Taking the surprise out of Iran’s low unemployment rate
The fact that after a long hiatus the Statistical Center of Iran (SCI) has decided to publish the results of its quarterly Labor Force Survey should be welcome news but instead it has been met with controversy and disbelief. The new report for summer 1390 (2011) shows a surprisingly low unemployment rate of 11.1%, down from 13.6% the same quarter a year ago. In the absence of data on the Gross Domestic Product (GDP) from the Central Bank in the last 3 years, analysts were hoping to find answers to questions about the economy’s health from unemployment data, but instead they were disappointed. Many reports, including one interview with a former SCI deputy director, dismissed the data as cooked. A closer look shows they are right to be skeptical of the lower unemployment rate, but not to question the survey’s veracity. Based on the published report I argue that the actual rate of unemployment maybe as much as 40% higher than what has been officially reported.
Before I get to the numbers, let me explain why the reported unemployment rate of 11.1% is hard to believe, and why the hasty questioning of the survey itself is misguided.
Although systematic GDP data are still lacking, it is no secret that, if growing at all, the economy is growing at half or a third of the rate it used to grow just a couple of years ago. A decline of 2.5 percentage points in the unemployment rate one year in means that last year about 600,000 more jobs were created than new workers entered the labor market (estimated at about 800,000 per year). That is 1.4 million new jobs. (A similar calculation probably lies behind the government claim of 1.6 million new jobs reported in the news). When the economy was growing at 6% per year, it generated half as many jobs per year. How can it produce 1.4-1.6 million jobs when it is stagnating?
The answer that seems most appealing these days is that the numbers are cooked. It is commonplace for reporters to question Iran’s official data and for experts to claim that the unemployment numbers are twice what the government reports. What is new this time is that the accusation of data manipulation came from a former SCI number 2 official, Mr. Khalil Saidi, claiming that the Ahmadinejad administration routinely questioned unfavorable numbers and on occasion ordered them changed.
This charge comes at a bad time for SCI, having already risked its reputation as an independent data collector (under the watch of the same official) by getting involved in the cash transfer program, which is part of the 2010 subsidy reform. An institution charged to collect data under strict confidentiality rules for the purpose of generating information on average Iranians entered the political fray by helping the government identify individuals to see if they deserved to receive cash transfer.
The accusation of data manipulation ignores how difficult it is to do so. For years now SCI has allowed researchers access to the raw survey data, which has become the reason why those of use who use the data have come to appreciate their quality and attest their credibility (mostly to deaf ears, I might add). If you examine, as I have, consecutive surveys you will find a high degree of consistency. It would take a lot of effort and skill to mess with the raw data in such a way as to produce smooth series for unemployment by, for example, sex, age, and province over the years. I certainly do not know how to do it and neither those who make this charge. As I argued in an earlier post, the unemployment rates for young women could not be much higher than the 50% the labor survey has consistently reported.
Another source of doubts about the accuracy of unemployment data arose is the change of survey instrument to measure unemployment, from the old Employment Survey to the new Labor Force Survey, and with it the change in the criterion for being unemployed, from two days of work last week to one hour only. A recent news report from the BBC Persian service claimed that change in the definition of the unemployed was initiated by the Ahmadinejad government and was politically motivated. It is easy to show that this is nonsense. The changeover from the old to the new survey happened before Mr. Ahmadinejad was installed in office, in summer 2005. The spring and summer rounds of the new survey had been collected before anyone from the new administration had set foot in SCI. The decision to change how labor force data is collected and how unemployment is measured was taken during President Khatami’s administration. The survey was designed and field tested in 2004 while he was still in office. The move to the new survey was encouraged by the International Labor Office in Geneva in order to make Iranian employment data conform to international standards. A top statisticians from ILO helped design the new survey, in close similarity to the labor force survey in neighboring Turkey.
The one-hour cutoff, which seems restrictive and is the source of controversy in Iran is standard practice elsewhere. I doubt that it makes as big a difference as people think, perhaps 1-2 percentage points lower unemployment. In any case, this change has nothing to do with the accuracy of the 2.5 percentage point drop in unemployment last year because both numbers come from the same survey.
Now, to explain the mystery of the good survey producing misleading results: A closer look at the released report shows that, in addition to lower unemployment, the size of the labor force is surprisingly smaller, especially for summer when school leavers start seeking work (SCI only reports the size of the 10+ labor force in its quarterly publications). Or perhaps this is not so surprising once you take into account the fact that the economy has been sluggish for the last three years and there is no end in sight to its sluggishness. The lower labor force is very likely the result of discouraged workers not rapid job creation.
The SCI report for summer 1390 (2011) puts the size of the active labor force at 23.4 million, lower than the same figure reported for spring 1390 (24.4 million). Assuming a small growth rate of 0.5% in the population of the 10+ age groups over one quarter, if the participation rate had not decreased we should have seen a labor force of about 24.6 million for last summer. This leaves a rather large gap of about 1.2 million ‘missing workers’ who could be safely assumed to have been discouraged and therefore in fact unemployed. So, instead of 2.6 million unemployed counted in the survey we could have as many as 3.8 million unemployed, in which case the real unemployment rate would be — deep breath! — as high as 15.2 percent.