The World Bank’s International Comparison Program (Price Indices #4) Transcript

Transcript (PDF)

On December 17th, 2007,[1] PPP GDP per capita figures for 2005 were dramatically revised. India’s fell by 38%. China’s fell by 39%. Singapore’s grew by 39%. And Congo’s[2] soared by 188%![3]

Similarly, on April 29th, 2014[4], the 2011 figures were dramatically revised. India’s grew by 33%. China’s grew by 21%. Singapore’s grew by 20%. And Congo’s grew by 32%.

What explains these dramatic revisions? The answer lies in the ICP.

The International Comparison Program[5] is run by the World Bank, with an alphabet soup of organizations helping out.[6] The ICP’s main purpose is to compare price levels across countries.[7]

The basic methodology is simple: Gather lots of prices. Then crunch the numbers. But of course, it’s not that easy. There are plenty of difficulties, both theoretical and practical.

In Part 1, we mentioned three cost-of-living indices. These are published once or even twice yearly. In contrast, the ICP is a far rarer bird.

Since 1970, there’ve been only eight editions of the ICP.[8] The 12-year gap between the 1993 and 2005 rounds is particularly glaring. Long story short, the 1993 ICP was an epic fail. It sparked a crisis and there were doubts as to whether the program would survive.[9] It would survive, but the next round of the ICP would take place only in 2005.

Before the 2005 ICP, price level comparisons were done using crude extrapolations of old data. On December 17th, 2007, preliminary results from the 2005 ICP were released. It turned out that — oops — the old data were badly wrong and had to be dramatically revised.

For example, India and China were much pricier and thus much poorer than previously thought. Conversely, Singapore and Congo were much cheaper and thus much richer than previously thought.

The revisions were enormous. Overnight, the PPP GDP per capita of 133 countries was revised by an average of 22%. In 54, the revision exceeded 20%.

Instead of loudly alerting the world to these dramatic overnight revisions, the World Bank swept them under the rug. The 2007 publication contained scarcely any hint of these dramatic revisions. Only in one short paragraph was it drily noted that China and India’s shares of the world economy had shrunk.

Worse still, an accompanying publication asserted that “the biggest change [was] for China.” This was an outright falsehood. The 39% change for China was huge. But 24 other countries had bigger changes!

Such shenanigans prompted one critic to charge the World Bank with peddling poverty and being no different from a snake-oil salesman. I have a more charitable explanation: the World Bank simply messed up.

In 2007, the World Bank could argue that the 2005 ICP data were newer, better, and more reliable than older data. It was thus quite the surprise when, on April 29th 2014, it released 2011 ICP results, showing that — oops — the 2005 ICP data were badly wrong. Once again, dramatic revisions had to be made. Overnight, the PPP GDP per capita of 141 countries was revised by an average of 25%. In 60, the revision exceeded 20%.

Once again, the World Bank swept it all under the rug. When one think-tank pointed out that with the new results, “global absolute poverty fell by almost half on Tuesday”, they were promptly flamed by the World Bank’s former Research Director.[10]

Why does the World Bank repeatedly get it so wrong? What exactly were the errors? These are questions we’ll try to address in the next video.


[1] This was the date on which the World Bank released 2005 International Comparison Program Preliminary Results (PDF). Also released was this accompanying “Background document – FAQ” (PDF).

[2] Congo-Brazzaville. Not Congo-Kinshasa.

[3] All data/stats quoted in this video are available at this statistical appendix. Note that on each of these two dates, it was clear that there’d be dramatic “overnight” revisions. However, their precise magnitudes would not be known till later work was done. The “overnight” revisions cited here are based on such later work. For 2005, see “Appendix G (revised)” — date unknown, read below. For 2011, see Inklaar and Rao (2017), Table 8 (dataset on AER website).

The revisions to the 2005 figures furnish yet another example of the World Bank getting things wrong, then sweeping things under the rug. Initially, in the final 2005 ICP report (PDF, published 2008), Appendix G listed the “overnight” revisions and found that these were very small. For example, the revisions to the PPP GDP per capita of India and China were −4% and 0%! However, at some later unknown date, a brief paper titled “Appendix G (revised)” was quietly released (PDF). This new paper revealed that the “correct” (well, who knows) revisions to the PPP GDP per capita of India and China were instead −38% and −39%! In this new document, there was no explanation, apology, or even acknowledgement as to why the original Appendix G had messed up so badly. It was simply stated, drily and without apology, that “This paper is a revision to Appendix G, Global Purchasing Power Parities and Real Expenditures.”

Oh, and by the way, if you go to the World Bank’s official ICP Reports page, you’d simply get the original 2005 ICP report (published 2008), containing the incorrect Appendix G. There’s no hint anywhere that that particular Appendix G is completely wrong. I chanced upon this revised Appendix thanks only to Google (try keywords like “icp”, “’international comparison program’”).

[4] It was around this date that the World Bank released Purchasing Power Parities and Real Expenditures of World Economies: Summary of Results and Findings of the 2011 International Comparison Program (PDF). I got the specific date of April 29th, 2014, based on this blog post “Global Absolute Poverty Fell by Almost Half on Tuesday”, itself dated May 2nd, 2014. (The Tuesday before that date was April 29th.)

[5] Initially called the International Comparison Project.

[6] See “ICP Governance Framework”, November 2016 (PDF).

[7] Actually, the ICP’s headline results are not price level indices (PLIs), but rather PPP conversion factors.

Here’s a quick example of what PPP (purchasing power parity) is. According to the 2011 ICP, the headline PPP for China is 3.506. What this (roughly) means is that 3.506 CNY can buy in China, what one can buy with 1 USD in the US. (Footnote continued on next page.)

To get China’s PLI, simply take its PPP, divide by the market exchange rate, and multiply by 100. So according to the 2011 ICP, the market exchange rate was 1 USD = 6.461 CNY. And so China’s PLI is 3.506 ÷ 6.461 × 100, or about 54.3. Meaning that according to the 2011 ICP, the general price level of China was 54.3% that of the US.

PPP is an unnecessary detour (at least in the context of my videos). What most laypersons are naturally more interested in are the PLIs. And so for pedagogical reasons (and also for the sake of keeping my videos short), I have chosen to completely omit any explanation of what exactly PPP is.

[8] The first five were “phases”. The sixth (1993) onwards, the ICP was conducted in rounds.

[9] See the Ryten Report (1998).

[10] Martin Ravallion was the Research Director at the World Bank from 2007 to 2013. The think-tank was the Center for Global Development and here’s where they pointed out that “global poverty fell by almost half” on a single day.

Do read the comments section, where they get flamed by Ravallion. It’s fabulous. It’s the academic economist’s equivalent of YouTube comments.