To many, China’s announcement of better-than-expected 6.9% GDP growth for the third quarter confirmed a growing conviction. No, not that China’s economy was hardier than it seemed. Rather, that the Chinese government’s GDP data are more or less made up.
Of all of China’s official statistics, GDP is probably among best-known for being, shall we say, internally consistent. But the history of China’s lower-profile economic data is littered with disappearing data, mixed-up methodologies, and freak aberrations. Here’s a roundup of the most notorious examples:
When in doubt, ditch the dataset
Housing is a pretty critical sector, ultimately driving more than a quarter of China’s economy. And while surging home prices signal strong impetus to keep building, when they’re rising faster than incomes, people understandably get upset. It’s easier for housing bubble fears to enter the national consciousness when there’s a national average—or, more specifically, an accurate one.
However, the National Bureau of Statistics (NBS), China‘s official economic record-keeping agency, only publishes the change in average new home prices for each of 70 major cities, including huge cities. It does not release a nationwide figure, nor does it release other figures that allow someone to extrapolate a nationwide average. (Averaging the averages, of course, would yield a figure that did not properly account for the differing number of home sales in each city.)
One online commentator wondered if the NBS hadn’t accidentally placed the decimal point too far to the left.
The NBS used to, though.
But that stopped after many in China suspected the national average was doctored to make home prices look less frothy than they actually were. Public outrage peaked in 2010, when the NBS reported that the national home price index rose just 1.5% for all of 2009. That figure squared with neither the steep climb in prices documented by an independent firm, nor, apparently, with prices prospective Chinese homebuyers were encountering. One online commentator wondered if the NBS hadn’t accidentally placed the decimal point too far to the left (link in Chinese), and actually meant a 15% jump.
In 2011, the NBS announced an overhaul in its home price methodology (link in Chinese), reflecting a much more complete picture of prices in the 70 cities. However, the NBS also stopped publishing the national index, explaining that the nationwide figure smoothed out important regional variation, according to Tom Orlik, a Bloomberg economist, in his book Understanding China’s Economic Indicators.
“That is certainly true, but the national average was also the most straightforward and widely watched measure of developments in China’s housing market,” he wrote. “Cynics concluded that the real aim was to do away with a controversial number that had been the basis of both rumblings of social discontent and questions about the statisticians’ professional integrity.”
There’s an even more brazen example of the NBS erasing a too-good-to-be-true data set, way back in 2006. It involved one of the most keenly watched growth measures, industrial production, Derek Scissors, economist at the American Enterprise Institute, recently pointed out to Quartz.
“Cynics concluded that the real aim was to do away with a controversial number.”
Up until late 2006, the government reported both the raw value of industrial output—from sectors like manufacturing, utilities, and mining—as well as the growth rate adjusted for inflation. The raw value tended to be implausibly high, Scissors said, likely because firms were overstating their output.
Then all of a sudden, at the end of 2006, the NBS stopped publishing the nominal value, releasing only the real, inflation-adjusted, growth rate. Despite having since implemented an expensive, highly detailed survey of industrial activity, the NBS reports to this day only the growth rate of headline industrial production, which is based on an index.
“The problem with indexes,” says Scissors, “is that they’re completely falsifiable.”
Errors of omission
When China’s statistics keepers are less heavy-handed, instead of scrapping a data set altogether, they merely leave gaping holes in it. Here’s one, for instance:
That comes from the People’s Bank of China’s quarterly report on urban residents’ income and price level expectations, as well as their plans for saving, spending, and investing. The survey, which tallies the views of 20,000 banking customers from 50 representative cities, anticipates inflation, wage growth, and spending habits—a slew of pretty useful indicators, in other words.
However, the reports from Q4 2008 and the first three quarters of 2009 simply don’t show up (link in Chinese) on the PBoC’s website, as you can see—just when the global financial crisis hit China and the rest of the world’s economy.
Messed up methodology
However, government statisticians also seem to have subtler ways of masking data trends, such as overhauling a dataset’s methodology.
To be fair, some of these tweaks are probably sincere efforts to improve data quality. However, sometimes it’s hard not to wonder whether officials are trying to hide something.
One case in point is inflation, says Christopher Balding, finance professor at Peking University. The composition and weighting of China’s consumer price index, the key economic indicator used to gauge inflation, has long downplayed the effects of rising home prices, he said.
“According to official statistics, from 2000-2011, when home prices were tripling in most cities, urban housing CPI grew 6%—not 6% annually, but 6% total,” says Balding. And that’s not all. “In recent years as the population has been getting richer, the food weighting has actually gone up.”
In other words, even as people spend more of their increasingly large disposable incomes on goods like cars and clothing, the government has weighted food costs heavier in the CPI index—a weighting that likely muffles the actual inflation rate.
