New directions in development economics

How to make the Millennium Development Goal on poverty effective




Matt Berkley

30 Mar 2003



Official statements about global poverty trends need revision, for several reasons.  


The first is that the dollar-a-day line currently applies both to adults and to children.   In reality, adults need more food;   and adults are becoming more predominant in the population of poor people as the years go by.   The old poverty line of 1990  -  a per-person consumption expenditure line for households  -  thus comes to represent a lower level of consumption adequacy.   


The second is that for the majority of countries, there are no adjustments for inflation of the goods and services which poor people buy.   The shifting percentage of people below an international dollar equivalent therefore represents an unknown trend in the percentage of people below the original consumption level.    The clear implication is that categorical statements about consumption poverty trends are inappropriate when based simply on consumption-expenditure trends.  


The third relates not to the trends in the percentages, but to the correct interpretation of these trends.   Clearly, it is not accurate to speak of people “rising out of poverty” if the percentage falls due to falls in birth rates.    More seriously, if poor people live longer the percentage will fall more slowly;   in a country where they make less progress in survival, the percentage will fall faster.    There is thus a need to check on survival rates before making statements that people have done better or worse by looking at the percentages.  


Several Millennium Development Goal indicators need some tweaking to make them into meaningful measures of poverty alleviation.   If the current methods can be shown to be inadequate, then that is perhaps to be welcomed.    If there were no room for improvement in poverty monitoring methods, then progress in improving poor people’s lives would be slow, as now.    If there is room for improvement, then social scientists can fulfil more of their potential in the formulation of effective policies.   


Welfare economics  -  the science of economic outcomes for real people  -  is more complex than it looks.   Here I raise some questions, which relate to three broad areas:   Birth, Life and Inflation.   






The level of consumption doesn’t tell you the level of consumption adequacy
(unless you know the level of need)


To meet their minimum needs, and so avoid poverty, people in 2015 will need a higher poverty line than the dollar level.   Why?  Because birth rates have gone down.   More of those people, proportionally, will be adults, and adults need more food than children.


This is an undiscussed but hopefully obvious flaw in many official statements about poverty trends. 


Let’s start with a definition.    Consumption poverty is not low consumption but low consumption relative to need.    This second is the stated intention of the World Bank’s dollar-a-day measure.    However, that measure in practice takes no account of whether the people are children or adults.  


The problem is twofold.  


First, proportions of children in the total population of poor people vary between countries.


Second, proportions of children vary in the same country over time.  


The definition used by the World Bank is this:   those people living in households where per capita consumption is less than what a dollar could buy in the United States.[1]   


A shorthand description such as “people on below a dollar a day” is strictly speaking inaccurate, for two reasons.     I would like to speak strictly here, in order to clarify what the statistic tells us and what it does not.    The first reason is that the dollar is an international dollar, not a real dollar  -  about which I say more below.   The second reason is that the method does not in reality count those whose consumption is below that level, for good reason:  if it did, it would count too many children and not enough adults as poor.    In reality, a more accurate and clear description is this:   the method counts the households where per-person consumption is less than the level, and then counts the number of people in those households.   A household’s members are either  all above or all below the line.  


A baby on a dollar a day in the USA would be far more likely to have their basic needs met than an adult.     That’s because babies eat less.   Where this method is applied precisely, in a country where there are a lot of children, the families will in fact be better off than the standard of living which the dollar line is meant to capture.   They don’t need so much to be above the consumption-adequacy line, because they save by virtue of more members being children.   Conversely, in a country where there aren’t many children per adult, the consumption needs in those families, per person, are greater.   A perfectly measured dollar line would therefore represent, in the adult-dominated country, a lower level of consumption adequacy.   It would miss out families who really are poor.  


