From: hal@finney.org
Date: Sun Oct 29 2000 - 21:55:23 MST
[I ask that these comments not be forwarded elsewhere as I have not had
a chance to develop them in detail. They are presented here for feedback
from Extropian list members.]
> Given that individuals discount the future, invariance requires that
> agents weight the past more heavily than they do the present. In
> other words, it requires that a thirty year old cares more about an
> apple that was consumed at age three than about an apple to be
> consumed today. moreover, the importance of that apple at age three
> relative to an apple today must grow at an exponential rate as the
> consumer ages.
I don't think this is a good example, because experiences at age 3 are
unusual in a couple of respects. First, many of us have considerable
childhood amnesia which will still carry over to age 3. This makes
experiences from that young age unusual compared to most past experience.
Second, the argument implicitly relies on the three year old's view
of the relative importance of his life at age thirty. But three year
olds do not have an adult's developed sense of rationality. An argument
about the rationality of discounting towards the past or future should
be based on rational actors.
A better example would be a 60 year old considering an apple eaten when
he was 30. However IMO that does not make the case as sharply as the
(misleading) example presented.
> This implication of invariance is difficult to accept. Rather than
> viewing past consumption as progressively more important as the
> consumer ages, it is more plausible that agents discount past
> consumption. When asked most people would prefer having completed an
> onerous task last week over having to perform one this evening; they
> would prefer a pleasurable meal this evening to one ten years ago; and
> they would prefer to be leaving on vacation to returning to work when
> their vacation is over.[Notes 5,6] Whenever one thinks, "I'm glad it
> is finally three o'clock and that lecture is over," it reflects
> discounting of the past. The disagreeable experience is less painful
> when it is in the past than when it is being experienced in the
> present.[Note 7]
These examples don't make the case at all well, and in fact they are
clearly a completely different phenomenon. According to reasonable views
of discounting, the degree to which the immediate future and past are
discounted would be practically nil. Yet most people would far prefer
having an unpleastant experience one minute in the past than having it
begin in one minute. This cannot be due to discounting.
IMO the real effect here is that we prefer situations in which our
life is improving. Whatever our current status, we want to improve it.
We aim for a situation where bad things are in the past and good things
in the future. In any case, whatever the reason for this preference,
it is not an effect of discounting.
> Beyond revealed preference, the strongest theoretical argument in
> favor of discounting, involves what Bohm-Bawerk referred to as the
> "brevity and uncertainty of human life." According to this view,
> people discount future felicity because they may not be around to
> enjoy it. An individual weighs pleasures at date t by the conditional
> probability of living until date t.
I think there is more to it than this. The future is inherently
uncertain. We are all familiar with the maxim, a bird in the hand is
worth two in the bush. Consuming goods in the present gives us guaranteed
value. Holding them until the future presents many hazards. We may
die, as mentioned above. But there are other things that can go wrong.
We could lose the goods or they could be stolen. They could go bad.
Our circumstances could change so that they are less valuable, etc.
Presumably such factors, acting through evolutionary pressure, have
programmed discounting into us. With a good enough theory we could
probably even predict or explain the observed discount rates, based on
conditions in our remote past.
> There is a parallel between mortality and forgetfulness. The past
> dies when we forget.[Note 8] Do we value today meals eaten ten years
> ago, if we cannot even remember what we ate? How is our current
> utility reduced by past pains that we cannot even remember? Imperfect
> memory justifies discounting the past in much the same way that
> mortality justifies discounting the future, and just as mortality
> suggests that we discount the far off future more than the near
> future, imperfect memory suggests that we discount the far off past
> more than the recent past.
I can't see how evolutionary effects could cause discounting of the past.
There is nothing you can do to change the past so it doesn't matter what
you think of it, unlike the future. The analogy between mortality (and,
as I suggest above, other sources of future uncertainty) and forgetfulness
is not accurate because the one has operational effects and the other
does not.
> Finally, heterogeneity in discount rates across individuals undermines
> the case for retrospective consistency. If, as is commonly believed,
> the rich and educated are more forward looking, do they place
> relatively _less_ weight on the past as invariance would require? To
> the extent that individual discount rates are endogenous, as argued by
> Becker and Mulligan [1997], does the weight on the past adjust to
> maintain delta = 1/beta? Or is the weight on the past also
> endogenous? After all, we spend considerable resources keeping the
> past alive, through stories, photographs, souvenirs, and diaries. If
> the weight on the past is endogenous, wouldn't it be a remarkable
> quirk of fate of [sic] this weight were exactly equal to the inverse
> of the discount factor?
I think there would be at least weak evolutionary pressure designed
to make delta = 1/beta, that is, to make us view our past decisions
as reasonable. Otherwise we would constantly be fretting over things
we can no longer change. Also, we might be tempted to change our
actions to bring delta into line with 1/beta, which would imply giving
relatively greater weight to the future. But as I have argued above,
there are already evolutionary reasons for us to discount the future,
and any factor which would introduce a bias into that discount rate
would reduce survival.
(In fact, this seems to be a strong argument against the authors'
proposal, in that it would seemingly apply to any society which has
evolved an optimal discount rate for their conditions. The authors
would step in and propose to manipulate things to decrease the discount
rate and give more weight to the future, hurting the survivability of
that society.)
> We conclude that in almost any reasonable formulation, future selves
> weight current consumption less heavily than does the current self.
> The implication is that tastes change over time. We now consider the
> welfare implications of this observation.
I don't agree that this effect can be aptly described as "tastes changing
over time". Your tastes don't change, all of your likes and dislikes
are stable. Furthermore, whatever changes you would experience in regard
to your future satisfaction with current decisions are fully predictable.
In consuming today, you are fully and completely aware that you will
regret it tomorrow, in the authors' formulation.
I have a couple of other points to make, the first being that if this
effect were real it would be both self-evident and widely recognized
("stop me before I consume again"), and second that there _are_ in fact
arguments along these lines which IMO would be more convincing (such as
a claim that our evolutionary heritage has left us with a discount rate
inappropriate for modern society, as it has with our tastes in food).
I will expand on these in a future message.
Hal
This archive was generated by hypermail 2.1.5 : Fri Nov 01 2002 - 15:31:47 MST