Michael Lebowitz writes:
I have just received some comments from a former colleague on the questions posed about the use of PPP. They include his comments in a letter plus an attachment which I have copied into the text below.
Please thank your colleague for his comments which show a generous effort; please also thank your other friend for pointing us to the Robert Wade article. For me, your colleague's comments illustrate the dilemma - even very well informed people assume PPP is just a statistical tool and not a economic model that is produced only through assuming the most severe and improbable "free market" models of price formation.
I can't tell how much of my postings he got to read, but I'll bet that he would have some doubts if he knew what was "under the hood" of the PPP. Without repeating my earlier comments, I will make a few below.
-------------------------
He writes:
I beg to disagree with the idea that the PPP method is "imaginary" and the Atlas method is "actual".
Agreed - at least the part about the Atlas method not being "actual". One is trying to compare "apples and oranges" (Manhattan with African villages) so there can not be a magic "actual" conversion number. However, and as you point out, at least exchange rates (used in the Atlas method) are "actual" for the international part of the economy and so inherently partly relevant and real. I (like authors on this subject) used the comparison with the Atlas method to show just how large a difference an (arbitrary) method can make.
As I explain in the attachment, the PPP exchange rate takes into account the price difference of goods and services between countries,or the purchasing power of a country's currency vis-a-vis the currencies of other countries (or the US dollar), whereas the market exchange rate does not take into account the price difference.
"They" say PPP does that (the Belassa-Samuelson argument). But it is not that simple (nor could it be, IMO). The actual PPP method does something very different than just "adjust for price differences" and one can only believe the PPP model is accurate if one also deeply believes in neo-classical General Equilibrium theory - and one of the most die-hard versions of it. Anything less and the numbers become weaker. I will illustrate this further down.
Take a simple example of Japan and the US. Say the market exchange rate is 110 Yens = One US$. Now take an equivalent basket--in quantity and quality--that contains a burger with fries and a drink. It costs 450 Yens in Tokyo and US$ 2.50 in New York. The PPP exchange rate is then 180 Yens = One US$ (450/2.50). There is nothing imaginary about the PPP exchange rate since it gives you the purchasing power of a country's currency vis-a-vis the US dollar.
Start with a smaller problem in this analogy: "burgers, fries and a drink" carry different connotations in different cultures that distort a purely physical comparison. When I was last in Japan (a while back) McDonalds had a cachet in certain young circles (hard to believe no?) so the burgers cost more than just a physical comparison would justify. Neo-classics wave off these price distortions as "imperfections" but...
The larger problem with the burger analogy is that it leaves out the next 2 steps in forming a PPP. First: which items do we put in the basket (or price vector)? A hamburger in the Kenyan basket or the Kenyan staple ugali in the Manhattan basket? Use the hamburger and Kenya is made to look more expensive; use ugali and New York looks more pricey (using both doesn't solve the problem since the difference isn't likely to be symmetrical and in any case this is impractical for 200 countries).
One universally accepted criticism of the in PPP/basket issue is the "substitution" problem (aka the Gershenkron effect) - poor people substitute poorer items in their basket but out of need not "preference". The PPP uses an price index method (called the Geary-Khamis method) that takes no account of this. Even the OECD (who had often been hardline over negotiations on this issue) ultimately refused to use it and turned to
Then comes a breathtaking leap. Manhattan and the Kenya have a small number of items in common. But the ratio between these items then gets applied to ALL the rest of the items in their respective economies that have never been traded (and some, like some labor services, that could never be tradable). The Gen. Eq. model that is produced assumes that this doesn't matter. So it treats the haircut sold in the village in Kenya AS IF it could be sold in Manhattan. Obviously this is a biased and unreal model that erases some important facts of life.
How much do these differences matter? Relax just one extreme assumption in the PPP model: include "substitution effect" which many (most?) neoclassics would since this is more of a statistical issue than an ideological one. The authors of the PPP acknowledge themselves that because of this effect PPP "overshoots" the mark, minimizing the difference between developed and developing countries. Dorwick and Akmal estimate this single change in the PPP model is estimated to produce 10% to 50% exaggeration in bilateral comparisons (using 1980 data while most of the exaggeration comes in the neo-liberal era). This effect alone is enough to reverse the conclusions of many of the neo-classical claims about the success of the last 10 years.
