Statistics and survival


Letter to the Economist newspaper, 10 October 2001





“This year the average size of fish in the pond is higher.  This confirms that, on average, the fish grew”.   Later, we realised big fish lived longer than small fish.


Perhaps survival rates are relevant to what is good for the poor (“Grinding the Poor”, September 27)  -  and not just for the obvious reason.  


Growth falls if the poor live longer.  

Growth rises if the poor live shorter lives.

Growth rises if inequality of life length between rich and poor rises.

Inequality of life length has varied greatly in poor countries.  

Among poor people, studies measure consumption [*] rather than income.

People who eat less have shorter lives.

People at the bottom of the poor group probably last less time.


Many people think the following two inferences are obviously valid. 


“Average income among the poorest fifth alive now is 1%  higher than for the poorest fifth alive last year.  Therefore, poor people’s incomes rose on average by 1%.”.


“The proportion of poor people went down.  Therefore, poor people’s incomes rose.”


Maybe it is time to find out.  Survival data, like cohort data, never give rise to the “fish” mistake.







* Later note:  My statement about studies measuring consumption was untrue.   The fact that I made this error may be of interest.  


For all large international studies, data were not on consumption but mostly on what people said they spent.     See for example:


“The 2004 survey by UNSD showed that of the 55 countries that responded to the question, 30 (55 percent) base their poverty calculations on expenditure data rather than income data.  The balance relied on income data, and a handful used both kinds of data.  The ability to spend is primarily determined by one’s income, but the two values are not identical since households often borrow, sell assets, or draw down savings when income is low. ” 

(Jonathan Morduch, Concepts of Poverty.  Draft chapter for UN Statistics Division Handbook on poverty measurement.  2005. )


National statistical departments carried out surveys.  Survey questions were mostly on spending.  Some questions were about what people grew and ate themselves.    In a minority of surveys questions were on income.    In all relevant cases, data supplied from national statistical offices for subsequent analysis by economists were on money.   


My error may have stemmed from a passage by David Dollar of the World Bank, author of the document quoted by the Economist about global poverty:    


“And let me emphasize that in poor countries like Vietnam, what we actually measure is what people really consume. We track how much rice, how much pork. Most poor people spend the overwhelming share of their budget on food. So, the issue is how much are they eating. So, we often say income distribution, but I want you to understand that we are measuring real consumption by people. So, when we talk about the income of the poor going up, think of it in very real terms. Malnourished people who had almost no protein are now getting more protein. That's the real issue that we are trying to get at.”  (Source: Transcript of Policy Forum, Cato Institute.  Is globalization good for the poor?  Tuesday, September 18, 2001. )




It is not clear why Dr Dollar claimed to know what people were eating in different countries.    Spending is not a measure of food.   


Suppose I know what you spent.  That does not tell me what you received in return.   I need to know relevant prices of food. 


Someone might say that on the face of it, it seems that for both global analyses and policy recommendations, the Governors of the World Bank, the British Government, and many economists worldwide, including some in well-known charities, confused spending with consumption.   


They might also say that on the face of it the economists and politicians, whenever they used the international money data to talk about poverty, went on to confuse the concepts of “consumption” and “consumption adequacy”.   


Suppose I know what you ate. 


I still cannot come to an opinion about how adequately you ate without thinking about your needs.   


Economists know need is influenced by economies of scale (bigger households are more efficient) and by age (bigger people need more).    You need more food if you do harder physical work, and if it’s colder.    One of the strangest errors of economists is in using per capita statistics for countries known to have falling proportions of children.  


Another part of the economist’s problem in relation to needs and food is this:  


Suppose I know about


a) relevant food prices (which are necessarily a bit subjective) and

b) food needs (ditto).


I still don’t know


c) how much money was left over for food after other expenses.   


Other expenses include rent, transport to work, and anything else which may vary across places, times, policies or circumstances. 


In real countries, a sensible thought might be that a lot of people have moved to cities and now need to pay rent.   

If the economist doesn’t think about this, then for this reason if for no other, the claim to have measured poverty would seem to be false.   


This raises philosophical questions about what kinds of expenses are necessary in different circumstances.    If poverty depends on both resources and need, and need is subjective, what does that make poverty?


But even if needs were measurable in principle, the question of whether they are measurable in practice remains.   


Even if we also ignore other aspects of economic gains and losses (assets, debts, communal assets, environmental assets) the following would still be true:  


The tradition in economics, in the largest studies reported in newspapers about global poverty, is of failing to take account of either the cost of dying or cost of living.


One line of argument which might be used is that in some time periods life length increases while birth rates fall, so while economists using the traditional methods, other things being equal, underestimate gains in “welfare” by omitting life length, they overestimate them at the country level by using per capita statistics and thereby counting today’s adults as needing the same as yesterday’s children.     It is not clear how the two might be compared. 


Note added 3 January 2006.