Here’s Why China’s GDP Numbers are Smoke and Mirrors

Foreign media tends to accept manipulated data at face value

Christopher Balding , July 17, 2015 9:05am

After China released above consensus GDP numbers showing 7 percent growth, many asked questions about the accuracy of Chinese GDP growth data.  Articles were written by The Economist and Financial Times, both publications I respect, essentially arriving at the conclusion that the numbers are good faith estimates, aren’t perfect, can be considered generally accurate, and have the trend right.  I understand this line of thinking but I think this view point fails to grasp the depth of data manipulation that is being thrust upon the world.  There are so many problems with Chinese GDP data that range from the methodological to the manipulative, I will only be able to cover a few aspects today (and this ended up being much longer than I expected for which I apologize).

I will begin with my own research on the matter.  In August 2013, I released a paper that studied Chinese inflation data and its impact on real GDP specifically focusing on housing CPI. Before turning to the results let me note a two key points.  First, this study did nothing more than build a time series of the underlying National Bureau of Statistics China (NBSC) data.  There are no fancy mathematics.  Second, housing CPI is different than the asset price of a home.  Though there are some rather minor and technical disagreements about how best to estimate housing CPI, as will be seen shortly, the discrepancy within the NBSC data is vastly in excess of anything that might be considered a technical disagreement or rounding error.

According to the NBSC, from 2000 to 2011 private housing CPI rose by 8 percent.  Let me emphasize very strongly that is not 8 percent annually but 8 percent total in twelve years.  In a period when official real GDP growth was averaging around 10 percent annually, official housing price inflation was a mere 8 percent total.  To provide some perspective, research covering a similar time period found total real estate asset price inflation of 200-300 percent.  One recent paper by the same team found “real house price growth has been high, averaging 10 percent per annum since 2004.”  It defies any reasonable explanation that home prices increased by such enormous amounts while the inflation attributed amount was so small.

Beyond the obvious data manipulation, there were a few interesting and less detectable ways that inflation was masked.  First, even though in 2000 China had only recently transitioned to a majority urban country, the NBSC gave the data an 80/20 urban/rural weighting.  For instance, total period rural housing CPI was 20 percent while urban was only 6 percent in 2011 with the weighted level at 8 percent.  Over the 2000-2011 time period to arrive at the national average, all years are assigned essentially an 80/20 urban/rural weighting, plus or minus a couple percentage points, to arrive at the national average.  Now we know, and no one disputes, that China only recently became a majority urban country but had a 64 percent rural population in 2000.  This manipulated weighting overweighted the already low urban housing CPI.

Second, according to the NBSC, only 12 percent of Chinese households are renters. When including all categories of private housing as defined by the NBSC, and excluding the undefined “Other” category, a total of 85.4 percent of Chinese households are counted as private housing occupants. This has a very big implication: even though the consumer price of renting increased a total of 53 percent from 2000 to 2011, this price applied to less than 15 percent of Chinese households according to the NBSC. The remaining 85 percent of households are counted as only seeing an 8 percent increase in their price of housing. This then invites another problem.  Given the explosion in apartment buying, asset prices would be much more closely linked to the inflation price.  However, again, this seems not to be a problem for the NBSC.

Third, the NBSC significantly underweights the cost of housing in the total consumer price inflation basket. According to one report, from 2000 through 2010, the NBSC gave only a 13 percent weighting to housing in the CPI basket.   To put this weighting in perspective, it gave a higher weighting to “education and cultural articles” and only a slightly lower weighting to “clothing”.  Most countries and even Chinese data indicate much higher levels of income dedicated to housing than 13 percent of 17 percent.  In 2011, the NBSC reweighted the housing portion to a 17.2 percent despite the fact that it grew significantly slower than all other components of the CPI. In other words, even though housing fell significantly relative to other items in the NBSC basket between 2000 and 2010, it was magically reweighted upward in 2011. The NBSC is implicitly saying its own statistics are unreliable.

Fourth, even if we compare Chinese macro data to other countries we are left with odd inconsistencies. For instance, China experienced some of the highest levels of money growth excluding high inflation states; while numerous countries had comparable or higher nominal GDP, China reports inflation levels over the 2000 to 2011 time period less than the United States and among the lowest of even all developed countries.  All of the Chinese real GDP story is dependent on low inflation data.

This matters for a very simple reason: it understates CPI and overstates real GDP.  Using very straight forward assumptions like using the rental price inflation and other straightforward procedures, this would increase annual inflation during this period by approximately 1 percent.  This subsequently has the impact of reducing 2012 Chinese GDP by approximately $1 trillion. To anyone who has spent anytime in China in recent history the question is simple: do you believe the price of renting an apartment has risen in urban areas by less than 10 percent since 2000?To anyone who has spent anytime in China in recent history the question is simple: do you believe the price of renting an apartment has risen in urban areas by less than 10 percent since 2000?

It cannot be emphasized enough that this is not a rounding error or an accident.  An 80/20 urban rural population weighting happens from an organizational decision to manipulate the data to obtain preferred numbers.  The key point is that in addition to straightforward data manipulation, there are multiple examples of more difficult to detect methodological manipulation that skew the results.

However, it would be a mistake to think this type of behavior has disappeared.  Just this past February a major statistical change was made to GDP numbers that was only uncovered by researchers at McKinsey after numerous people, myself included, expressed surprise at the rapid increase in Q4 2014 consumption.  The NBSC went so far as to note that consumption data in 2014 was not comparable to previous years and this change was found only after inquiries were made by people questioning the data.  Without the methodological change to consumption, 2014 GDP would probably have been about 1 percent lower. Without going into detail here, even the change itself seems incredibly dubious to produce the magnitude of change claimed.  In other words, there are continued examples of statistical manipulation at the NBSC.