What actually happens is this.   In every country, people around the dollar line live in households with differents proportions of children.    The per-person method always falsely counts some big families (by which I mean here ones with more children) as poor when they aren’t, and falsely counts some small families as non-poor.   In the real world, a fixed per-person consumption line fails to adjust for the fact that children are cheaper for a family than adults are.   But that doesn’t matter, if the average family size is the same as it was when the consumption line was drawn up.   It does matter if the average family size is significantly different from the norm.    Then, the line ends up falsely including or excluding too many people.   


This mathematical effect of child-adult ratios on per-person need  -  and therefore on the appropriate level for a consumption line to capture consumption poverty  -  is quite big.   We can easily see that by thinking about a couple with two children compared to a couple with three.    The couple with two children needs significantly more per person for both adults and children to achieve the same minimum standard of living as the other family.  


Let’s think about a planet where in 1980 there were significantly more children per adult than in 2000.   Suppose we know the changing proportion of people below a certain level of consumption.   Does that tell us the changing proportion of people in consumption poverty at the old level?    No.     The old consumption level  -  in particular the old meal size  -  was averaged out over more children.   Children don’t need as much food as adults.   Now there are more adults per child.   So the old meal size isn’t enough to meet the old standard of consumption adequacy.   


But the intention of research on consumption poverty is to measure consumption adequacy, not the size of people’s meals in general irrespective of the size of the people.   So for measuring consumption poverty, any particular consumption level is either appropriate to the old demographic profile (or, if you like, the average size of children-and-adults-jumbled-up in the past) or to the new mix of people.     But it can’t be appropriate to both, unless the proportion of children is about the same as before.    And so it can’t accurately be described, when comparing different countries or times, as a measure of consumption poverty.    On the planet I live on, in countries where most of the world’s population live, the proportion of adults has risen dramatically over the last 40 years.  


Voluntary family planning is thought by many to be a good thing.   But it isn’t the same as economic progress for real individuals.    A fall in birth rates causes statistics about the economy  -  population averages or poorest-fifth averages  -  to rise faster than the average gain to real people relative to their actual age.    Most of what I say about headcounts also applies to economists’ studies using population averages.[2]    But I’m digressing.    Let’s stick to headcounts for the moment.  


In reality, if the global proportion of people below a fixed consumption line stays the same, and birth rates fall, there are in fact now more people below the old standard.    It is hard to escape the conclusion that the World Bank’s global figures for both a) the trend and b) the current prevalence level of extreme poverty are significantly in error.   The real trend cannot be as dramatic, or the prevalence as low, as the official statistics claim.    The only real question here  -  leaving aside some other methodological ones which I mention later on  -  is by how much the statistics are wrong.  




Question:  “Why do you think it’s a big effect?”

Because of two things.  One, the ratio of adults to children has changed greatly.   You can check this in United Nations reports on population.   They have some nice graphs showing rapid increases in adult-child ratios (usually thought of the other way round, i.e. ratio of young dependants to working-age people).    Regionally, the ratios will change in Africa last.    The ratio changed dramatically in China from the late 1970s to the early 1990s.    Globally, the scale of the fall in numbers of children has been so dramatic that the food needs per person are obviously significantly higher.    We might also think about the fact that food needs per person are increased by any increased longevity of adults, as well as the falling birth rates.    That’s because if adults live longer, there are more of them around at any one time.    The longevity of poor adults  -  however we define them  -  is not given a lot of official attention.   So we wouldn’t be very wise to assume much about how long they live   -   probably best to ask people who know a country well rather than try and extrapolate from whole-country statistics.    Bear in mind that poor people are, according to official statistics at least, mostly malnourished  -  say 1100 million poor according to World Bank and 800 million malnourished according to UN.    I am saying that at the old level, which is the appropriate level of consumption inadequacy, we could safely say, if there were no other methodological issues to consider (see below!) that there are many more than the 1100 million.    But the point is that it’s not a safe assumption that poor people have extended their lifespan in proportion to the change in life expectancy in the country as a whole.  

