From: Matt Berkley [mattberkley@dsl.pipex.com] Sent: 30 April 2003 22:42 To: Ravi Kanbur Dear Professor Kanbur We spoke two years ago about the influence of death rates on static welfare measures (the proportion of people in poverty, population averages and averages of “poorest fifths”. I have some further comments on measuring poverty trends - and on both development economics and welfare economics - which I think you may find of interest. I could publish these observations in academic journals, but perhaps they are more urgent than the up-to-two-year timescale would allow. Perhaps my observations account for some of the differences between perceptions of economists and non-economists as to how well or badly poor people are doing. Matt Berkley Millennium Goals and poverty trends Matt Berkley 30 April 2003 I would like raise some questions about official statements on poverty trends. These comments are fairly brief. These questions raise further questions as to whether economists are the right people to analyse trends in poverty. The main points are: 1. Official statements infer that a greater fall of the proportion of people in poverty shows more benefit to the poor. But the proportion falls faster if poor people fail to live longer. It falls more slowly if the poor live longer. The proportion is therefore only a guide where we know about death rates. Since global child mortality indicators are doing badly, and the proportion of poor people is falling at the intended rate, we might ask whether there is a connection between the two. Where children often die, the adults are not likely to live very long either. The proportion also depends on birth rates and migration. Unless you exclude changes in these as causes of the change in the proportion, you cannot tell how many people rose out of poverty. 2. Official methods assume that the changing proportion of people living on under a consumption line shows how many now live under the same level of consumption poverty as in 1990. But that cannot be true. There are in 2003 more adults per child than before. Adults need more food than children. Poor families spend most of their money on food. Therefore, a fixed consumption amount per person represents in 2003 a worse level of consumption poverty than in 1990. This effect is amplified by another consequence of falls in birth rates: a smaller household is less efficient, raising the costs necessary per person to cross the original poverty line. 3. Official statements treat the 2003 proportion of people below an international dollar a day as showing the changing proportion below the same level of consumption as the 1990 line represented. It cannot show us that. Costs have risen for one specific reason; and the international dollar is of unknown value to the poor relative to its value in 1990. First, there is increasing urbanisation. Costs in urban areas are usually higher. That is because a) prices are higher there and b) you need to spend money on more things (for example, rent; transport to work). So the poor people of today need to spend more money than the poor of ten years ago, if they are to reach the same level of consumption adequacy as before. Whether these costs are outweighed by some other factor has not been analysed. Second, no-one knows the value (in consumption terms) of an international dollar today compared to in 1990, or what it will be worth in 2015. The World Bank’s assessment of the proportion of people below an international dollar is an accounting exercise. The dollar line is not a consumption line. It is not a consumption-inadequacy line. Whether it by some fluke coincides with a consistent consumption-inadequacy line is not known. For such a fluke to occur, it would be necessary for prices of poor people’s goods to have fallen, to compensate for a) the higher proportion of adults, b) the smaller size of households and c) the higher proportion in urban areas. The World Bank has simply counted money, whose purchasing power for the poor is unknown. It has presented this accounting exercise as representing gains to the poor. But even if we a) knew the effects of birth and death rates, and b) adjusted for the needs of children, and c) adjusted for increased costs due to urbanisation, we would still not know d) whether an international dollar now buys more or less of the goods which poor people consume. Consumption expenditure figures, which are what the World Bank uses in the vast majority of cases, do not and cannot tell us the amount people consumed. Expenditure statistics refer to the numbers on the coins, not the amounts bought. Consumption expenditure figures adjusted by the overall rate of inflation in a country (the World Bank method) do not tell us the inflation rate for the poor. A consumption expenditure line is not accurately described as a poverty line. Poverty is inadequacy. But an expenditure line does not even tell us how many people are now below the original level of consumption (of adults and children jumbled up, and ignoring how much people need in the new circumstances). An amount of money is only worth what it can buy. What can a poor person on an international dollar buy in 2003 per day, compared to 1990? No-one knows. How many people in 2003 are below the same level of consumption poverty represented by the 1990 dollar line? No-one knows. I would like to make a suggestion. It is possible to find out whether poor people are doing better. Governments which want to improve the lot of the poor do not need complex statistics, but common sense. Common sense says that poor people die of hunger, and that helping to keep people alive is a more humane aim than using statistics about the economy which look “better” if poor people die; and financial statistics which look “better” even if the proportion of people below the original poverty level goes up. There are several deep confusions in academic economics over what the changing proportion of a) poor people, or b) people under a consumption line, or c) people under an expenditure line, tell us about economic gains or losses to them. I hope that the above clarifies some of the issues. There is also a deep confusion among economists over whether