A non-economist's observations on development economics Matt Berkley Draft 11 August 2004 Summary: Official statements on health, poverty and hunger in the human species, viewed together, contain some puzzles. Partial solutions to these puzzles may lie in the words used by social scientists to describe research conclusions, and in the words used by governments to describe their aims. The article presents an analysis of economists' measures of human prosperity, and a partial solution to hunger in the human species. * * * * One day, I saw a claim in the Economist newspaper that the incomes of the world's poor people rose in line with national GDP per person. I thought that this was a useful finding, if it were true. So I looked at the document. The authors, from the World Bank, discussed "average benefits" to poor people from different policies. One of their conclusions was that spending on social services did not show any special benefit to poor people. They wrote of economic gains accruing to the "poorest fifth". Looking at the method, I thought to myself: "But they haven't followed real people at all. They've confused "the average for the poorest fifth was 1% higher at a later date" with "on average the poorest fifth of people had rises of 1%"." and "Of course they can't know the average income trend for the poorest people. In a country where poor people die earlier, the average for the poorest fifth goes up. You could get results like this if enough poor people died in particular countries". and "Is this the way economists do their work? Because it is the wrong way". * * * * I wrote down "How do exchange rates affect the measure of extreme poverty? If exchange rates are not involved (i.e. it's not a dollar measure) then how is the purchasing power of the poor's money assessed?". The World Bank authors made no mention at all of this, and seemed to have treated everyone in each country as having the same inflation rate. * * * * Around the same time, I thought "I don't agree with the aim of poverty reduction, because it is faster if poor people die earlier". Since many millions die early, directly and indirectly from malnourishment, the figures may be influenced by changes in death rates, in the wrong direction - in one country, many countries or in the world as a whole. It seemed to me that existing projections for numbers of poor people in 2015 must be based on particular assumptions about survival rates. I brought the longevity error to the attention of a number of internationally known economists and others - including, in rough order: John Burton, senior economist at DfID, in a workshop at a DfID Development Forum, Simon Maxwell of the Overseas Development Institute, Jonathan Morduch of Princeton, Claire Melamed of Christian Aid, Edward Anderson of Sussex, Robert Wade of the London School of Economics, Caroline Anstey, head of media relations at the World Bank, Srikanth Puranam, speechwriter for the President of the World Bank, Michael Schultz, Chief Social Development Adviser at DfID, Frances Harper, statistician at DfID, Maria Eugenia Bonilla-Chacin, Jeffrey Hammer, lead economist at the World Bank, Jeffrey Sachs of the WHO Commission on Macroeconomics, Brian Hammond, head of development statistics at the OECD; Eric Swanson, head of the World Development Indicators project at the World Bank, Ravi Kanbur, Kenneth Arrow, Frances Stewart, Sudhir Anand, Andrew Smith, then Chief Secretary to the Treasury, Clare Short, then a Governor of the World Bank, Carol Howard of the Economist newspaper, Amartya Sen, Anthony Giddens, Dan Hausman, Daniel Little, Uskali Maki, Richard Jolly, Bernard Williams, Anthony Atkinson, James Wolfensohn, Tony Baldry, Chairman of the International Development Committee of the House of Commons, Hilary Benn, now a Governor of the World Bank, Gordon Brown, British Alternate Governor of the World Bank and Governor of the IMF, and many others. After Professor Kanbur confirmed that even Professor Sen had not written about this problem, I sat down to try to understand it better. I realised that a) all statistical averages of the living (including GDP per capita) are affected the wrong way by survival rates during the previous period; b) a country's people would look more prosperous to an macroeconomist if life length inequality increased, and less prosperous if life length inequality narrowed; c) most people are below per capita income, so the macroeconomist counts most people as doing better if they die; d) the longevity error was relevant to an old puzzle in development economics: economists say that in places such as Cuba, Sri Lanka and Kerala people are very poor, but the people live to their seventies. Part of the answer is that the poorer people are still around in the figures. d) purely because of the longevity error, all economists' measures of the prosperity of peoples are wrong. Utility in economics The economists' concept of utility is not in practice a Benthamite (greatest good for the greatest number) concept. Bentham, the original utilitarian, thought that if something was good, then to have it for longer was better. The economists' concept of maximising the use of scarce resources was based on something else. Purchasing power of the poor I forgot about inflation because I was preoccupied with the longevity error. From the start I could see the World Bank economists were wrong to look at only the survivors each year, but I was not sure that there was no standard method in economics for adjusting for poor people's purchasing power in different countries. There is not. Adam Smith noted an observable difference between the purchasing power of poor people for food and overall inflation in 1776. He called the purchasing power of poor people "the real recompence of labour". Economists two hundred years later were assessing policies for the poor by adjusting the statistics by inflation rates dominated by the rich. When an economist says about a policy, after looking at their figures for different countries, "incomes of the poor rose", they have not adjusted for the price of food. Food needs Another thought which occurred to me at the start was that it was very strange that economists used per-person statistics. This seemed strange because in reality the proportion of children has varied between countries and also across time - especially in China. Children need more food than adults. Other needs Another thought was that economists looked at "income" (in reality consumption expenditure, with some income statistics and some data on the monetary value of people's own food) but not necessary items of expenditure. People need different things in different