A non-economist's view of some World Bank aims and policy research
28 October 2004
Title changed (word “some” added); formatting changed and contact details added: 8 September 2008
Five mistakes by social scientists may help explain some puzzling features of international statistics.
The existence of the errors may be of some significance for future policy decisions.
1. Longevity error
I once sought out the World Bank document "Growth is Good for the Poor", having seen one of its conclusions in a newspaper. The document claimed to measure average benefits to people under different conditions and policies.
I was surprised that the authors confused
"the average for the poorest fifth rose 1%"
"people in the poorest fifth had average rises of 1%".
Those are two different statements.
I thought to myself "they can't know the average outcome of each policy, because they don't know how many of the poorest survived in each country".
I also thought "if this is the way economists do their work it is the wrong way".
Nor did I like the idea of "poverty reduction", for the same reason: if more poor people die, the "reduction" will happen faster.
The confusion between "the average rose" and "on average people had rises" is a fundamental error in the theory of economics.
So is the error concerning "poverty reduction".
It is also a fundamental error in the thinking of the FAO, whose head complained about the number of hungry people going up slightly. The number will go up if the most hungry live longer.
2. Omission of relevant prices
A second thing which puzzled me about the World Bank policy document was that the authors showed no sign of having assessed the purchasing power of poor people's money.
To omit to estimate relevant prices under each policy would be very odd. Later I confirmed that the World Bank did not estimate prices for poor people in different countries or under different policies. So the policy assessments were based on the wrong inflation rates.
No economist has compiled prices of staple food in different countries for past years. So all public claims as to the relative merits of policies for the poorest people, from such international data, are misleading.
The British Secretary of State for International Development announced to Parliament on 26 April 2004 that global poverty would be halved if a certain level of GDP per capita were reached. But he has not estimated how GDP per capita correlates with food prices. So exactly what he means by "global poverty" is not clear. Economists, politicians and the World Bank talk about this issue as if they had data on the cost of living, but they have estimated neither prices nor needs. The cost of living is a function of both.
The usual survey method aims to measure consumption expenditure. At present, therefore, the most prominent aims of DfID and the World Bank amount to reducing the proportion of low spenders. The fact that people spend 1% less does not tell a researcher that they received 1% less. The fact that they spend more does not tell a researcher that they "rose out of poverty".
3. Claiming to measure economic need without estimating food need
Thirdly, per capita figures are not suitable if the proportion of children varies. That is because adults need more food than children. To assess need without assessing food needs is a mistake.
4. Cost of living depends on both prices and quantities required
Fourthly, other aspects of need were also omitted: such as need for rent.
5. Omission of asset changes in claiming economic "benefits"
Fifthly, changes in land ownership were not taken into account. But if you inherit or lose land, in reality that is an economic gain or loss.
World Bank policy documents have claimed to measure "average benefits" without looking at whether people lost or gained land, or whether they increased or decreased debt.
The same applies to the World Bank statements about the level of global poverty.
I later confirmed that these five errors were indeed common practice in macroeconomics. Economists often did not know the difference between cross-sectional and longitudinal statistics, and were steeped in a tradition of using survey data without looking at the cost of living for the target group.
6. Errors may explain puzzles
I think these errors may provide partial explanations of
why Millennium Goal indicator 1 is ahead of the FAO indicator (the FAO adjust for the falling proportion of children's meals needed, and the Bank do not);
and why some countries have low GDP and high life expectancy.
They may also help explain why the World Bank statistic is ahead of health indicators. Given the uncertainty about prices, survival rates, items required, and assets, it might be that the World Bank is being overpessimistic about progress: in such a scenario, it could be that people in the World Bank target group are living longer, and/or food is getting cheaper and/or other expenses lower.
However, let us bear in mind the bigger picture, including poor progress on health goals. If people are getting richer, then would they not be eating better? If they are eating better, would they not be getting healthier? The Bank's continued insistence that poor people are doing well, and that great progress has been made since 1981, without data on food prices or survival rates, seems doubtful.
In any case, the Bank would need to adjust for children's meals in order to comply even with standard practice in economics. This would reduce the claimed "poverty reduction". The FAO and the Bank cannot both be right about the food.
In relation to the life length-GDP puzzle, the official figure for GDP per capita depends on the national price index. But national price indices are mathematically dominated by items which have more money spent on them. So if a government makes food cheap, the economist, using the inflation rate dominated by luxury goods, thinks people are not much better off.
As a result, GDP per capita, as a measure of basic human welfare, has a tendency towards bias against countries where food is cheap. It also has a tendency towards bias against countries where
life is long,
basic services are cheap,
land ownership is high and
rental needs are low.
The tendencies are an inevitable consequence of the economists leaving things out which matter to people in real life.
These tendencies apply also to macroeconomists' policy advice from international studies.
The policy assessments have an undeniable tendency towards bias against policies associated with
cheap basic services,
high land ownership,
low need for rent, and
low personal debt.
I am not claiming that the biases affect any statistics in any significant way. What I am saying is that the tradition in economics of equating "income" with "profit" without estimating necessary expenditure is unjustified. The macroeconomists have not estimated inflation rates for poor people, or the cost of living for anyone.
I am also saying that economists lack basic training in the difference between cross-sectional statistics (statistics about the economy or segments of the economy) and longitudinal statistics (statistics about what happens to individual people).
There is no "average outcome" where survival rates vary.
Since it is impossible to assess what economists' statistics are telling them about consumption levels, it would seem wise to look elsewhere for guidance as to the effects of policies.
An alternative to the present situation (in which social scientists can make fallacious claims without fear) is one where social scientists have stricter rules.
I propose five axioms for the social sciences. A rather poorly-written argument for these appears at www.mattberkley.com/5axioms.htm .
+44 (0) 7868397699