Two, in many countries there are many people whose consumption levels are around the dollar lines.    So if we raise or lower the consumption line, we bring in or exclude a lot of people.    Even a raising of the poverty line by a few percentage points would make a significant difference to many conclusions about global poverty trends.   And it seems to me that the increased proportion of adults will raise the food requirements by quite a few percentage points.    It’s usually accepted by economists that 70% of poor families’ expenditure is on food.   If that’s true, then the food needs make a big difference to the overall adequacy of a given consumption level. 





Question:  “What about changing economies of scale in households?”


Good question.   Economists, like everyone else, know that bigger households are more efficient.   You buy in bigger quantities, and you economise on fuel and so on.   


So how does this change as couples have fewer children?    Well, the picture is slightly complicated by the fact that in today’s smaller families, the adults who are becoming more predominant (numerically if not in behaviour) consume more.    So the economies of scale from having extra adults in families are more than those of having more children.   


In other words, the extra economies of scale from the additional children in the past may not have been very big, because the additional children didn’t consume very much anyway.     I think that the main distorting effect of consumption-level statistics comes from jumbling up adults and children, not from ignoring economies of scale.  


The apparently clear facts about China, and about other countries, might lead us to reassess conclusions of social scientists based on “headcount” measures of consumption expenditure.    One such conclusion is in the World Bank’s publication “Globalisation, Growth and Poverty”.    Others are in the World Development Indicators and in a variety of other publications from UN organisations.  





Question:  “Even if the adults now aren’t as well fed as we thought, isn’t the following correct:  



1) the World Bank statistics are basically reliable, and

2)  the percentage of people below the line goes down,


3) consumption levels among the poor, per capita, have risen 


-  which is some kind of advance?”

This question raises quite a few others about what consumption-expenditure headcount trends can possibly tell us.    So a comprehensive answer needs a reading of the next sections, as well as consideration of some other mathematical determinants of the percentage at the end of the period.   For example, if the non-poor have a slower reduction in birth rates than the poor, or start living longer, the proportion of poor people goes down independently of what happens among the poor.   


But there’s a wider point.    The aim of poverty research isn’t to measure consumption levels, but to measure poverty.     Consumption levels are of only marginal interest, because the whole point is supposed to be to improve people’s standards of living.    All concepts of poverty are designed to capture not a minimum consumption level, but a minimum standard of living relative to need.    And this is just common sense.  


Suppose you were a child in the 1970s in a family whose members’ consumption level was each 10% below the minimum; and in 2003, when you’re grown up, your family’s consumption level is still 10% below the same minimum for each member.
But, as in real families in most countries, there are now fewer children in your family. 

In childhood, you had lots of brothers and sisters.  And the family was counted by a social scientist as poor.   In adulthood, your own offspring aren’t as numerous.   What does the social scientist say?    The per capita method shares out your own consumption (which is higher because you are an adult  -  and has to be, otherwise you wouldn’t be alive at all) among fewer children.   So the per capita method may well count you, and all your family’s members, as “not poor”.     A couple less children than your own parents had, and it may very well count your whole family  -  whose basic needs are each 10% unfulfilled  -  as not poor.  


The average consumption level   -  for adults and children jumbled up  -  is higher, but neither adults nor children in your family are any better off.   The fact that the average is higher is of no use whatsoever to your family.    It may represent an interesting fact for economic planners, but it is irrelevant to a sensible statement about your family’s poverty.    In fact, if you had had more children, the family wouldn’t have been much worse off anyway, because children don’t consume much.    So it’s not even true that your family is better off by virtue of having fewer mouths to feed.  


The question of adequacy is made even more important by the fact that we are here talking in many cases about people who are barely subsisting.    The dollar line is not set at a dollar in local currency.   It is in theory what a dollar could buy in the United States.   In real countries, a person below this line is several times worse off than they would be if they had a real dollar a day.  


Lower food adequacy means earlier death.    This links with the section below about longevity and cross-sectional statistics (which means statistics taken of people alive at particular times.).    I don’t think cross-sectional statistics are at all suitable for conclusions about how people who are vulnerable to early death have done   -  whether we are looking at medical patients, old people or malnourished people   -   unless we have some information on survival rates.    We would like to believe that poor people in the world are living longer.    But since there are now more adults per child, even if all the other usual assumptions are right, people now on the dollar line are consuming less for their age than those on the dollar line before.    The degree of vulnerability because of that decreased consumption adequacy is a cause for concern.  