places. Climate is one reason. People need different things according to whether they live in cities and pay rent, or have to take the bus to work, or need medical or other services. Assets Macroeconomists also ignored asset changes in inferring economic "benefits" and "losses" to people. Debts Macroeconomists also ignored both debt levels and debt interest payments. Structural bias in macroeconomics For all these reasons, economists' progress assessments, policy assessments and policy advice concerning poor people have necessarily been structurally - but not necessarily statistically in any significant way - biased against policies where: 1) the majority below the mean survive longer and/or 2) retired people survive longer and/or 3) the poorest survive longer and/or 4) prices rise for the rich and/or 5) poor people's wages from the rich rise and/or 6) food prices are low and/or 7) quantities required are low and/or 8) assets do not decrease for the poor and/or 9) debt is kept low. I do not have statistics on any of these. So if an economist says to me "I don't think any of these are serious problems" we both begin in the same position, because neither of us has statistics. But the economist has a harder time explaining 1) the discrepancies between their claims of prosperity and life length statistics for some countries; 2) discrepancies between World Bank statements on economic trends for the poorest in the world and FAO statements on hunger trends; 3) discrepancies between reported progress on Millennium Goal indicator 1 and many of the other 47 indicators, including health indicators. I should add in relation to structural biases 1-3 and 7 above that both the value of living, and the cost of living, are subjective: the value of living entirely subjective, and the cost of living at least partly subjective. Neither can be measured using statistics. The value of staying alive is not comparable to the value of anything else. The cost of an "equivalent standard of living" depends on what you specify as needs. I should also add that macroeconomists have not only used methods which have structural biases; they have also misdescribed statistical trends. To claim to know economic gains for poor people without knowing about survival rates, food prices, food needs or other needs is fundamentally mistaken. To claim to have good evidence on the "average rise" by looking at the "rise in the average", where the people are described as the poorest fifth of people in poor countries, is a fundamental mistake. The same applies to claims to know how many "rose out of poverty" in different countries. That is not knowable without survival rates. More importantly, nor are aggregate gains and losses in any respect inferrable without survival rates. The economist's statistics are not aggregate statistics. They are selective statistics, and should be labelled as such in the absence of good information on survival rates. It is worth noting that the standard measure in economic surveys of the poorest people is not income but expenditure. The fact that people spent 1% more does not tell you that they received 1% more. The common use by macroeconomists of the words "income rise" gives an unwarranted emotional effect to the reader of economic reports. "Expenditure rise" is more accurate. To claim "income" rises, even in a country where survival rates are known to be constant, on the basis of inflation rates dominated by rich people's purchases, is grossly misleading. If the economist has some reason to think that the inflation rate for poor people did not vary from the overall inflation rate, then such an inference or assumption requires labelling. If not, writing of income rises or falls for poor people is misleading. These statistical rises and falls (usually, in expenditure) are after adjusting by an inflation rate dominated by items bought by the non-poor. The tradition in macroeconomics is to fail to note this in adequate detail. Reply to the economist or politician My reply to the economist or politician who implies that the biases are not a problem for existing policy recommendations is to say: A. It is undeniable that the statistics have been misdescribed for decades as economic gains, and policies recommended to governments: without survival rate data, or consideration of the theoretical problem of longevity, even for the poorest; without data on prices for poor people, or any substantial theoretical analysis of the problem of inferring gains from policies without looking at food prices; without assessments of food needs; without assessments of other needs; without assessments of asset or debt changes. B. You cannot possibly know that these things are not a problem either individually or together. International data on survival rates, food prices, items needed, assets and debts have never been compiled. Because of the widespread belief among macroeconomists that what was being measured in expenditure surveys was poverty, the macroeconomics profession has ignored the need for evidence for its own inferences. C. The value of life and some aspects of the cost of living are not quantifiable in any case. D. There are four puzzles in international statistics which are perhaps hard to explain except by reference to the biases in economists' methods for inferring poverty and prosperity: 1. The health-and-poverty Millennium Goal puzzle: If malnourished people get richer, surely they eat better. If they eat better, surely their health improves. So how can this be: the World Bank says the poor are doing well, while the WHO say health progress is bad? 2. The hunger-and-poverty Millennium Goal puzzle: The FAO and the World Bank use similar survey data for the poverty and hunger counts. So how can this be: the FAO say progress on hunger is bad, but the Bank say the poor are doing well? 3. The "long life, low GDP" puzzle: How - apart from by confusing survival of poor people with "worse poverty" - did economists end up saying people in Cuba, Sri Lanka and Kerala were very poor, despite the fact that they lived a long time? 4. The health-and-surplus puzzle: Rich-country governments repeatedly state intentions to meet global health goals. But they keep failing. Why? The GDP-longevity puzzle Why do economists say that people in Sri Lanka, Cuba and Kerala are very poor when they live to their seventies? An undeniable part of the explanation is that there are more retired and poor people still left alive every year. Some of the rest of the explanation might be due to child survival rates. But if Cubans live to 76, the adults are surviving pretty well too. They can't be eating too badly. So a reasonable conclusion