Are China's economic statistics reliable?
Are China's economic statistics reliable?
6.28 on a scale of 1-10
On a scale of 1–10, where 1 is a “complete Bernie Madoff-style fabrication” and 10 is “100% accurate depiction of reality”, China is probably at around a 6.
At one extreme, you have one set of people who think that every data point in China is made up and completely unreliable and that you cannot trust a thing. This view is wrong and often based on circular logic that is grounded in pre-existing notions. Ironically for an economy as big and complex as China’s it’s probably easier to collect real data than to make it up at this point.
At the other extreme, you have those who treat every number coming out as gospel, are too optimistic or those who merely cherry-pick data to support their point. The mistake this group is making is that they cannot or choose not to spend the time to dig deeper.
On balance I think the truth is somewhere in the middle. Today, I think the process by which the National Bureau of Statistics collects and compiles data is “pretty good” by global standards. However, there are definitely biases in the data collection system and you cannot merely take the numbers at face value either.
Ultimately, I have less of an issue with the reliability of the statistics than with not having enough context to properly interpret the numbers and get to a conclusion that is both meaningful and correct. This is not saying that people cannot come to different conclusions after looking at the same set of data but that a lot of conclusions and opinions I see out there are often based on a very shallow understanding of the complex topic that is China.
A March 2016 article from the Economist discusses one interesting case of data collection bias as it related to reporting by local officials on population data:
A recent paper from America’s National Bureau of Economic Research uses fertility rates as a way to test this theory. Economists have found a relationship between GDP growth in an official’s fiefdom and subsequent promotion, but it is difficult to know how accurate the GDP figures are (a question that haunts anyone following the Chinese economy). Population data are different: in addition to the figures provided by local officials, China conducts a census every ten years, revising population data all the way down to the village level. That makes it possible to pinpoint where bureaucrats have been fiddling the statistics.
Examining data on 967 mayors in 28 provinces from 1985 to 2000, Juan Carlos Suárez Serrato and Xiao Yu Wang of Duke University and Shuang Zhang of University of Colorado, Boulder, find that officials who claimed to have suppressed population growth were rewarded. Mayors who reduced the local birth rate by one child per 1,000 people per year by their own count had a 10% greater chance of being promoted.
But the relationship between fertility rates and career trajectory disappears when using the census data rather than the figures reported by the local officials themselves. Mayors who received promotions were no better or worse at curbing population growth than those who did not. The way to get ahead in the Chinese bureaucracy, it seems, is to falsify statistics. It makes you wonder what other data have been doctored.
Most people know that one of the major policy goals for China during the 1980s was to reduce the population growth rate; hence, the famous One-Child policy etc. The researchers in the above paper found moderate correlation between success in lowering the population growth rate and promotions of local officials. The problem was that once you double-checked the locally reported data with the supervised census data that came out every ten years, the correlations went away. It was a pretty clear signal that local officials were biased in reporting local data.
We see this same dynamic play out with the mismatch between reported provincial level GDP data and national-level GDP data. Basically, if you take the growth rates at a provincial level and try to re-construct national growth rate, the provincial data tends to be much higher. That is because there can be multiple data collection processes going on in parallel. Some people look at this incongruity and conclude that you cannot trust any data, national or provincial. But I would draw a different conclusion: National-level data is more reliable than data calculated at the local/provincial level as there are fewer incentives to skew data.
The national-level data is far from perfect but contrary to what many think, this is not due to outright accounting fraud or keeping two sets of books but for different reasons.
Any economic statistician will rely on a significant amount of estimation. Even in today’s modern digital era, there is simply no way you can calculate everything completely bottoms up. There are large swathes of any economy that are “informal” meaning that there are very few paper trails to measure economic activity.
Standard practice is for large swathes of informal economic activity to be estimated by statisticians according to some common standards. For example, here is an excerpt from an OECD report that discusses how China tabulates its GDP metrics:
Data for China’s quarterly GDP accounting by expenditure approach mainly comes from two channels. One is the statistical survey information, including the overall survey and the sample survey information; the other is the accounting information.
Essentially, these “statistical surveys” are used to form estimates on economic activity. And anytime you have estimates you open yourself up to human interpretation, bias and error. And the less developed an economy is, the more weight the informal economy will have on the overall numbers. As developing countries like China tend to have larger informal economies compared to developed economies, this means that China is going to naturally rely more heavily on surveys and estimates. Which leads me to another key takeaway: As China's economy develops and the formal economy plays a bigger role, the reliability of its economic statistics is improving.
I think the accuracy of China’s economic data collection has changed significantly in the last five years, in no small part due to the widespread proliferation of digital payment platforms like Alipay that can capture a much richer set of transaction data than existing credit card terminal-based systems. Society has been moving along pretty quickly, and your understanding of China from even just a few years ago may no longer be that relevant. I experienced this firsthand when I went back in 2015 to discover that China had gone from a largely cash-based economy to a cashless society in roughly a three-year period.
People often question how China can release GDP numbers so soon after the end of the given period. They joke that this is a sign that the number was already pre-ordained in the last Five-Year Plan. People also question how China’s GDP numbers can be so smooth.