The important point is that the market exchange rate seems to be a valid conversion factor for settling payments between countries on account of trade, debt, aid, etc. and the PPP exchange rate seems to be a valid conversion factor for comparing the "standard of living" of people in different countries.
But shouldn't it be clear that the "standard of living" of Kenya and Manhattan can NOT be directly compared ? Mostly (not entirely) they exist in separate universes that never meet, so a true statistical measure in money terms is out of the question (one would have to try 'utility' or 'labour values'). PPP (or any comparison model) creates models of worlds AS IF the universes do meet and those models draw on their respective beliefs on how prices are formed. Different theories of price formation (conservative or liberal; neo-classical or classical or even 'moderate' neo-classical) produce very different numerical opinions. The World Bank et. al. present ONLY one numerical opinion, drawn from the most extreme model, and increasingly make it difficult to notice that the numerical opinion is not a "statistic". The "statistic" is then used to "prove" the success of the original assumptions.
Now please turn to the data shown in my attachment Table. In the GNI differences between the high income and middle + low income economies for any year (1996, 1998, or 2002), our focus should be on the ratios of the GNI of high income countries to the GNI of middle + low income countries under the Atlas and PPP methods separately. I see little change in the ratios between 1996 and 2002: the GNI gap between the high income countries and the middle + low income countries does not change over time (compare the 1996 and 2002 data).
GNI (Atlas Method): in 1996 the ratio is 4.41 to 1.00 and in 2002 the ratio is 4.18 to 1.00.
GNI (PPP Method): in 1996 the ratio is 1.36 to 1.00 and in 2002 the ratio is 1.30 to 1.00.
The fact that the ratios of GNI between the high income and middle + low income countries in each year differ so much under the two methods is simply because the Atlas Method does not take into account the price differences between countries and PPP Method does. There is no indication that the income gap between the rich and poor countries has narrowed. However, the income gap is larger with the market exchange rate compared to the income gap with the PPP exchange rate.
----
The attachment:
Gross National Income (GNI) of Countries, 1996, 1998, 2002
GNI (Atlas Method) GNI (PPP Method)
Billion US Dollars Billion US Dollars
Economy
1996 1998 2002 1996 1998 2002
High Income 23,772 22,592 25,596 20,574 20,745 27,516
Middle Income 4,141 4,401 5,056 8,305 8,834 15,884
Low Income 1,597 1,842 1,070 6,809 7,678 5,269
World 29,510 28,835 31,720 35,688 37,136 48,462
Source: World Bank, World Development Indicators, 1998, 2000, 2004.
The effort you have made is appreciated, and it may well be my confusion, but I am fairly sure the numbers you attached do not work out to the ratios above. (I think the ratios should be 4.14--->4.17 vs. 1.36--->1.30 which would show rising inequality under Atlas vs. declining inequality with "their" preferred PPP). Also, I wonder if it is such a good idea to have used '96-2002 which is pretty much 'peak to trough' (only half a business cycle)? I suspect PPP has a bias in the cycle as well as a secular pro-neoliberal bias, and of course inequality changes need more than just a few years to show trends? For example for 1991-2001 (I don't have ready access to 2002 on CD) readily show this.
BUT, it doesn't matter to the larger point: the World Bank and like-minded academics (presenting only PPP) DO try to show that inequality has declined (see my earlier postings for reviews of the lit. e.g. "Contradictory Trends in Global Income: A Tale of Two Biases" by Dorwrick) and, as illustrated even by your numbers above, this can't be supported by a fair use of a wider range of comparison methods. The public should see statistics clearly separated from numerical opinion in government documents.
In summary: PPP is not a statistic based on something that can be measured or sampled. It is a model based on imagining two economies merging. The foundations of that model use extreme free market assumptions and the most exaggerated price index. For popular consumption no other versions are calculated or even hinted at. Then the (predictable) results are used to support those assumptions as policy guidelines.
Hope this helps.
Paul
