Last week saw the release of the new 2011 Purchasing Power Parity (PPP) rates for GDP produced by the International Comparison Program (ICP). The ICP is a major global statistical operation. The Global Office is housed in the World Bank but the ICP is implemented separately in each region by designated regional counterparts.
I have been scratching my head over some of the new numbers, as they are quite surprising. Most surprising of all (at least from my first review) is the number for India. If you believe the new PPPs then India has emerged last week as the world’s third largest economy, with much less poverty than we thought. Both observations have attracted media attention in the last week. But should we believe the new PPP?
The easiest way to think about PPPs is to use the price-level index (PI) given by the ratio of the PPP exchange rate to the market exchange rate. You can think of this as a measure of how cheaply one can live in a country with $US. (The inverse of the PI is often called the real exchange rate.) India’s PI from the prior ICP round, for 2005, was 0.333, i.e., the PPP exchange rate was one third of the market exchange rate. (For the U.S. the PI is unity.) The new PI for 2011 is 0.323. (Note that the 2011 ICP report released last week gives a different figure, but they changed the base to the world, rather than the U.S. My number has the PI for the U.S. as unity for both 2005 and 2011.)
The first thing that surprises me is that the PI for India has not risen nearly as much as I would have expected, given all that economic growth we have heard about—indeed it is surely remarkable that India’s PI has actually fallen. When a developing economy grows as much as India’s has in recent times one expects the PI to rise. That is mainly because (as is the case in India) growth tends to come with higher real wage rates. So goods that are not traded internationally get more expensive. The PPP moves closer to the market exchange rate. (In rich countries the PI is typically unity or higher.) This is hardly surprising; indeed, it is the main reason why we measure PPPs in the first place—to allow for the fact that non-traded goods are cheaper in poorer countries.
That is what we have seen in the past for India; the 1993 PI was 0.20 and the shift from 0.20 to 0.33 in 2005 was consistent with the growth we saw. And we have seen the same thing in most other growing developing countries. For example, the PI for China rose from 0.42 to 0.54 between 2005 and 2011. This relationship between changes in the PI and growth rates is what I dubbed the Dynamic Penn Effect (DPE) in my paper “Price Levels and Economic Growth,” which found this relationship in the data from prior rounds of the ICP for 1985, 1993 and 2005. I have confirmed that the paper’s results also hold with the new ICP round. If India’s PI had moved consistently with the DPE then we would have seen a PI of 0.38 in 2011. You might not think that is a huge gap, but in PPP terms it is large. Indeed, once one deflates GDP by the implied PPP, going from a PI of 0.38 to 0.32 is equivalent to adding a whopping 18% to real GDP.
The second thing about the new India PPP that surprised me is how far out of alignment it is with the domestic inflation rate in India relative to the U.S. If I had simply adjusted the 2005 PPP over time for the differential inflation rates (using GDP deflators) I would have expected a PI for India in 2011 of around 0.43 instead of 0.32. Adjusting solely for inflation this way is how the World Bank’s World Development Indicators updates PPPs between ICP rounds, although it is known that this method does not work as well for updating PPPs as allowing simply for the DPE (as shown in my aforementioned paper). But the point remains: this gap is huge!
In fact there seems to be a puzzling South Asia effect here. All the countries in South Asia have unexpectedly low PI’s in the 2011 ICP. The South Asia effect represents a proportionate decline in the PI of 16% (with a standard error of 2.6%) relative to 2005, controlling for the GDP growth rate.
So what is going on here? It has proved hard in the past for researchers to look inside the “black box” of the PPPs. So far the 2011 ICP is no different. Indeed, we are given almost no information about the underlying micro price data and only minimal information about how the data were collected. (Only quite aggregate price data, for what are called the “basic headings,” were ever released from the 2005 ICP round.)
Absent data, I can only imagine some possible explanations for the puzzle. We know that the PPPs for a number of countries are urban biased in that the price surveys tend to be done in major cities or their suburbs. With Shaohua Chen, I demonstrated this urban bias in the China ICP data for 2005. (The paper can be found here, which also came up with a corrective.) It is also known to be a problem for Latin America and Eastern Europe. Maybe South Asia’s ICP price surveys are biased the opposite way, leading to an underestimation of the PI relative to other countries. South Asia has had a tradition of regular rural price monitoring from local markets for calibrating price indices. Possibly the ICP surveys for 2011 in South Asia did a better job in drawing on this infrastructure, while urban bias remained elsewhere, There are other possibilities, related to the methods used for imputing prices for the stipulated goods that turn out not to be available in local markets. Replacing the missing prices in the field with lower quality substitutes that are available can readily create a bias since the higher quality goods will tend to be missing in poorer places.
These sources of bias are informed conjectures on my part. The point is that we need to know much more about these ICP price surveys. Until then I warn everyone against assuming that India is now the third biggest economy in the world or that poverty is much lower in India than we thought. More research is definitely needed. That research should not be difficult in principle, but it will require greater openness from the ICP’s regional offices and its global office. Publicly funded data on prices and how they were collected should not be a secret.