From a human point of view, speaking for myself, I would say I was still poor if my family’s basic needs weren’t met considering their needs now.    I wouldn’t want to be counted as not poor just because there were fewer children in my family now even though they were all at the same level of deprivation, which is what the per capita method does at least for some families  -   and, it seems to me self-evidently, for many.  



Question:  “How does this affect our view of the Millennium Goal on halving poverty?”

The short answer is that as currently measured, it wouldn’t be a halving of the proportion of people at the original level of poverty.  


A halving of the proportion of people below a fixed consumption line  -  which is what the dollar line aims to be (though see below on prices)  -  is not a halving of the proportion of people in consumption poverty, but something less than a halving.   


The description “halving poverty” or “halving the number in extreme poverty” is incorrect.    How far it is from the truth in individual countries depends on both a) prices and b) the changing ratios of children to adults among poor and poorish people in each country.    All the percentages need adjusting.  



Life length

The fall in the proportion of poor people cannot tell you mathematically the number who rise above the poverty line
(unless you have other information)

Inferred conclusions from social scientists from changing headcounts are in any case in need of quite a lot more thought. 

For one thing, even where the child-adult ratio is constant, the change in the percentage of poor people doesn’t tell us how well or badly poor people did.    That’s true even if we know the change in the depth of poverty (poverty gap ratio) or a number of other measures of poverty based on statistics on people alive at the time. 


Let’s think about a country which enables hungry people to live longer.    That might happen through food subsidies, or better health care, or various other measures.    That country will, other things being equal, see its percentage of poor people fall more slowly.    That’s a good result for poor people, but some social scientists would say the poor have done badly because there are still lots of them there!     Oddly, the fall in the percentage is the basis of the main indicator for the Millennium Goal on poverty.   


What’s the appropriate response for a social scientist, or any thinking or caring person, to the mathematical relationship between life length and the percentage?


We could cross our fingers and hope the percentages aren’t affected very much.  But:

  1. The percentages are sometimes put in complex equations (see “Assessing Aid” from the World Bank).    Are the results of the equations affected a lot, or not very much?   God knows. 


  1. Where poor people die of AIDS and/or malnutrition, not only do their own deaths reduce the headcount, but also the birth rate falls  -  because those people don’t have any more children.   Their children, who would have been counted as poor, are never born.   In other circumstances, a falling birth rate falling might be a good thing for all concerned.   But in this case it’s hardly the sort of thing we want our statistical system to count as success in helping the poor.    There might be other economic effects too of people dying of AIDS, or of malnutrition.    Again, even if for some gruesome reason the survivors end up better off, we don’t want to confuse that with a better outcome for the poor during the period.   

    We know that some countries have started better health care systems than others.    And we know that in some countries children have started to survive much longer, and adults too.    Those people will make the headcount look “worse” by staying alive.   Do we really want to ignore this, to tell ourselves it’s not a problem?    


Usually what seems to happen in countries is that first people begin living longer, then they have fewer children.    Poor people insure themselves by having more children, until the time when they see that children are now surviving better.    That’s quite sensible.    What this means for the percentage of poor people is that when child survival among the poor goes up, for a time you see more children and a rise in the headcount.   After some years, you see fewer poor children being born.   


Let’s say you’re in the government of a poor country.   If you don’t tackle child mortality, you may see a fall in the proportion of poor people (and make the mistake of thinking that you’ve done well).    But poor people are going to have more children, which means that in the medium term you’re going to have more poor people anyway.   


