Tyranny of numbers

A few weeks ago, in this blog and in opinion pieces (here, here and here), I argued that during the three decades since the end of the war with Iraq (1988), Iran’s economic growth exceeded that of Turkey, such that by 2012, when US sanctions intensified, living standards in the two countries were very similar.  My analysis, which surprised some and angered others, is because of the particular data I used to measure GDP per capita (which I also refer to as the living standard).  GDP comparison is not rocket science but most journalists (and even many economists) often get it wrong.  So, in this post I try to explain why it is important that we use data specifically intended for such comparisons.

The essential choice is between a measure of GDP (or GDP per capita) based on Purchasing Power Parity (PPP) dollars and another based on the official or market exchange rate for the US dollar.  For most developing countries, the difference is large enough to change the comparison living standards, especially of poverty rates.  The difference in GDP values are usually in the order of 2 to 1, with the PPP conversions showing the living standard of a given country relative to the US to be twice as high as when using other exchange rates.  In the case of Iran the difference can be as high 3 to 1.  This difference stems from the fact that a dollar in Iran today (as 15,000 toman to the dollar) buys 2 to 3 times as much as a dollar in the US.  In other words, the cost of living in Iran is much lower than in the US, and the PPP calculations try to correct for the difference.

So it is that according to PPP dollars, in 2010 the average Iranian enjoyed a similar standard of living as the average Turk, and, more significantly, that this living standard was 60 percent higher than it was in 1975, at the peak of the pre-revolution oil boom.  The same comparison done with dollars not corrected for differences in the cost of living show Iran’s GDP per capita to be lower that its pre-revolution level by as much as 30 percent.

To see this comparison for the last 25 years, consider Figures 1 and 2, which are both taken from the website of the World Bank World Development Indicators. The first figure shows the comparison in 2011 PPP dollars and the second in constant US dollars.  The two graphs tell very difference stories, even for the period after 1990, which both share.

Figure 1. GDP per capita in USD PPP for Iran and Turkey

Figure 2. GDP comparisons using constant USD.

For most countries, the difference in PPP and constant US dollar values of the GDP arise in large part because the cost of living, which depends on the cost of labor, is lower in poor countries.  Trade helps equalize the price of traded goods, but non-traded goods in poor countries cannot rise to the level in rich countries because labor is not mobile internationally.  In the case of Iran, the gap is also due to the vast amount of energy subsidies (gasoline, diesel, natural gas), estimated at 20 percent of the GDP, which lower the cost of living in Iran relative to, say, Turkey and the US.  In these countries, prices for electricity, gasoline, diesel, and natural gas are several times higher than in Iran.

A technical note
The Gross Domestic Product (GDP), or its close cousin Gross National Income (GNI), is measured by aggregating the value of all goods and services produced in a country in one year.  Values are simply volumes times the unit price, so the GDP is a number quoted in local currency.

The constant price series corrects for changes in prices over time, which is important.  Statisticians try hard to measure the value of goods and services produced in different years with the same set of prices, which is why they divided the values for later years by the GDP deflator or the Consumer Price Index.

The more difficult problem is to correct for differences in the purchasing power of a dollar in different countries, which arise from differences in the cost of living.  Traded goods, like shoes and TVs, may differ in value but for reasons that are easy to correct, such as sales taxes.  But prices of non-traded goods, like haircuts and doctor’s visits, can differ because of differences in the cost of labor.  Correcting for the latter requires patient accounting in which the same set of prices (usually prices in the US) are used to evaluate production of goods and services in different countries.  So, for example, an inexpensive but decent haircut in my hometown of Blacksburg costs about $10, and the same in my childhood hometown of Neishabour, Iran, costs about 20,000 tomans (each toman is 10 rials).  The exchange rate that equalizes the values of these haircuts is 1000 toman per dollar, or less than one-tenth of the current free market exchange rate. There are as many PPP rates as there are goods and services, and when they are all averaged they yield the PPP conversion factor which WDI reports.

If you got the basic point about PPP’s, you can stop here.

A bit of history
The International Comparison Project ICP) started in 1968 by University of Pennsylvania faculty members Irving Kravis, Robert Summers and Alan Heston. The project was supported by the UN and started with comparison of 10 countries. By the time I joined the faculty at Penn, in 1977, it was already well known, and was responsible for new research on empirical growth.  The project has since moved to the World Bank and University of Gottingen and California at Davis.  It now includes data for close to 200 countries. Nobelist Angus Deaton who has advocated the use of PPP comparisons for a long time whites about the ICP in his wonderful new book, The Great Escape.

There are at least four PPP series for global income comparisons:

• The World Development Indicators, published and frequently updated by the World Bank, are the most readily available.  There are two benchmarks, 2005 and 2011, each recalibrating the series to US prices in a different year.  The 2011 benchmark, which is quite different from the 2005 benchmark but has the support of Nobelist Angus Deaton, goes back to 1990 only.
• The Maddison Project, which has been recently updated for better cross-country comparisons, using multiple benchmarks (calibrating for prices at different points in time). Because of its long historical reach (going back a thousand years), it is used frequently by historians.
• The Penn World Tables, which is currently in its version 9.0 (an update is expected soon), is the source most frequently used by development economists for cross-country regressions to help them understand what factors are correlated with economic growth.

Figure 3 shows that the last two series follow each other rather closely. I prefer to use the Maddison Project data 2018 version because it uses multiple benchmarks and therefore takes better into account the carnations in prices over time. It also has the longest reach.

Both series show a period of robust growth during the 1960s, a sharp decline after the revolution and the war with Iraq, followed by a robust recovery after 1995, thanks to the oil boom of 2000s.  In 2011, when per capita GDP peaked at $18,024 according to both series, living standards were twice as high as in 1975.  As I have noted elsewhere, oil revenues per person were nearly four times as high in 1970s compared to 2011 ($6,393 in the highest point in 1970s vs $1,711 in 2011)

How reporters get it wrong
This post is already too long, but I cannot end it without noting how Iran’s fluctuating exchange rate has resulted in some very crazy reporting on the country’s living standards.  A particularly egregious example is this one from the Centre for Economics and Business Research (CEBR), which claims to provide “clear insight into the world economy and the growth prospects and rankings for each of 193 countries to 2033.” In its latest report, we read that  Iran’s “dollar GDP has fluctuated dramatically – it reached $575 billion in 1990, fell back to $49 billion [sic!] in 1992, gradually recovered to $577 billion in 2011 before falling back to $375 billion in 2015.”

When speculator money and fleeing capital raise the price of foreign currency in Iran, as they did after Trump’s decision to leave the nuclear deal in May 2018, the living standard of Iranians did not fall automatically.  The economy shrank but not nearly at fast as the rial depreciated.

Finally, to see how easy it is for untrained reporters to make big mistakes in reporting economic performance, consider the graph below (left) reproduced from a recent BBC Persian service which uses Iran’s own GDP data to show that Iran has been doing much worse than Turkey.  The graph to the right is the same comparison but with the appropriate data, PPP adjusted GDP per capita from the Maddison Project.  Here, too, we see that Iran falling behind Turkey after the revolution, but this time it is able to catch up (again, thanks in large part to the 2000s oil boom) and actually surpass Turkey briefly, before falling behind again once sanctions kicked in.