From: Matt Berkley	mattberkley@dsl.pipex.com	
To: James K. Galbraith	Galbraith@mail.utexas.edu
RE: Politicians and statistics	

Dear James
 
Thanks very much.   As usual, there are so many separate points that we are discussing at the same time that it’s hard to write on all of them in a short letter.  
 
So for the moment here’s a comment on D + S, and some on inflation and on public economics.     
 
On the mistaken use of D + S data to support categorical statements, there are two groups of people who have done this:
 
One is economics professors discussing claims about growth and distribution, and policies and distribution.   The other is the economics PhDs and professors who generated the claims.    Some of these are in the World Bank and some outside it.    
 
The production of statistical claims by social scientists at the Bank usually comes with disclaimers from the authors, stating that they are not expressing an official view.   However, the Bank has a choice as to which pieces of research to promote to a wider audience.    Those chosen pieces of research are supported by a certain amount of work by public relations staff, and mentioned by representatives at press conferences.   
 
For example, the Chief Economist has in a press conference made the statistical claim about economic gains to poor people under conditions of economic growth.   “Growth is Good for the Poor” was for an extended period (several months, I think) the main headline on the World Bank’s website for all development issues and all development news.    
 
Other claims based on D + S data are also promoted regularly.    Some caveats about the difficulty of using different types of data all together are noted in official publications (the World Development Indicators, if I remember right, are an example of this).   
 
The claims are referred to in books funded by the Bank, and are widely mentioned in publications from, say, DfID and H.M. Treasury (on whose website on evidence for economic policy two papers by David Dollar are mentioned as the sole representatives of their genre).   
 
In summer 2000 David Dollar was in London and spoke personally with Clare Short about his findings.    It is not clear to me that the choice by Bank administrative staff to fund that trip was worth the cost.    Who funded his speech to the Cato Institute in September 2001?    If it was part of his work, on what grounds was he chosen to receive funding for his time rather than someone else?     If the D + S dataset had many caveats to begin with, why did the Chief Economist fail to take this into account in his own public statements and in choices as to which employees should receive funding for promotional work?    
 
It is true that the D + S data are not official, but the work of employees at the World Bank.    But here again, in such a situation there is an administrative decision to be made by an organisation as to which research results to publicise.    Here again, the public pronouncements on the dataset have not always been fully descriptive of its quality.    (I’m not sure I saw much in the way of caveats in D + S’s original article in the IMF magazine, though I could check again).   
 
Now, some people who have based claims on dodgy D + S data are Bhalla, Sala-i-Martin, Milanovic, Ravallion, Dollar, and Stern.     
 
I can’t complain about the public funding of the first two.    But there is a serious question to be asked about the public funding of the rest.    It is a question of economics.     How much did the work of these people cost, and did it give value for money?    
 
Well, you and me and Atkinson, but apparently not enough others, know the answer.    Were the funding decisions and the publicity decisions appropriate given the state of knowledge at the time of the research based on D + S?     I can’t see how they can have been.    
 
You see, if we start from the assumption that the Bank’s research department should have similar standards to those at a university, we find that it is lacking.   Many people are cynical about this, but I think it’s possible to present the issues in a different way.     What are the academic standards in the main conclusions from the Bank about the economics of poverty?     Very poor.    What are the costs overall in promoting these conclusions?     In terms of academic research budgets, very high.   
 
So really I’m doing public economics.    I’d like economists to join me in saying that the Bank’s research output on some questions which many economists think are major is severely lacking, if we look at it as a research institute in Washington.    Ignoring the politics, the output is defective, but expensive.    It’s a performance issue.    Aid budgets are designed to deal with competing claims.     In economic terms, the value of a piece of research needs to be weighed against the value from a development project.     And quite a few lives could have been saved with the money spent on these incomplete pieces of research.     
 
The omission of data on inflation means that we really have no idea what the results signify.    Or do we?    I can’t see any work which suggests that the differential for the inflation rate for grain (over or under the overall rate) has had any consistent mathematical relationship with income of the poor, or with growth, or trade.    
 
Can we assume there is no systematic relationship?    No.   If a country uses food fields for coffee, what happens to prices of food?    Can we just measure incomes of the poor and infer from the income the degree of economic gains or losses in comparisons between countries with more and less trade?     I can’t see how that is a sensible approach.    
 
If we look at the other independent variables in Dollar’s studies, we might think about why he should expect no systematic relationship between those and the inflation differential as well.      Basically, we have no idea how food prices changed with growth or trade or anything else.    Even if the overall variation in food prices between countries were small, there is no reason to think that the correlations aren’t there.    Now, those correlations might be in either direction.    But what Dollar and Kraay did was to assume a one-to-one relationship between incomes of the poor and economic gains to the poor  -  which include food price trends.     So that assumes a particular relationship between [growth] and [food price differentials (ie difference from the overall rate)].    
 
That hypothesis is, as far as I can see, a pure guess.   I see no reason whatsoever for thinking it, rather than another, should be preferred in these kinds of studies.    The question of whether growth raises incomes is absolutely inseparable from its effects on food prices   -   if we are talking about economic welfare of the poor.    
 
These studies only looked at part of the equation for wealth.    The other part -  the changing cost of living  -  is unexamined.     Now, I could hope or pray that the statistics fell out in the right way, so that income really does show economic gains.    But the world is big, and the circumstances widely varying.    So I can’t see the reason for any assumption whatsoever about price trends to poor people under particular conditions.    I certainly can’t see any reason to think that price differentials are always roughly zero, whatever the independent variable.    Dollar and Kraay’s conclusions needed two null hypotheses  -   one about income, the other about prices.    The fact that this has been swept under the carpet by many academics is not a reason against assessing the Bank’s research in terms of value for money.    Survival-rate data are far cheaper to process, and I argue that they provide a better option in terms of costs and benefits than income data without price data.     
 
It seems to me that contrary to what Ravi Kanbur and Kenneth Arrow have written to me, there are price data available.    Not necessarily good data, but surely if rice contributes to the CPI, then someone must have written down a price for it.    And there are many sources of information about whether rice became much more expensive or much cheaper in a year.    
 
Now, if the international organisation charged with the eradication of poverty, with a board composed of finance ministers, can’t find a way to obtain estimates of past grain prices, it is perhaps not providing competent research services to the taxpayer.   
 
Matt