Keneth Miles has been posting quite a bit on the Castles critique of the IPCC economic projections used in estimates of global warming. I started a long piece six months ago, but have been too busy to do all I wanted. So, in the best blogging spirit, I’ve decided to post what I have and let the debate go on.

For those who don’t want to read a complex and lengthy post, my conclusion is

No-one can predict with certainty, but the IPCC estimates don’t seem noticeably different from those used in other long-range forecasts. It seems unlikely that they are biased towards overestimation of likely growth in emissions.

Ian Castles is a former Australian Statistician a distinguished member of the Academy of the Social Sciences in Australia. He has been particularly influential in promoting the International Comparisons Project. The idea of the project is to make estimates of living standards more comparable internationally by converting between currencies on a ‘purchasing-power parity’ (PPP) basis rather than on the basis of fluctuating market exchange rates.

Although the goal can never be achieved perfectly, the idea is to find a conversion rate at which an Australian dollar with a given purchasing power in Australia, is converted into a US currency amount with a similar purchasing power in the US (and similarly any bilateral exchange rate. Most estimates of a PPP exchange rate for Australia-US are around 70 cents, while the market exchange rate has fluctuated between 50 and 80 cents over the last decade.
PPP measures provide a better basis for comparison of living standards than do exchange-rate adjusted GDP measures. This is not surprising, since the purpose of GDP statistics is to measure economic activity rather than living standards.

Castles has criticised a number of official publications that have used exchange rate rather than PPP conversions in comparing living standards. His most recent, and best-known critique was applied to the Human Development Report (HDR) of the UN Development Program.

The HDR report used exchange-rate adjusted comparisons of national income per person to argue that global inequality in incomes was large and growing. Castles used PPP measures to argue that HDR had overstated the level of inequality and that inequality was actually declining. UNDP defended the use of exchange rate measures on the basis that they are a measure ofa country’s weight in the global economy. In addition to these central issues, Castles raised a large number of unrelated criticisms of HDR. A panel of eminent statisticians eventually reviewed the dispute and concluded
(i) Castles was right in arguing for the use of PPP measures in comparisons of living standards, and in saying that the HDR overstated the level of inequality (though even Castles measures show great inequality)
(ii) No clear trend in inequality could be determined from the available data.
There was a mixed report on the quality control issues, but the general impression was that by raising them, Castles had tended to cloud the issue and add unnecessary heat to the debate.

Most recently, Castles has criticized the scenarios used by the IPCC in projecting emissions of greenhouse gases. Assessment of Castles’ critique is difficult because:
(i) any projection of economic performance over the next 50 to 100 years is subject to great uncertainty
(ii) the IPCC has put forward a number of different scenarios illustrating possibilities rather than a projection of further income
(ii) Castles’ criticisms are not contained in a single document but in a number of short pieces (a letter to the IPCC and an opinion piece in the Canberra Times)
(iii) as in the case of UNDP, Castles has included a range of criticisms unrelated to his main point about PPP and exchange rate adjustment.
Despite these difficulties there are two major reasons for skepticism abouth the Castles critique.

First, whereas the UNDP report was primarily concerned with assessment of standards of living, the IPCC analysis is designed to provide projections of emissions of greenhouse gases, primarily derived from the burning of carbon-based fuels. PPP measures are designed to produce more accurate estimates of living standards than GDP measures converted on an exchange rate basis. There’s no particular reason to prefer them in estimating consumption of fossil fuels which are internationally traded.

Second, the main burden of the Castles critique is that the IPCC has overestimated the likely growth of poor countries. This is exactly the opposite of the critique he made of the UNDP, namely that both living standards and growth rates for poor countries had been underestimated.

Castles’ objection is that, since exchange rate measures understate the living standards of currently poor countries, they allocate insufficient weight to those countries in computing the average rate of growth for the world as a whole. But this objection, at best, partially cancels out the fact that PPP measures show higher growth for these countries. This can be seen by comparing the IPCC estimates Castles criticises with IMF estimates for the period 1994-2003
Castles’ objection to the IPCC is

If the Maddison estimates of per capita income in each region in 1990 are used to provide ‘correct’ weights in the base year, and the growth rates assumed in the SRES A1 scenario for each region are projected for the 1990-2050 period using these weights, the assumed growth in global average income over this period turns out to be nine-fold rather than five-fold (equivalent to an average annual rate of increase of 3.7 per cent rather than 2.8 per cent for the 60-year period).

Of the IMF estimates, Castles says:

An indication of the possible extent of understatement can be derived by comparing the IMF’s estimate of the average annual rate of growth in world GDP between 1994 and 2003 at market exchange rates (2.8 per cent) with the corresponding estimate of this growth measured on the correct basis using  purchasing power parities (3.6 per cent).

In other words, the IPCC estimate of growth for the future is almost exactly equal to the IMF estimate, whichever weighting basis is used. The switch to PPP weights implies that growth rates for the world as a whole have been higher in the recent past than is normally supposed, and that any plausible measure of average growth in the future will also be higher. It doesn’t have any real effect on the validity of the IPCC projections.

The real substance of Castles criticism is that he expects growth to slow down substantially relative to the period since 1950.. He says

The maintenance of an average rate of growth of 2.8 per cent per annum in the output of goods and services per capita for a 60-year period would in itself be an extraordinary achievement. This is double the average rate of 1.4 per cent annually in per capita output which was achieved during the period 1850-1990 (Angus Maddison, The World Economy: A Millennial Perspective, OECD 2001, pps. 264-65), which was in turn about 20 times the estimated rate of growth in per capita global output during the eight centuries from 1000-1820

But of course, growth since 1950 has been substantially faster than than for the period 1850-1990 as a whole. In estimating likely growth for the period 1990-2050, experience since 1950 seems a lot more relevant than for the period 1850-1950, let alone 1000-1820.

No-one can predict with certainty, but the IPCC estimates don’t seem noticeably different from those used in other long-range forecasts. It seems unlikely that they are biased towards overestimation of likely growth in emissions.