From: hal@finney.org
Date: Wed Nov 01 2000 - 12:56:59 MST
Another aspect of the reasoning in the paper that bothers me is that the
notion of regretting past decisions implicitly assumes new information
that was not present in the past.
Someone may regret not having saved more in the past. That would
make more available in the present. But if the reason for the extra
consumption in the past was due to the riskiness of the future, then the
decision may have been appropriate even though later regret was possible.
This happens if the risks did not actually occur.
A person who regrets not having saved more ten years ago would probably
have a very different view if he is on a plane about to crash. Then,
he will be very glad of his past consumption. In fact, he will wish
that he had consumed even more, lived life to its fullest, rather than
saving for a future which, as it happens, will never arise.
Not only death, but the many other risks I have described can give rise
to this effect. It does little good to save if you are paralyzed in
an accident and unable to enjoy your savings, or if your entire life
savings are lost due to theft or misfortune.
When we ask people today to look back and see if they regret having
consumed in the past (equivalent to asking whether the "past discount
rate" is less than the reciprocal of the future discount rate they
adopted back then), we are asking people who got lucky. From their
present perspective, they know that none of the risks that they were
(possibly unconsciously) considering in their past consumption decisions
actually arose, at least not in the most extreme form.
It's not right to require reciprocal past-future discounting for time
consistency, when we are looking at all past people but only a selected
subset of corresponding future people. The selection effect occurs
because we are only asking living people, we are probably only asking
relatively healthy people, and we are only asking people who have not
been subject to a devastating economic crash in the past few decades.
Hal
This archive was generated by hypermail 2.1.5 : Fri Nov 01 2002 - 15:31:48 MST