Usually these tweaks seem designed to smooth out big swings; Balding notes that China’s data are among the least volatile of any country’s. Sometimes, however, seismic surges in growth appear out of nowhere—like the El Capitan-esque pattern of highway freight traffic growth in 2009.
Anne Stevenson-Yang, head of J Capital Research suspects these occur when the government statisticians need to flatten out the drop in an important headline statistic that has been dragged down by a closely watched data point. In the chart above, for example, that phenomenal growth in highway activity offset a sharp decline in railway figures, keeping overall growth in commercial freight aloft.
Why didn’t China’s record-keepers just fudge the railway freight data (see messed-up methodology, above) and call it a day?
Because it is closely watched, probably. And things like electricity production and railway cargo can be cross-checked and independently verified through comparison with other data. For example, to reach their own conclusions about electric power demand, independent research firms have been known to count China’s coal-hauling trains or deliveries of coal to power stations.
When these higher-profile figures can only be massaged so much, you tend to see “sudden and mysterious growth” of harder-to-verify substitute indicators—e.g. highway freight—that serendipitously keep headline growth rates aloft, says Stevenson-Yang.
Highway freight isn’t the only example. Behold the periodic leaps in hydropower growth that have prevented dropping coal plant output from scuttling overall electricity production.
A general disdain for transparency
Then there’s the problem of shuffling how data are classified, and where these metrics are reported—which happens particularly at one of China’s most important sources of economic information.
“My greatest frustration is with the highly untransparent and randomly changing balance sheet data” at China’s central bank, says Victor Shih, professor of China’s political economy at the University of California, San Diego.
More and more of the People’s Bank of China’s liquidity operations, the money it pumps into China’s economy to keep the banking system churning, are turning up in vague categories like “other assets” and “other liabilities,” he says.
These items might offer important clues to the state of the PBoC’s finances—such as how much money it is spending trying to keep the yuan’s value stable. What actually makes up an item called “other foreign assets” is completely unknown, says Shih, but it’s possible that the central bank is using it to mask movements in its foreign exchange reserves.
“To me, it is galling that a country whose currency may become a global reserve currency is so vague about its central bank balance sheet,” he says.
The old bait-and-switch
China’s wildly swinging stock market offered up a new example of China’s commitment to data manipulation. For many years, a key indicator of investor interest has been the number of new stock-trading accounts opened, which China’s state-owned market clearinghouse, China Securities Depository and Clearing Corp (CSDC), reported on a weekly and monthly basis.
Then in June 2015, it stopped reporting account numbers altogether.
In its place, the CSDC began reporting the new number of investors—a metric it had only introduced in weekly reports in May 2015 and monthly reports in April 2015.
The CSDC told Quartz that the reason for the change was a new April 13 regulation that allowed investors to open up to 20 stock accounts of virtually any type (before they had only been allowed to open more than one account for specific types of investment), which made new account data less valuable.
However, the reason CSDC gives don’t make much sense once you compare the brief period of time where they reported both “investors” and “accounts” data. As you can see, the mid-April rule change caused no market increase in the average number of accounts per investor.
Regardless, declining to include both metrics makes it impossible to compare how many retail investors are active in China’s stock markets since May 2015 with historical trends.
Finally, there’s China’s freakishly flat unemployment data. Decade after decade, through ups (entry into the World Trade Organization) and downs (the Asian financial crisis, state enterprise restructuring, and the latest global financial crisis), its surreal calm persists.
Since unemployment gives rise to social tumult, this might well be the most sensitive statistic of all—even more so than GDP. That would explain why the official data are kept both low and eerily level. (Also, the rate only reflects urban employment, eliding the fortunes of China’s the nearly 275 million rural migrant workers, even though they make up as much as half of the urban workforce.)
What’s baffling, though, is that the unemployment rate isn’t totally flat. Here’s a zoomed-in look at the trend since 2002:
These tiny movements hint that employment statisticians may have been trying to reflect at least some change in labor market realities, while still keeping the overall trend within a politically acceptable range.
The all-importance of maintaining a soothingly stable appearance shows up in most the other examples too. That might signal an articulated policy of top-down data obfuscation. Yet the sheer range of statistics tricks listed above—and they aren’t the only ones—also hints at creative ad hoc fudging perhaps done by China’s statisticians out of self-preservation.
Whatever the case, the system that compels the data-doctoring seems to also thwart corrections. As AEI’s Scissors observes, when a Chinese statistician takes over a new post, the odds are very good that the dataset he’s managing is wrong.
“Do you fix them? Do you renounce the old figures, which is politically very difficult? Do you fix 2014, but then 2013 is all wrong?” he says. “This is not an easy problem to fix.”