Throughout the years, China has been a serial GDP reviser with a non-exhaustive list of changes in 2004, 2008, and 2014 though it is quite possible there are others that are either public or non-public with all revisions making significant upward changes.  Furthermore, the NBSC has admitted that it has not standardized the data so almost any year to year comparison over more than a couple years is worthless.  In fact, even when back to back years are non-comparable due to methodological changes, there is no mention of this anywhere on the NBSC website.

There is also indirect evidence that Chinese data is less than accurate.  Let me give you two examples.

First, Chinese economic data has virtually no variance.  Look at almost any headline number of economic activity from GDP to unemployment or others and you will see essentially a straight line.  Every economy naturally has fluctuations and statistical noise.  Chinese data when compared to almost the entire world, has the lowest measures of economic volatility whether measured yearly or quarterly.  Especially for a fast growing economy, this makes no sense.

Second, China is probably the fastest country in the world to release quarterly and annual GDP.  How many people can even say they have paid their electrical bill for the month of June but China has somehow released national economic statistics for a country of 1.3 billion people, as it has done for many years. A mere 15 days after the quarter, China has such efficient statisticians that they have collected a sampling of data and crunched the numbers?  Makes you wonder why companies can’t release quarterly numbers in a few hours.A mere 15 days after the quarter, China has such efficient statisticians that they have collected a sampling of data and crunched the numbers?  Makes you wonder why companies can’t release quarterly numbers in a few hours.

The problem with deconstructing NBSC data is that it takes lots of work to sort through their math and even now they provide largely only top line data of categories.  A large-scale economic industrial census was leaked (circumstances unclear) of firms in 2007, which became a gold mine for economists starved for data studying the Chinese economy.  However, other than the NBSC essentially saying “trust us”, analysts have little way to verify official data.  By comparison, many other economies, such as Japan, the United States, and Europe, have significant numbers of large observation datasets with significant granularity which, although they may not be the data used by statistical agencies, allow economists to conduct research and replicate official data.

Consequently, many firms have taken to estimating Chinese GDP using some variation of the Li Keqiang Index, but the primary drawback to this method is that you are ultimately relying on government data in some manner.  Even when I note that the electricity consumption growth was 1.1 percent YOY for January to May 2015, I am relying on a self-admitted state organization. The challenge becomes how do you estimate change in the Chinese economy for 1.3 billion people?

When studying the Chinese economy, it isn’t enough to say we believe GDP or other data is wrong, as we need to have comparable data that allows us to say with confidence what a better figure is.  Actually, in the fall, I plan to begin a major research project to better estimate Chinese GDP and inflation.  However, for the moment if we attempt to break free from relying on government data, common sense should tell us that Chinese GDP growth is much weaker than advertised. Let me give you some examples.

Let’s accept for a moment China Electricity Council’s number of 1.1 percent YOY growth in electricity consumption. This number seems to diverge widely from YOY GDP growth of 7 percent.  Put another way, producing 7 percent more as a country and consuming only 1 percent more electricity would require a monumental shift in production techniques or industrial restructuring for a country of 1.3 billion people in one year.  Imports of fuels like coal used for electricity production have collapsed furthering the story that significant problems exist.

Luckily, there is completely independent data that indicates the Chinese economy is much weaker than official data.  For instance, look at trade partners where exports to China have absolutely collapsed across a range of products from consumer to industrial inputs and commodities.  Official trade data with imports declining 15.5 percent, which given import over invoicing implies a larger drop probably closer to 20 percent, is not indicative of robust economy growing at 7 percent.  Not only are prices collapsing due to lack of demand from China but volumes are dropping.  Countries that grow at 7 percent do not reduce imports by 20 percent.

Then look at a wealth of corporate data, whether domestic Chinese companies or foreign multinationals.  Almost universally, they are reporting flat to declining revenue and profits in China.  In fact, corporate profits in China rose 0.6 percent, with 96 percent of that growth from investing in the stock market.  In other words, absent the stock market boom, Chinese firms saw no profit growth.

Then consider the official industrial production growth of nearly 12 percent for the second quarter 2015.  During the second quarter, the HSBC PMI showed constant contraction but somehow official data not only grew but grew well above estimates.  Now the data is not directly comparable but to have such stark differences does invite questions.

Finally, I am always puzzled when I hear people talk about the “weak” Chinese economy when they turn around and talk about 7 percent growth.  The official growth estimate for 2015 has declined from only 7.1 percent to 7.0 percent, but somehow this 0.1 percent has sent commodity prices plummeting, close trading partners like Singapore into recession, caused electricity consumption to flatten, and cars sales to evaporate.  There seems to be an inability to reconcile facts and logic surrounding the Chinese growth story with official GDP numbers.

There is no doubt that the Chinese economy has grown rapidly in recent history.  However, there are enormous discrepancies both within NBSC data and then when comparing it to outside data.  These discrepancies come from both poor quality statistical management but also unquestionably data manipulation.  In other words, the trend is generally accurate but in some cases is also off by such a magnitude as to be considered manipulated.

Li Keqiang made his now famous remarks about GDP growth in the context that the data he received as a provincial leader was unreliable so he focused on measures that were more difficult to manipulate.  The risk here is that Chinese leaders and other interested parties are receiving manipulated data.  While I believe the Chinese leadership does have access to better data, I also believe they face enormous gaps in the quality of data they are receiving in order to make informed decisions about the economy.  Only last month, an initiative was announced to improve labor market data as the official unemployment rate has been nearly unchanged for more than a decade.  If Chinese leaders are telling the world how poor the statistical agencies in China are, imagine the reality.

Christopher Balding

Associate professor at the HSBC Business School of Peking University Graduate School.