A Chinese worker polishes coiled stainless steel rods at a plant Dongbei Special Steel Group Co., Ltd. in Dalian city, northeast Chinas Liaoning province, 8 February 2014. AAP
The world’s second-largest economy has become the second-most watched and yet investors, politicians and economists are never quite clear what it is they’re looking at. China’s premier, Li Keqiang, is not so sure, either: leaked diplomatic cables have him confiding that all data other than electricity consumption, volume of rail cargo, and amount of loans disbursed should be considered “for reference only”.
The problems with China’s statistical system are manifold. It starts off with the status of China’s statistical authority, the National Bureau of Statistics (NBS). A “work regulation” stipulates that the NBS is to implement “important decisions and instructions of the Chinese Communist Party Central Committee and the State Council”. A former NBS commissioner makes no bones about the NBS’s allegiance: “the government statistics organisation primarily serves the needs of macroeconomic decision-making of Party and government leaders at each administrative level, and is responsible to the Party and government leaders at each administrative level.”
Some data are well known to be problematic. Added up, provincial output routinely exceeds national output. And how can a subset of industry — the “above-norm” industrial enterprises – those with annual sales revenue in excess of 5 million yuan (just short of US$1 million) —- produce more output than all of industry together? The NBS resolves the problem by discontinuing publication of value-added data on the subset.
It also isn’t particularly trust-inspiring that the NBS database, with highly confidential information on all individual above-norm industrial enterprises, is available for purchase at taobao.com, while China’s Statistics Law strictly prohibits the release of statistical data that allows the identification of an individual reporting unit.
How good can the GDP data produced by such a system possibly be? It is worth remembering, though, that measuring the size of the economy is not easy, anywhere. Across OECD countries, official (nominal, or “current-price”) GDP is estimated to under report actual GDP by between 8-30%, with a low of 8-10% for countries such as the US and Japan, and a high of 24-30% for countries such as Greece, Italy and Spain.
One can expect China, a rapidly developing and transitioning economy, to fare no better. Indeed, in the 2004 economic census an attempt was made to include previously ignored economic activities and to improve the quality of the result. If the 1978 data were revised proportionally to the new activities found or included over three censuses, then China’s 1978 GDP would be 34% larger, and China’s average annual real growth rate in the reform period one percentage point lower.
That is the big picture. Numerous specific allegations have been levied against Chinese GDP statistics, but the evidence invariably does not bear out significant and persistent problems. Double-checks more often than not confirm the plausibility of the official data. In other words, if we make a reasonable comparison between the figures for real growth in GDP components against the actual volume of physical output, then those official real GDP growth rates look rather good.
There’s also no huge red flag when you look at the way China incorporates inflation into its growth calculations. The measures used are unpublished but can be implied by looking at official GDP numbers together with the official real GDP growth rates. There are a variety of published price indices against which to compare it, and at either extreme you would see real GDP growth at one percentage point lower or higher. The official implied measure is right in the middle of the two extremes.
In the final instance, one can simply ask the question: given known statistical regularities in the type of data that we are interested in, do the data that come out of the NBS somehow look different, as if they have been tampered with? Applying a statistical technique for ferreting out fraudulent data (Benford’s Law), the Chinese nominal data pass the test. The Chinese real growth rates pass the test. The inflation measure – known as the implicit deflator – shows an abnormal pattern: the second digit is zero a bit too often (as in 3.0 percent inflation). But exactly the same is the case for the comparable US data. The likely culprit? Rounding."
What is left? Quite some uncertainty about the nominal GDP values of the early reform years. They are probably underestimated because the socialist statistical system wasn’t designed to capture small-scale, private economic activities. The nominal GDP values of today are probably quite accurate, with the same caveats as apply to OECD countries.
The real growth rates aren’t too bad, either. They may be off by a percentage point. In the short run, that’s not the end of the world if the economy is growing at 8-10 percent. And as long as real growth rates are overestimated one year and underestimated another year, things might even balance out in the long run.
Fudging the numbers, US style
But have no doubt: the NBS is not a transparent institution. The final data each year are cooked up by a small circle of people, perhaps no more than the Party cell members in the NBS. The temptation to fudge data in order to please one’s employer may be overwhelming. In a dictatorship, it may be irresistible.
What one could easily imagine, for example, is that the NBS in some years massages real growth rates up or down a bit by manipulating the impact of prices. And in later benchmark revisions, when the particular year’s data are out of the spotlight, cleans up its act. It can’t happen too often or the statistical checks on the data would reveal it —- unless, of course, the tampering is done in full consideration of statistical data regularities.
If it’s any consolation, the US appears no better. US inflation went down by a percentage point and real growth up by a percentage point thanks to the Boskin commission. Browsing shadowstats.com, one will find that president Lyndon Johnson would review the gross national product (GNP) reports before their release and if he did not like it, would keep sending the estimates back to the Commerce Department until they got the numbers “correct”. And under the first Bush administration, a senior member of the Executive Branch approached a large computer company with a request to exaggerate computer sales in its report to the Bureau of Economic Analysis: “thanks to the heavy leverage of computer deflation, reported GDP growth enjoyed an artificial spike.”
As China’s premier said in those frank comments in the US diplomatic cables, GDP statistics are “man-made and therefore unreliable.”
Carsten Holz does not work for, consult to, own shares in or receive funding from any company or organisation that would benefit from this article, and has no relevant affiliations.