a) “average income rose 1%” is the same statistic as b)“people on average had income gains of 1%”. It isn’t. “The average for the population of 2000 was higher than the average for the population of 1990” does not and cannot tell us that “the people gained on average by 1%”. In a rich country, if retired people live longer, the average gain is larger than the rise in the average. That is because the population average falls if retired people live longer. Oddly, economists often think that they are talking about “utility” - in the philosopher’s definition of benefits to people over time - when they use population averages of welfare measured at separate times. These definitions of utility are not the same. A philosopher such as Jeremy Bentham would not have recognised a modern economist writing of people’s “utility level” (what they have now) as representing the same thing as he meant by “utility” as consequences. Let us assume for the sake of argument that a person’s annual income represents their standard of living. If people on below-mean income (which is most of us, in any country) go on living, then average income will fall. The economist, thinking that population averages show the amount of benefit or loss to people on average, will think that people have done worse. Jeremy Bentham would say they have done better. If people on below-mean income begin to die at a higher rate, the economist and the philosopher will each come to the opposite conclusion. And yet economists think that the aim of increasing GDP is based on Bentham’s utiliarianism! Quite odd. Economists and averages The reader may be able to work out from the above that a conclusion such as “Growth is good for the poor” is not supported by consumption-expenditure statistics on “poorest fifths”. a) Unless you know the inflation rates for the poor, you can’t know how much they gained or lost on average. That is because the numbers on the coins don’t tell you how much poor people were able to buy. Economists study the economy. They do not have statistics about the inflation rate for the poor. They may know the change in nominal income of the poor (as a theoretical segment of the economy) but even if they know about age structure and needs, they don’t know about changes in real income (what the money can buy). b) Unless you know changes in consumption need (due to changes in age structure, including the changing ratio of adults to children), you can’t know how much they gained or lost on average, compared to people of the same age before. That is because i) adults need more food than children, and ii) people of different ages usually have different incomes. A demographic shift among age groups, even confined to adults, will change the population average. If there are now more people at higher-earning ages the population average can rise even if everyone is worse off for their age than before. c) Unless you know the increased or decreased needs for expenditure, you can’t say whether the people were better or worse off in terms of consumption or income. That is true even if you somehow were able to work out the changing value of their money. Contrary to the beliefs of many in the economics profession (based on an imaginary model of the world where no-one is born, grows older or dies; where the inflation rate for the poor must be the same as that for the rich), we do not know whether increases in average incomes were or were not associated with economic gains to poor people. But even if they were, that would not tell us that increasing output per person was in any way an important policy for helping the poor. d) Unless you know how the later average for the “poorest fifth” was affected by death rates, you can’t know the average gain or loss during the period. We do not know whether growth was good for the poor or not. What we do know is that the poor - or rather the majority - are bad for growth, at least the definition of growth which is in terms of average income or output, rather than total income or output. Suppose an increase in average incomes were associated with some benefit to the poor in the past. Would that tell us that policies to increase average incomes would be beneficial to the poor in the future? The supposition amounts to saying that increasing the incomes of the non-poor helps the poor. An assertion that “increases in average incomes will be a central element in increasing the standards of living of the poor” is in any case a different assertion from “growth is to some extent associated with higher living standards for the poor”. The question then would be, how much - compared with other options? We do not know. Therefore we do not know from past studies whether increasing other people’s incomes is a central part of helping the poor. We do not even know if it has been associated in the past with benefits or losses to the poor. It is a theory with no empirical evidence to support it. It is a somewhat odd theory, in that it would seem more sensible to help the poor rather than the non-poor if the aim is to help the poor. Before the invention of statistical methods, a king could either be a bad king or a good king to the poor. He did not need statistics to be either; nor did his subjects need them to know the difference. Statistics can give a more accurate impression of the state of human progress than you can get by walking around and talking to people, and using your eyes. Or they can give a less accurate impression - giving the receiver of statements about statistics an opposite impression to what is happening. To understand the relationship, if any, between statistics and what is going on in the real world, we need to understand where the numbers came from, how the people did the sums, and how they came to conclusions. We need to understand what people who present us with statistics are saying, and why. Accurate description of statistical findings is more important than the numbers. Inaccurate description is equivalent to getting the numbers wrong. Economists commonly misdescribe their statistical findings as to consumption gains to real people. Let us hope that either a science of welfare economics will emerge, or common sense will prevail.