might be that consumption adequacy in Cuba is better than the economists think. If we look again at the list of structural biases, we can speculate that a) the economists are about right (though I am not sure the word "right" is appropriate where the inferences are about human welfare, and the value of life and the cost of living are necessarily subjective) b) the economists overestimated economic prosperity or c) the economists underestimated economic prosperity by use of the methods I have described as structurally biased. Where health indicators show good progress, I think c) is the only plausible answer. We know that the economists' methods have inbuilt biases. We do not have relevant statistics to test the biases. But we do know that people live longer with better health in Cuba. Undeniably, people in Cuba are economically richer than the economists say. Undeniably, the continued existence of people in the majority of people below the mean kept the statistic for GDP per capita down to some extent. That may not be a huge effect. Undeniably, economists have used methods which are in other respects structurally biased against certain kinds of economic outcomes. The simple explanation for why economists say people in Cuba are very poor is that the biases have operated to a significant extent. In relation to the prosperity of people within nations, per capita income has a bias towards richer people, since there are in every country fewer people above the mean than below it. Where income ratios between rich and poor are lower, the mean becomes more representative of the typical person's income (closer to the median). In countries where income is more unevenly distributed, per capita (mean) income is less representative of the typical person. In countries where incomes are more unequal, other things being equal the price of food is less adequately reflected in the overall inflation rate for the country. That is because the purchases of the rich, being numerous and/or expensive, swamp the effects of food price changes. We might expect a country which deliberately makes incomes less unequal to have lower GDP per capita (because more poor people survive); while in countries where incomes are more unequal, both the inflation rate and per capita income are less representative of the typical person. That is another reason why GDP per capita is not a measure of the prosperity of the people. Both "mean income" and "mean inflation" are measures which are biased towards the rich. Since average income (and/or GDP, GNI, GNP or such measures) is adjusted by the overall inflation rate, we have the product of two biased measures. Another reason why economists might end up saying that Cubans, Sri Lankans or Keralans are very poor might be connected with international exchange rates. In a country where the mix of items is very different from the USA or another comparison country, the inferred exchange rates may be skewed. If governments spend money on health projects rather than stimulating growth in wine bars or junk food industries, the economist may be unhappy, because s/he uses international dollars which are based on a very different mix of expenditure habits. Global food needs Economists' claims about whether poor people in the world had done better or worse - and their claims about which policies helped them more or less - suffered from not only the longevity error, the food price error, and the extra-items error but also the food needs error. Unless the one-child policy had no effect on poor people, the economists have progressively underestimated the amount of food needed by poor people in China. Unless birth rates have not fallen among the world's poor people, the economics profession has progressively underestimated their food needs. Partial solution to FAO-Bank Millennium Goal puzzle [How can the hungry be doing badly and the poorest doing well?]: The economists' error about food needs appears to be an undeniable partial solution to this puzzle. The FAO in 2003 claimed bad progress on hunger and the World Bank in 2004 claimed good progress on poverty. The FAO adjusted their line upwards each year because of an assumed lower proportion of children's meals. The Bank did not. Partial solution to health-poverty puzzle [WHO says health progress bad. Bank says poverty progress good]: See list of biases in economists' inferences about prosperity and poverty (above). Bank failed to estimate food needs (same solution as with FAO-Bank puzzle). Bank failed to estimate food prices. Bank failed to distinguish between prices and cost of living. Bank data known to be unreliable as different methods used in different countries. Food price problem compounded by use of international dollar known to be unreliable even for non-poor, and influenced by irrelevant trends in other countries. Possible that WHO stats dodgy as well. But less complex and less error-prone methods (hopefully). Partial solutions to health-and-surplus puzzle: Why are health goals not being met in a world of plenty? A partial solution may conceivably be found in two facts: 1. The aims of governments have been to reduce the proportions of poor and hungry people, rather than to increase consumption adequacy Note: The governments confused "poverty reduction" with "poverty alleviation"; and "fewer hungry people" with "hungry people eating better". 2. The methods of governments have been to reduce the proportion of low spenders. Note: The standard measure of what economists call "poverty" or "income" is expenditure, not income. Expenditure is not a measure of prosperity, and not a measure of deprivation. Possible partial solution to hungry people's needs: Focus on increasing life length of hungry. Statistic is cheap, difficult to get wrong, transparent, relatively incorruptible. Politicians would have less excuse to blame scientists for mistakes. Survival rate information necessary in any case for determining nutritional value of food to people in different countries. Food which carries disease not very valuable. Survival rate data essential to evaluate health services. Survival rate data necessary for all social sciences where conclusions presented as implying that people did better or worse. Survival rate data necessary to make any sense of hunger or financial statistics in human terms. If aim is to help poor, rich countries should not lend to countries which do not make intense effort to help poor stay alive. Give up "poverty reduction" idea. Cuba, Sri Lanka, Kerala alleviated poverty, not "reduced" it. 48 Millennium indicators not needed - too confusing. Admit mistakes or else compound them. Economists not needed. Good food needed.