What about the other option?    If you do improve child survival, you may see the proportion fall more slowly for, say, ten or fifteen years.    More children survive during that time.    But in the medium term, the birth rate falls anyway as people begin to have confidence in their children’s survival chances.    So some slightly longer-term thinking gives a better result all round.    Poor people don’t want to have too many children.   They just want to make sure they have enough even if some die.   So helping children survive is quite a good idea all round.    It has a temporary effect on the percentage, but the aim is to help the poor, not the statistics.  




Question: “Are changes in longevity big enough to change the percentages?”  


No-one knows, as regards past statistics.   No-one can possibly know, in relation to future statistics.   If you don’t know the effects of changing longevity on the percentage, then you can’t say whether the fall in the percentage is good or bad, or how good, or how bad, for the poor people.    As in other areas of science, the investigator has to explain why they think one type of real event (rising or falling consumption) caused a statistical change, and not at least partially another type of real event (changes in life length).    It seems counterintuitive to me to say that 100% of the fall in the poverty headcount in a country like Uganda, where life expectancy and child survival fell, was due to raised incomes.    But that is the current practice in development studies.   It is current practice to say that a rapid fall in the headcount shows more “poverty reduction” and by implication that the poor did that much better compared to in other countries.   But that doesn’t follow.  


It is in practice hard to account for the various possible causes of headcounts changing.   There aren’t many data about longevity of the poor.    But that is fundamentally irrelevant to thinking about the mechanics of the statistics.   


In a wider context, I would like to question whether “halving poverty” is a sensible aim, given that

a)      many millions of poor people die early of malnutrition each year, and

b)      this number can vary.  


I suggest it is a sensible aim only if i) we know that poor people are making good progress on surviving longer;   ii) we know about price inflation for poor people.  

Currently, we don’t know either. 

I would also suggest revisions to the Millennium Goals on slum populations, school attendance and water access, for similar reasons.   They all deal with reducing proportions of people.   In the age of AIDS, population projections for several countries are vastly reduced from what they were a few years ago.    A rough-and-ready solution to this  Millennium Goal problem is fairly easy:    use available information on death rates, and come to a sensible judgement about whether adequate progress is being made in keeping poor people alive a bit longer.    In countries where that isn’t happening, the changing percentages aren’t much of a guide as to how many people met the relevant criterion.   


In the same vein, we might think about other terms such as “poverty reduction” (applied to a fall in the headcount); “eradicating poverty”; and the World Bank’s phrase “our dream is a world free of poverty”.     I think that, even if we ignore the problem I’m describing here, this phrase is too idealistic.   I think that a practical, rather than a utopian aim, is more likely to inspire employees to success.   But back to my main point.    The Bank makes it clear that its aim is to help poor people.    In this world, many die of hunger.   I suggest that a clearer mission statement for any such organisation would refer to “raising living standards of poor people”.  


I would like now to say a couple of things about living standards.   There are some people who object to economics almost in principle.  But if there are ways of measuring how much more hungry people ate, then that’s fine by me.   The real problem with economics is that it needs reform.   As with all social science, and all science, it’s as important to say what you don’t know as what you do know.    Clear terminology, with clear referents (meanings) goes a long way towards clear thinking and sensible conclusions.     If you don’t say what your assumptions are (see above on consumption poverty, and longevity) then you are very likely to end up saying things for which you have no evidence.   


In the case of measuring consumption poverty, the obstacles are rather more numerous, and high, than many non-specialists realise.   One problem is that of comparing the value of money, and/or goods consumed, across countries.    That’s a vast problem in itself, not yet solved.   Even the people in charge of the International Comparison Programme, whose general country price indices are used for monitoring the dollar-a-day headcounts for the Millennium Goal, say that they aren’t suitable for poverty measurement.   There are also several large problems about the comparability of different surveys using different questions about different goods bought by poor people in different countries;  about the truthfulness of respondents;  about how to extrapolate for missing data when only data from years far apart are available;  and so on.    These data are often so poorly regarded by investigators that no estimates are given for margins of error.    Despite this, it is common practice in economics to make categorical statements about average gains to real people or numbers of people “rising out of poverty”.    This is essentially a “crossing the fingers” approach to science, as well as a closing of eyes as to the various factors which can cause recorded statistics to change.  


