Subj: Growth and poverty Date: 17/12/02 To: r.thomas@open.ac.uk Dear Ray Thank you for your interest in my work. I thought I'd send you a short version of some of what I say in my long article. On growth and poverty Economic growth statistics are a function not of income gains alone, but of income gains and demographic change. Quintile income growth statistics are not a function of income gains alone, but of income gains and demographic change. So why do economists say "incomes of the poor rose x%" when they do not know this? Matt Subj: Re: Development economics Date: 03/12/02 To: R.Thomas@open.ac.uk Dear Ray Thank you for the reply, and I appreciate your interest. What I am doing is raising theoretical issues in economics and statistics - pointing out the logical steps missed out by anyone who infers average gains from cross-sectional data, in countries and quintiles where demographic changes are unknown. I am therefore casting doubt on validity of the main assumptions of development statistics and economics as disciplines. The effects of demographic changes are simply left out of the logic of inferring average gains. Why this is so, when birth rates and death rates obviously vary so much, I don't know. If economic growth in China is influenced to some extent by changes in the ratio of adults to children, then to that extent, the rise in per capita income does not reflect the average gain in income. It's just that there are more producers and fewer consumers. Whether anyone thinks this is good or bad is irrelevant to a scientific investigation of whether people are on average better off for their age. The life course of everyone's income can remain the same (or get worse), and yet the average income in the whole country can go up simply because of a change in demographic composition. There is no theory in economics to stop an economist saying "the incomes of the poor rose" even if a dictator makes everyone's income go down, as long as the dictator kills enough poor people. This sounds incredible, but it's true. My thoughts on this started when I read "Growth is good for the poor". My first reaction was to think "the results might have been due to the poor dying earlier in some countries than others". There is no theory in economics to deal with this sort of question. I've asked the Nobel Prize winner about it, and many others. The then head of development studies in Oxford, Frances Stewart, said to me "I'm sure demography has a lot to do with it" in relation to Dollar and Kraay but didn't appreciate that this raised a fundamental question about economic theory. The equation (the percentage rise in average income for the geographical area) = (the percentage average income gain to individuals) is not a mathematical certainty. It's contingent on there being zero demographic change. On denominators, my point would be that the denominators (for both growth statistics and the fall in the proportion of people below the poverty line) are different people at the start and end. Clear examples: China (different proportion of children after one-child policy) and Uganda (fall in life expectancy in 90s, likely to be mostly for poor people). Per capita figures are just that. They only tell us about average individual gains if a) we know that demographic changes were minimal or b) we somehow calculate the effects on later averages of demographic changes. This way of looking at it seems to me to be the only way that makes sense. The alternative - currently, usually taken for granted but with no theoretical support - is to assume that demographic change had no effect. Why? I can't think of a reason. Social scientists assume that demographic changes had no effect unless proved otherwise, but the burden of proof is surely on the social scientist's shoulders. Otherwise we don't have science. These things are fundamental questions of social science theory and philosophy. On the practical implications for studies of inequality and so on, the parts of the document where I talk about "the three bi-directional cogs of economic thought" amount to three ways in which the logic of economics discriminates against poorer people. Certain kinds of bogus answers are therefore bound to emerge under certain conditions. Nobody has any model for deciding when these conditions apply. My basic point is that longevity statistics are free of all these problems, and if the aim of policies is to improve people's lives, longevity is the one sensible indicator to aim for, especially when studying people who don't have enough to eat. Last year I spoke to Eric Swanson, who is the head of the World Development Indicators project for the World Bank. I mentioned to him that if more poor people died, for example from AIDS, the proportion of poor people would fall faster. He said that people were better off after the plague, and that "we could find out how many people died, but I don't know if it would be a good idea to say it". Three months ago I mentioned some of what he had said to Robert Johnston, who is the Chief of Statistical Services at the UN. He laughed and said "Well, Larry Summers tried to do that with AIDS when he was at the Bank". These are the kinds of things I was referring to in my previous email as having found out which can't be substantiated. I think it's more profitable to raise questions of theory, partly since it is not just people who want to deliberately misuse statistics who make the mistaken inferences. Matt Subj: Happy new year Date: 28/12/02 To: R.Thomas@open.ac.uk File: D:\Current articles\Dear Ray 2 .doc (44032 bytes) DL Time (32000 bps): < 1 minute Dear Ray Thank you very much for your encouraging and helpful, comments, and for the dancing Santa, which will help me with my own dancing in the future! Thanks also for the introduction to John Parsons, with whom I'll be in touch. I take your points about making the focus of my observations clearer. I'm writing about several things, among them: 1. problems for statisticians in using cross-sectional data to infer aggregate individual change (JRSS type stuff) 2. specific problems for economists (PPP, children's consumption). Demographic change does confound attempts to measure meaningful changes in aggregate poverty, unless d.c. is known to be minimal. But more important is the other way round: per capita wealth changes (e.g. income) do not fundamentally measure aggregate trends in individual wealth. Whether they do or not in practice is a task for the social scientist. I want to get away from talking about poverty and talk about economic theory, because a) that is where the thinking needs to be done, and b) the considerations apply to rich people as well. Certainly all current statements by economists and statisticians about world poverty are inferred, not calculated. It seems to me that there are several books I could write on the various topics. Most of my time since spring 2000 on this has been spent trying to find out other work that has been done on the theoretical problems, especially the differential mortality problem - but in reality this was just a way of spending time adjusting to the idea that development economists have missed out some very important things. It's only recently that I realised about PPP - mostly through the work of Reddy and Pogge - which I think at the moment is the final piece in the jigsaw of development statistics. I've attached a letter with more detail. Cheers Matt