Draft letter not sent
Subj: RE: Fundamental problems in welfare economics
Date: 06/06/03 06:58:01 GMT Daylight Time
Sent from the Internet (Details)
Dear Professor Arrow
Thank you very much for replying. I understand what you say about the
polemical and arrogant tone. My message read as if I were saying that
such things as the inflation rate for the poor and adult equivalence scales
were left out of the theory of welfare economics, which is not what I meant
to convey. I say more about this below.
Perhaps I, like many people, am prone to blame my own weaknesses and
frustrations on others. In my own case, I have failed to bring
attention to a serious flaw in development economics, while at the same time
cursing other people for similar failings. It is quite hard to be a lone
voice in a crowd, knowing something and being unable to communicate it.
What I realised two years ago was that when studying hungry people,
economists never took account of the possibility that their results might be
influenced by changes in longevity.
A year before that, in 2000, I was a trustee of a family fund, hoping to
further justice, partly economic justice. I thought that the debate on
globalisation could do with better statistics on poverty trends, and I
thought that somehow the fund could help with this if the money were very
well targeted. It was around this time that I read "Growth is Good for
the Poor". First I read the article in the Economist, to which I
subscribed for many years, describing the conclusion of the research. I
was ready to accept the conclusion, and approached the article itself in
what I would describe as a naive way. When I saw the document I was
astounded. There seemed to be gross distortions of the truth. I could
hardly believe that such a piece of writing could be broadcast to the world.
I was shocked by several aspects. One of the thoughts I had was "but they
haven't followed real people - they've just looked at the averages, which
don't show us how well or badly people did; maybe the figures come out
that way because poor people died earlier in some places". I remember
thinking to myself that I hadn't realised that was the way economists got
their conclusions. These were longitudinal conclusions, but the data
I thought to myself
[the rise in the average for the poorest fifth]
[the average income rise for people in the poorest fifth]".
They measured the first one, but said they had found out the second one.
At the same time, there seemed to be many questionable things about this
document. I have a list of fourteen flaws, several of them pointing to
factors in the analysis which are completely unknown. The factors were
part of the analysis, but not the logic of the document: the authors
simply left out discussion of awkward questions about the flaws. The fact
that there appeared to be so many things wrong confused me. I suppose I
like to see the whole of a problem, insofar as that is possible. I could
not understand this problem, because there were so many facets to it.
Part of what I could not understand was how these things could go
uncorrected, both in the sense of tradition forbidding such poor work, and
in the sense of not being immediately corrected by prominent economists in
the next edition of the newspaper.
To cut a long story short, I knew I was right - this was not the way to
measure the progress of poor people. Nor was reducing the proportion of
poor people, which I found offensive as a concept. I told a senior
economist in the UK Department for International Development at a meeting of
NGO people and experts with DFID staff, and was laughed at. I contacted
several economists who seemed to have done relevant work. I could not
bring myself to question the World Bank staff themselves, because I regarded
them as liars. No-one could be stupid enough to have made the arguments
they had made.
I could not find anyone with whom I could discuss the life-length issue
sensibly. A head of department for development studies at a very
well-known university said to me that the problem was data availability.
I knew it wasn't. The problem was that people were not thinking about
the problem. This was a theoretical matter, not an empirical one.
I spoke to various experts about the flaw, but they did not shed any more
light on it than I could shed myself. The experts included Professor
Kanbur. When I learned from him that no-one had thought about it, I
worked furiously for several days to try to understand it. I realised
that since poor people kept dropping off the bottom of the income
distribution, the figure for average income was kept artificially high.
I expect that it was then that I realised most people would make average
income rise by dying.
A good way to measure economic welfare in a country might be to multiply the
median income by mean life expectancy - provided income were adjusted by
a) age structure, b) the inflation rate for the person on median income, and
c) extra necessary items of expenditure, and some account were taken of d)
significant changes in assets.
Such a way would, however, value richer people's lives as more valuable than
poor people's lives. That would be so even if some extra monetary value
were assigned to a year of life. That kind of problem arises with any
measure which combines life length and a static measure of welfare.
Bentham said that the felicific calculus should include the duration of
pleasures. But what weight do we assign to an extra year of life?
When we are 1 year old? When we are 5? 10? 15? 35? 55? 95?