Anyone familiar with these technical problems knows of the limitations of the studies which rely on assumptions about them.    The considerations I have raised above about demographic change  -  and those below on prices  -  are additional to those considered by most experts.    So we see there is quite a complex set of assumptions commonly made in development economics.    Births, deaths, prices, international comparison.   All these may confound (mess up) conclusions which rely on the simple assumptions about what caused of statistical changes.   


If we’re honest, there are many unknowns in the economics of poverty.   Actually, that’s true anyway whether we’re honest about it or not.    Did people do better, or did their expenditure rise slower than prices, so that they were worse off?    A study such as “Growth is Good for the Poor” makes no attempt to answer such a question.    Did changes in birth rates cause a different composition of people in the poorest fifth, confounding the claimed gains or losses to poor people?    Don’t know.    Did the average for the poorest fifth rise more slowly because the poor survived longer?    Don’t know.  


My solution for poverty outcome measurement is unconventional, cheap and easy for those who want to try it.   It has been mentioned by Rowntree, the pioneering researcher in England 100 years ago, and Sen.   It may gain in relative popularity as we examine more closely the credibility of consumption statistics.   It’s at the end. 







The consumption-expenditure trend doesn’t tell you the trend in consumption level

(unless you know about prices)


There is a common misconception in economics that consumption expenditure among the poor is the same thing as consumption level (and consumption adequacy).     It isn’t.    Consumption may turn out to have the same trend as expenditure, or it may not.    We can adjust the expenditure trend by the consumer price index, as the World Bank does, but in that case consumption will have the same trend as expenditure only if the rate of inflation among those things the poor buy is the same as the inflation rate for the whole country.   


In a globalising world, and in most worlds there have been in the past, that equivalence can’t be assumed   -   see Adam Smith, writing about “the real recompence of labour” in 1776.    He noticed prices paid by poorer people falling.  


How does such a disproportionate fall in prices affect the poor?    Let’s suppose a price fall benefits them a lot, but doesn’t affect the overall price index much.   That will happen in a number of circumstances in real life.    The poor then appear to the economist not to have benefited much  -  unless the economist knows the change in the price index for poor people’s goods.    So when an economist says that they have found a change in the headcount measure of poverty, we can ask them whether they adjusted for inflation.    Otherwise, if they measured income changes, we know that they haven’t been adjusted to what we can call changes in real income.   The real (purchasing-power) value of the income rise or fall may be in the opposite direction to what the non-adjusting economist says.   


And if the economist used consumption expenditure, which is the most common measure in countries where most people live, they don’t really know that consumption of the poor went up or down 1% just by looking at the consumption expenditure statistics.    The fact that I spent more (or less) money this year doesn’t tell a social scientist that I consumed more (or less).    And so it doesn’t tell the social scientist whether I should be counted as poor.    It just tells them the numbers on the faces of the coins I handed over.[3]     That’s a different statistic from the amount of food, fuel, clothes and water I bought.    The overall consumer inflation rate takes many other things besides those into account, but if I’m poor, prices of those things have nothing to do with the amounts of various things which I can buy with my money.  


The headcount measures in the vast majority of countries, in the World Bank’s methodology on global poverty, don’t take into account the inflation rate for poor people’s goods.    So we really don’t know, even for a country where the child-adult ratio is constant over time, and longevity at all levels of consumption doesn’t change, what the change in the headcount means in terms of consumption adequacy.    We might know the low-consumption expenditure headcount, but not the low-consumption headcount at the same consumption level as the old line.  


