs. 25, now vs. 50, 25 vs. 50, and 25 vs. 50 but
from the perspective of someone in 25 years. For the 50 year
steps, all the delays were double.
Option A: Save 100,000 lives [species] now
Option B: Save ,000 lives [species]
25 [50] years from now
Option B: Save ,000 lives [species]
50 [100] years from now
Option A: Save 100,000 lives [species] 25 [50] years from
now
Option B: Save ,000 lives [species]
50 [100] years from now
Now imagine making this decision in 25 [50] years:
Option B: Save ,000 lives [species]
25 [50] years from then
Results
Implications
Fairness
Democracy
Conclusion
References
1. Portney, P. R. & Weyant, J. P. (Eds.) (1999).
Discounting and intergenerational equity. Washington:
Resources for the Future.
Mean compensating ratio (based on geometric means) as a function
of time period and time of decision. (``At 25'' means the decision
is made at 25 years.) The right column shows the annual interest
inferred from the ratio.
Time period Compensating Annual % in years ratio interest 0-25 2.30 3.40 0-50 3.78 2.70 25-50 2.31 3.40 25-50 at 25 2.50 3.74 0-50 3.21 2.36 0-100 5.54 1.73 50-100 3.07 2.27 50-100 at 50 3.51 2.54
1Supported by NSF Grant SES98-76469. Send correspondence to Jonathan Baron, Department of Psychology, University of Pennsylvania, 3815 Walnut St., Philadelphia, PA 19104-6196, or (e-mail) baron@psych.upenn.edu.