Government statistics should be open to scholarly scrutiny: a deep data dive with Xu Xianchun
Three professors discuss the purpose and nature of Chinese government statistics at book reading
On Aug. 15, 2023, the decision by the National Bureau of Statistics (NBS) of China to suspend the release of youth unemployment rates caught many by surprise. The NBS spokesperson's explanation for this decision, which cited differing public opinions on the scope and methodology of statistical calculations, was met with sharp skepticism and stinging cynicism.
This development has reignited discussions about the government's transparency in statistical reporting. According to China's leading statisticians, while the NBS has made significant efforts to enhance the quality of statistical data, there remains a crucial need for China to draw lessons from the statistical systems of many Western governments. These systems often feature vibrant and open debates between the government and scholars.
Statistical data is foundational for effective and sound economic choices, and the government should either make corrections or provide explanations when official figures are in doubt. The discrepancy between official figures and public perception is not uncommon, usually owing to people’s predominant exposure to specific sectors of the economy. This highlights the importance of providing statistical breakdowns that can help mitigate personal biases and boost public acceptance. It is essential for statistical bureaus to tune into the actual needs of the public and the market, and remain responsive by offering clarity on statistical indicators.
The in-depth discussion mentioned above can be found in a more detailed format at a book reading event co-hosted by the National School of Development at Peking University and Social Science Academic Press. Actually, The event took place on the evening of Jun. 18, 2023, nearly two months before the controversial NBS decision. In today’s piece, we are pleased to bring you a full-text translation of the book reading discussions.
The book in focus was 透视中国政府统计数据：理解与应用 A Probe into China’s Official Statistics: Comprehension and Utilization, also featured on The East is Read in the Aug. 14 post.
The author of the book, Xu Xianchun, Distinguished Fellow at National School of Development of Peking University, was invited to deliver the keynote speech. He then talked with Wu Xiaoying, Research Professor of Economics at National School of Development and head of the Growth Lab, and Huang Yiping, the Deputy Dean of the National School of Development and Director of the Institute of Digital Finance. The following is the dialogue of the three professors.
Wu Xiaoying:
I want to start with a story from the mid-1990s, when I wrote a systematic argument questioning the industrial growth rate released by the NBS. In 1996, Professor Xu, who was then working at the Department of National Accounts of the NBS, translated my piece and circulated it within the NBS. He even introduced me to his colleagues from various departments when I later arrived in Beijing. Despite my sharp critique of the official statistics, Professor Xu and his colleagues treated me with great openness and honesty. This marked the beginning of my enduring friendship with Professor Xu. It is his extensive expertise in China's national economic accounting and insights into practical issues that deepened my understanding of China's statistical data.
To highlight Professor Xu's influence, let me recount an experience involving translation of Angus Maddison's book, "The World Economy: A Millennial Perspective." Maddison sought my advice in the choice of translators, and Professor Xu Xianchun was on my suggested list. Maddison instantly chose Xu. “I like him!” he said. When I asked him why, he said China was such a large country with such complex economic situations, and it was a daunting challenge to transform an outdated national statistical system designed for a planned economy and under Soviet influences to align with international standards and integrate into the international market and society. But Xu and his team accepted this challenge and worked head-on. Plus, translating Maddison’s book required a comprehensive understanding of GDP-centered statistical intricacies, making Professor Xu an ideal translator.
As you all know, later, I too joined the translation project. Maddison came to China especially for the book launch and was keen to visit Hangzhou, which used to be the capital of the Southern Song Empire, in the most prosperous era of Chinese history. Xu, who was engaged in other events in a different city, caught up later. We all spent a wonderful time together at West Lake and "Songcheng" (Song Dynasty Town).
Now, focusing on tonight's discussion about Professor Xu's book, it dives deep into the notion of government statistics as a public good, which might sound dry to many. If you're genuinely interested, you probably depend on these statistics professionally or are frequently puzzled by the national statistics' shortcomings. But seen through another lens, many more people can become interested in the topic because the challenges faced by China's transitioning economy touch everyone. The Chinese economy is vast and complex, which not only makes its transition more difficult but also makes issues within this process more intricate, hence an ever-growing need for reliable statistical data and indicators based on such data. This would not only help policymakers better grasp the nature and extent of the issues, align themselves with international counterparts, and draw up more targeted policies, but also inform the public or every economic entity to make prompt adjustments to address the potential impacts of economic transition.
To be more precise, economic transformation essentially revolves around increasing productivity to offset rising costs. This isn't just about technological innovations; it's primarily about institutional restructuring, both of which carry significant price implications. Understanding such issues requires accurate input-output statistics and corresponding price statistics; it's also essential to have accurate statistics on the primary factors, namely the flows, stocks and the corresponding price statistics of capital and labor. Economists have various theories and methods on how to understand and analyze economic transition through these indicators, but if the data they rely on is flawed, all work will be to no avail.