What we do know is that where prices rise, or fall, for such things as water, the costs or benefits to the poor will not be adequately measured.   In other words, the poor will falsely appear not to be much better or worse off unless the social scientist knows about price trends and puts them into the equations.     What we don’t know is how much this skews the published headcounts, which claim to measure not consumption expenditure, but for some reason consumption level.    We can’t in reality get there without knowing price trends.    And we can’t get from there to conclusions about consumption adequacy  -  consumption poverty  -  without knowing demographic trends.    Economists coming to conclusions about economic gains and losses usually make assumptions that demographic trends  -  changes in birth rates and death rates   -   have zero influence on cross-sectional statistics, which clearly isn’t true for most countries in the world in the last 50 years.  


The rise or fall in income per capita is a function of two cross-sectional statistics.   It isn’t the same statistic as the average gain or loss expressed as a percentage, which is a longitudinal statistic.   In conditions where demographic change has zero effect on the average, the two have the same number.    But they aren’t the same thing.   The average percentage gain is a function of the rise (or fall) in income per capita and the impact of several types of demographic change on income per capita.    It is not calculable from the change in income per capita.   Without calculations based on demographic change, it  can only be inferred from the trend in income per capita via assumptions about demography.    


The economics of poverty measurement is an interesting subject in itself.    I don’t think it’s cost-effective at the moment.    If you read some of the documents from the International Comparison Programme, you might find the complexity of the task of price comparisons quite surprising.   How much do poor people pay for food?    Simple question, hard to answer.   Where?   In the north of the country, or the south-east?    In the villages or the towns or the cities?    In more remote places or more accessible ones?    In small quantities or bulk?    For what quality of produce?    When?  In the drought season or the flood season?    In a year of plenty or a year of dearth?    How many people pay which price?   


But then also this method of assessing poverty needs consumption expenditure data, and then conversion to consumption level, and then to consumption adequacy.   The second and third of these, it is more true than not to say, are not done currently by the World Bank.    Nor are calculations as to how many of the hungry people died, making the statistics look better.   


What is the output from this method?    If we think about all these technical problems, the output  -  in relation to the large studies which have been used to shape “poverty reduction” policies  -  is inconclusive in some cases, and grossly misleading in others.  The work on growth and poorest fifths, like all such work from economists, is inconclusive.    The work on global poverty trends is grossly misleading.    In relation to the policies themselves, the claimed success is just that.    Is growth good for the poor, or are the poor often worse off due to higher prices, masked by the omission of price data by the World Bank staff?     I have no idea, and nor do they.


No-one in the world knows what price trends were for people in poorest fifths in poor countries over the last few decades.    So no-one knows what the trends were in consumption.   So the question “is economic growth associated with higher consumption among people in poorest fifths” can’t be answered.   


So we do not know, from the studies, whether the policy outputs   -   the effects on poor people  -   are good or bad.     What we do know is that if we apply the World Bank’s own stated definition of poverty as inadequate consumption to its conclusions based on consumption expenditure, the results will look very different.   


Economic growth, which for some reason the World Bank claims is essential for “poverty reduction”, is not a policy to help the poor, but a policy not to help the poor.   Whether this policy not to help the poor is associated with on average very small gains or very small losses to poor people is essentially irrelevant.   The claimed gains  -  or, to put it another way, the claimed losses from low economic growth  -   would be very small if the economic research methods were credible.   But they aren’t, except to people who don’t know what the research left out.  


What are the inputs?     The inputs are not only the expenditures on the salaries of World Bank staff, but also on those of all staff who use the statistical conclusions in the United Nations Statistical Division;   the United Nations Development Programme;    UNESCAP;  UNESCO and several other UN organisations.   





Cheap solution to poverty outcome measurement problems:
Measure life.  

If you live longer, then at least you’ve eaten enough to keep you alive.    A social scientist can’t argue with that.    


People whose consumption is less adequate live shorter lives.   

Whole-country life expectancy statistics, and child survival rate statistics, may give some indication of life length changes among the poor.    It depends on the country, its proportion of poor people and its health inequalities, but since malnourished people last less time, there is often a bias in these statistics towards measuring what happens to poor people.    Child mortality is concentrated, not surprisingly, among the poorest.   And we might expect adults whose children die more often to have worse health themselves.    Worse health, and worse life expectancy for adults, is, again not surprisingly, largely a result of malnutrition.    If you look at causes of death in WHO documents, you find that people are recorded as having many different causes of death.   But it’s a fact of life that disease hits hungry people far harder than other people.    The real cause of death is mostly malnutrition, because other people get the same diseases and live.  