That is not a scientific question. It is a moral question.
Social science is the science of human societies. Moral philosophy is
something else entirely.
Professor Kanbur's paper on differential mortality and poverty measurement
confuses the two. I realise that there is a demand on economists to
produce answers as to whether things are getting better or worse. But
the only valid scientific statements from a social scientist are those based
on data, theory and assumptions whose appropriateness for the purpose has
been thought out.
Any system of welfare economics which did take into account life length
would be highly subjective. It would have to place a monetary value on
each human life. But there is no such value. We would have to invent
it. I would not dare to put such a value on life. I would not dare to
place a higher value on a rich person's life than a poor person's life.
I suggest that the lives of people with little are more valuable, to them
and to us, than we think. Necessity is the mother of invention.
People who have little make do with little. My own preference is to
spend time with people of many backgrounds, because you get different angles
on life. Poor and apparently ignorant people have much to teach me.
They deal with real life, if anyone does. A hungry person has a healthy
set of priorities. They may well be more sane than many of us who have a
surplus of food, money, options, moral stances, thoughts, fears.
I suppose I have had a surplus of fears about what I was looking at. To
be saying something so obvious, and yet apparently at a high level of theory
in a respected academic subject, seemed ludicrous. My image of the world
changed. I had had a faith that people who were in charge of such
things as economics understood them. It was as if the world had gone
mad. Of course you can't use the Sen index or the FGT index or the
average income for the poorest fifth as an indicator of economic gains to
hungry people. It still baffles me: why would anyone think that you
could? The omission of information on differential mortality seemed to
me to be the devious workings of the selfish gene: a drive towards natural
selection, through false logic in the brains of people whose genes were
Above, I said that my message read as if I were saying that such things as
the inflation rate for the poor and adult equivalence scales were left out
of the theory of welfare economics. That is of course not true. They
are, however, not required as a theoretical basis for conclusions from
academic economists on aggregate economic gains or losses for poor people.
In that respect, the theory underpinning practical work within welfare
economics certainly is defective.
It is quite a big assumption to make that in both China and Brazil the
inflation rate for basic food was the same as that for the whole economy,
for all periods studied. It is also quite a big assumption to make that
there are no systematic effects, for example, between disproportionate rises
or falls in the price of basic food compared to the overall inflation rate,
correlated with the variable whose effect the social scientist is trying to
measure - such as rises in income per capita, or trade volumes.
It is also quite a big assumption to make that average food requirements did
not increase due to changes in the age structure. This flaw appears in
the World Bank's Millennium Goal indicator for the proportion of poor people
(and another indicator, the share of income in the poorest 20%). It is
certainly a flaw, because there are now fewer children per adult in many
countries in Asia and Latin America, and we do not know that the change has
not affected poor people. The effect of a fall in the birth rate is to
make total consumption requirements fall and average consumption
requirements rise. The dollar line is inappropriate for a world where
the proportion of adults is increasing.
What would I use to calculate welfare? I would not be so arrogant as to
try. If I were measuring economic power, then I would take into account
inflation at different income and asset levels, income, assets,
environmental assets (both natural and made by humans), freebies, subsidies,
debts and maybe more. I would not judge economic inequality by income
inequality. If I look around me, at my life and those of others, I see
people with low income and high assets. Are they rich or poor? I see
people with high income and high expenses. They could move to a cheaper
area or a cheaper country, but then they would not have the job they want
with the social life that they want. They might not get a job which pays
so much. If they have children in a local school, then they do not want
to move them. Housing costs where I live are high. Housing costs, and
other costs, in other parts of my country are lower. What is economic
inequality between my town and those? Suppose wages and assets are half
what they are in my town. Are the people half as well off economically?
That has two answers. One depends on prices and necessary expenditure.
The other doesn't. In terms of absolute economic welfare, the answer
may be no. In terms of relative economic power, the answer may be yes.
If prices and wages for poor people rise in tandem, then in one sense they
have done no better. But in another sense they have: the goods of the
non-poor are now more affordable (in terms of percentage increases in income
needed to buy them) than before, for those who manage to scrape together a
surplus. Have poor people done better? Well, no and yes. You
would need both statistics, to answer two separate questions. Life is