A long and careful documentation of Michael's point is in Miller, George,  
2000, *On Fairness and Efficiency*. Bristol: The Policy Press.  Miller is an  
English Epidemiologist in high standing as such, who publishes in The Lancet  
now and then along similar lines.  I recommend his book highly.  
  
A weakness in some of the postings herein, thus far, has been casual  
empiricism run riot.  In a scatter of points you can always find pairs that,  
taken separately, run against the regression line derived from the whole  
scatter, and James has done that (so, perhaps, have some of his antagonists,  
but I just read James recently).  Miller looks at the whole data set in  
England.  
  
As to life expectancy, or at least actual lifespans as experienced, when you  
plot them against income you get an L-shaped curve, with the angle at about  
$10,000 as I recall.  That is, lifespans rise rapidly with income up to  
about $10,000, then flatten out.  This says that those on the high end of  
the flat curve can be taxed to help those on the steep part of the curve  
without lowering their own lifespans, but extending those of the poor.  
  
Pareto and those in his echo-chamber tell us you cannot compare utilities  
among persons.  But you can compare lifetimes, and measure them without the  
heavy definitional problems that beset measures of income and wealth and  
happiness.  The finding seems pretty clear.  
  
Myself, I would use landownership rather than income on the abscissa.  
Unearned income extends your lifespan more than wage income, because you  
needn't work for it.  But I have to survive in a world where people insist  
on collecting data where people are ranked by income (whatever they mean by  
it), so I do the best I can with it.  
  
Mason Gaffney