In any case, in order to interpret cross-sectional statistics  -  in order to come to conclusions about poor people doing better or worse  -  we need to know about survival rates.    In the 21st century, we should be capable of more than the “zero assumption” current in economics, that life length never affects cross-sectional statistics.   At least, we should be capable of considering such difficult and inconvenient questions.     If we aren’t capable of doing this, then we aren’t capable of providing a credible structure for conclusions about economic gains, or credible conclusions.    If someone says “But maybe these results are caused by poor people living longer, rather than doing worse economically” there is at the moment nothing an economist can counter with apart from crossing their fingers.   I don’t think this is good enough.    Nor am I enthusiastic about the current jargon of “poverty reduction” whose precise meaning is often unclear.     I hope that more people will begin to feel a deep sense of unease at this phrase, when applied to a reduction in the proportion of people with less than a minimally adequate standard of living.    I hope for that, not because of some accurate knowledge about what the statistics mean   -  except for the case of the increasing average size of poor people over the years and so the decreasing adequacy of the same consumption line  -   but because of a deep unease that difficult questions are being ignored by bureaucracies and academic torpor.    Where a question that might be very relevant hasn’t been considered at all, it’s generally quite a good idea to consider it.   


I am not saying that particular poverty statistics are affected paradoxically by changing longevity among the poor.   I am saying that the logical relationships have not been considered;   that I therefore do not trust comparative studies claiming to show outcomes of different policies for people who are extremely poor;   that the conclusions of such studies are inferential in nature, not mathematical;   that the inferences about economic gains are based on assumptions about life length which have no empirical or theoretical basis.    How important to the conclusions are these assumptions?     I do not know.    Nor do the authors.   


What is self-evident is that there is some boundary, some level of shortening or lengthening of poor people’s lives beyond which particular types of statistical conclusions are affected paradoxically.    To me, it’s not really a paradox.    I don’t expect cross-sectional statistics to tell me about longitudinal trends (trends for real people).    There is no necessary mathematical relationship between them whatsoever.    The set of people in a geographical area at the start of a period is a different set of people from the set of people there at the end.    What was the average gain or loss to real people, given two averages for two sets of people?    I don’t know.    If retired people live longer, the average goes down even though people on average do better.  


One day, there will be a science of economic welfare which acknowledges such truths.    In the meantime, given the uncertainties I have outlined as to the interpretation of trends in cross-sectional economic statistics on poor people, I suggest that, other things being equal, a country has helped the poor more if it has added to the length of what are, almost by definition, short lives.   






[Note 13 Feb 2007:  This is an old article which may not represent my current views.  For instance, I no longer believe that there will be a science of economic welfare.  There are assessments of welfare or well-being or prosperity or poverty.   There are judgements of these.  There are scientific methods which can be employed in support of such judgements, concerned with things like adding up numbers.  But the concept of assessing or judging benefits is a very different concept from that of measuring them.]

[1] The dollar amount has been updated, so that it is now intended to capture what $1.08 would buy today  in the US if prices were at 1993 levels.  

[2] And to how we would like to define “utility”.    Utility in the sense of

a) best consequences (i.e. the best personal history) for the greatest number of people

is not the same thing as

b) higher average in a later population.   

Why not?   Well, it might be the same if demographic changes are minimal.   But if some people begin living longer, that affects the average.   If the worse-off people   -  which may be the majority of people  -  live longer, the average will fall but utility, in the sense of (a), which is the philosopher’s usual definition, will be more!    This needs a bit of sorting out in the theory of economics, because the practice uses averages.  

[3]  For many people, consumption expenditure measures the value of food they produce themselves.   But many poor people pay for water, and other goods and services whose prices vary across time and space.