These are all very fair questions and the easy conclusion is that the numbers are simply manipulated or fabrications — i.e. an indictment on the data collection process itself. But I think the reality is once again more nuanced.
First, on how China can release its numbers so soon after the end of the period — the OECD report I referenced above goes on to provide details around the data collection and GDP accounting process:
Process of China’s quarterly GDP accounting by expenditure approach is in three steps: initial accounting, initial check and final verification. The initial accounting is conducted 15 days after each quarter, which relies on the following information: monthly and quarterly statistical information from departments concerned of NBS, import and export statistical information from the Customs, implementation condition information of financial budget expenditures, price index information of Chinese foreign trade, etc.
The initial check is carried out 45 days after each quarter, which revises the data of initial accounting based on the further collected relevant information.
The final verification accords with the statistical yearbooks and the detail information of final accounts of financial expenditures, and utilizes the final verification data of the annual GDP by expenditure approach for the benchmarking adjustment to the quarterly GDP from the initial check through the year, thus forming the final verification data of China’s quarterly GDP by expenditure approach.
So it turns out the first release of GDP figures is merely an estimate, not a final figure. We make estimates too (WSJ: U.S. Third Quarter GDP Revised Up to 3.5% Gain). The numbers will ultimately be confirmed in a three-step process as more data filters in. It can take up to three years for the final numbers to be confirmed, which is why China’s annual GDP statistics only run through 2014 as of February 2017.
As to how China’s GDP figures can be smooth, this is better explained by the level of control the government has over the financial system paired with its desire for stability above all else. Compared to the United States, China’s policymakers have access to more direct policy tools that they can use to stimulate or dampen economic activity. For example, one of the biggest policy tools that the PBOC wields is its ability to control the amount of credit in the system. But while our Federal Reserve also wield similar tools, they are ultimately less impactful due to the Chinese government’s control over the banking system and a large portion of the economy through state-owned enterprises.
So when Chinese policymakers want to generate economic activity, it is pretty easy and the response from the economy is relatively quick. Within months of the collapse of Lehman Brothers, Chinese stimulus money was hitting the real economy in a big way. Meanwhile here in the U.S., even as interest rates have been low for much of the preceding decade, American businesses still have not really loosened the purse strings and significantly increased investments in things like capex, preferring to use their excess cash to do things like share buybacks that generate very little real economic activity.
So the key takeaway here is that ability to exercise control over its economy by Chinese policymakers is the main reason why its GDP figures are smooth — vs. outright data manipulation/fabrication. There is real economic activity underpinning the figures, activity that has been stimulated/dampened by policymakers. But not all all economic activity is good. I will explore this in the next section.
The importance of context
Let’s continue along with the GDP example. So based on my reasoning, the GDP figures are probably fairly reliable. But are they good?
Earlier this month, the National Bureau of Statistics released the latest figures — the Chinese economy grew 6.8% in Q4 and 6.7% for the full year. Some viewed this positively (ABC: China surprises with 6.8pc economic growth) while others viewed it negatively (CNN: China posts weakest annual economic growth in 26 years).
Who is right? And more importantly, should we even care all that much about headline GDP growth?
This is where context becomes important. For example, if you dive one layer down, you will have probably heard the theme about how China is in a long process to shift the primary driver of economic growth from investment-centric activities and exports to consumption-centric activities and the services industry. The implication here is that certain parts of the economy will grow more slowly than other parts of the economy. So you will not be able to get a comprehensive view on this just by looking at the top-line number.
The next-level analysis would be to look at the various component breakdowns of GDP. For example, in the NBS’ press release, we learn that the “Primary Industry” grew at 2.9%, “Secondary Industry” grew at 6.1% and the “Tertiary Industry” grew at 8.3% in Q4 2016. You might need to do a little side research to figure out what these mean, but at a high level, you will eventually figure out that Primary and Secondary is more tied to the “Old Economy” (e.g. agriculture, mining and manufacturing) while Tertiary is more tied to the consumption and service economy. In that sense, the fact that Tertiary grew at a significantly higher rate is probably a good sign that the economy is making progress on transitioning from an investment-centric to consumption-centric one.
Or is it? What if it took a tremendous amount of debt issuance to create the economic activity that ends up getting counted as GDP, with the implication that a fair amount of that incremental debt is not backed by high-quality economic activity. Well, now we have to start looking at the debt side of the ledger and that is even more complicated and technical — perhaps something I will save for a future post.
And once you answer that question, there is probably an even more important question as to whether such short-term noise really matters in the big picture of what ultimately may be a decade-long economic reform program under the Xi Jinping administration.
So even if you've gotten comfortable with the reliability and process by which statistics are calculated, you still need to understand the bigger picture surrounding those numbers. In the case of the GDP numbers, knowing that China grew at 6.8% last quarter tells you nothing about whether this should be viewed positively or negatively. The lesson here is that you need to both dig deeper and evaluate it within a broader context to really figure out what is going on.
This was originally published in Quora in February 2017.