From: hal@finney.org
Date: Mon Feb 19 2001 - 12:10:54 MST
Damien forwards from http://ap.tbo.com/ap/breaking/MGAKYLXFDJC.html:
> Claims by some scientists that humans in this century will have a life
> expectancy of 100 or even 120 are not realistic and not supported by the
> trends measuring the rates of death, said S. Jay Olshansky of the
> University of Illinois, Chicago.
We've had some discussions previously about these trend-based analyses.
Robin posted a reference last year,
Age-specific mortality rates have fallen at a steady exponential rate for
the entire last century. The classic source on this is
Modeling and Forecasting U. S. Mortality, by Ronald D. Lee,
Lawrence R. Carter, J. American Statistical Association,
87(419):659-671, Sep., 1992.
It is true that if you extrapolate this century-long exponential trend
forward you get the results of Olshansky.
I think we need to challenge the validity of exponential extrapolations
as a guide to future changes. My bet is that if you could look at most
data sets over a longer period, you would find that steady exponential
changes are NOT a good guide.
Unfortunately I don't have much specific evidence for this, but an
excellent example is population growth. Robin has an analysis at
http://hanson.gmu.edu/longgrow.html which shows that population growth
has gone through a series of exponential curves. Some have lasted for
millennia, others decades. In each case we saw an exponential curve go
through a "phase transition" to a new curve with a higher exponent and
more rapid change.
My guess is that the same thing will happen with age specific mortality.
We will enter a new regime in which the decreases in the past of about
1% per year in mortality may increase to 10% or more per year due to
a breakthrough in knowledge. Biology researchers often claim that the
advent of genomics will be so dramatic as to represent a new paradigm
for knowledge growth, so a breakthrough is plausible.
Such a decline in mortality rates would drastically increase lifespan.
If we could hit the 10% mark, then every 8 years mortality would decline
over 50%, which would roughly counteract the age-related doubling of
mortality every 8 years (figures from Curt Adams) and give us indefinite
lifespan.
Given that long term change is a series of exponential curves with
increasing exponents, and that we are on the verge of a breakthrough
in biological understanding according to the researchers involved, it
is entirely reasonable to project that we will see an increase in the
rate of mortality improvement, even though it has been constant for the
past century.
The question, then, is how valid is the model of piecewise exponential
growth, over a range of data? Are there other examples than population
which have been analyzed over a long period and where this model applies?
Or contrariwise are there examples where we have seen steady exponential
growth no matter how far back we look?
Hal
This archive was generated by hypermail 2.1.5 : Sat Nov 02 2002 - 08:05:57 MST