To enhance the quality of national statistical data, I'd like to emphasize two key points, speaking from years of global experience:
National statistical data are public goods. We need to popularize a somewhat forgotten common understanding that statistical data collected, compiled, edited, and published by the government are not solely for government use but rather for public service. Unlike privately-owned products that serve private interests, these statistics are completely public goods. Therefore, they must be "interest-neutral," devoid of biases towards any country, government, individual, or group. While national statistical agencies produce and manage data with public resources, they are not and should not be the exclusive owners of statistical data. Governmental bodies utilize and interpret statistical data for either the interest of the nation (the public) or the interest of their own power (the regime). Similarly, individuals, institutions, and groups can also fully utilize and interpret statistical data for their own benefits. Both approaches align with the nature of statistical data as public goods.
However, while we expect central governments to uphold data impartiality, it's not rare for local authorities to skew figures, like GDP growth rates, for political gains. Such distortion not only misdirects macroeconomic strategies of the government but also misleads private investors, resulting in resource misallocation. By reiterating that such data is a public good, we also encourage public scrutiny. This compels national statistical agencies to unequivocally and transparently share their methods for constructing and verifying statistical data and, when necessary, to respond to major public doubts through independent oversight and reexamination (such as sampling).
Under these principles, that is, within the framework of the United Nations System of National Accounts (SNA), statistical data of every country, being public goods, should be made accessible for the collective benefit of humankind. It shouldn't be limited to the interests of individual nations or blocs. Given today's multifaceted geopolitical scenario where countries fail to agree on rebuilding a rule-based international order, this principle is indeed paramount.
Openness and competition are necessary to improve statistical data quality. This connection might seem obscure or irrelevant to our discussion, but it's not the case. Both public and private sectors recognize that competition enhances product quality. Monopolies in public services often lead to inefficiencies, subpar quality, and inflated costs, hence the competitive public procurement policies in numerous government departments. In the context of governmental statistical data, it’s crucial to make the process of data generation and quality testing fully transparent to professional users; and the pivotal step is to welcome skepticism and recommendations from scholars at academic institutions and other government agencies. Scholars can then access a greater amount of fundamental data in their studies and communicate with statistical personnel of the NBS on the issues they discover.
Formal exchanges should take place at public academic seminars, which offers manifold advantages. It compels statistical entities to publicly address critiques and seek improvements; it also encourages scholars to propose alternative methods; furthermore, it spurs academic competition, fosters precision in identifying issues, refines existing statistical indicators, and potentially devises new ones.
We should learn from the statistical system in many Western governments, for instance, the U.S. which houses five independent statistical agencies, each having fairly sharp critique of another. The U.S. statistical agencies maintain regular interactions with academic institutions, but also frequently challenge academic viewpoints. These dynamics contribute to improving the quality of statistical data. Prominent international societies on economic statistics and macroeconomic analysis, such as the International Association for Research in Income and Wealth (IARIW), the International Input-Output Association (IIOA), World KLEMS (KLEMS), and the Society for Economic Measurement (SEM) further exemplify this collaborative yet competitive approach. Renowned economists like Simon Kuznets, Wassily Leontief, Dale Weldeau Jorgenson, etc. founders of these societies, organized rigorous debates within these circles. The annual gatherings of these societies see a blend of researchers from global statistical agencies and top-tier academic scholars in different thematic sessions, whose competitive exploration has led to several major revisions of the SNA. Instead of complaining that the West dominates the rules, why can't we proactively participate in the rule-setting process?
Huang Yiping:
Professor Xu's book is very professional, and while not everyone here might grasp its intricacies, this level of expertise is what we truly need now. I recall Professor Wu Xiaoying's stance as a challenger of national statistical data. Roughly 30 years back, he presented a lecture at the university where I worked. By that time, he had authored an intriguing piece, "How Wealthy is China Really?" [“How Fast Has Chinese Industry Grown?”], dissecting China's economic growth rate. It was very interesting. His research was triggered by skepticism of the official figures and gained notable international attention. So I've always admired the research conducted by Professor Xu and Professor Wu from different perspectives, including Professor Wu’s collaborative work with Maddison on China's long-term economic figures.
I believe statistical data is indeed extremely important. Having worked across government sectors, academia, and the market, I am often asked about the differences in my work in different settings. While the work is somewhat similar, the types of data needed and perspectives of statistical analysis do differ.
It is often said that Professor Xu is the producer of statistical data and we are the consumers. This highlights the indispensable role of statistics in our professions. However, China's statistical data, especially the earlier records, is indeed in doubt. Foreign investors and experts often question the reliability of the data, and we, while engaging with them, have to spend loads of time explaining and validating its authenticity.
This gives rise to two issues: the imperative for reliable data and the necessity to enhance data quality. Many may not be aware of the efforts statistical departments have taken to enhance data quality. That is why books published by scholars like Professor Xu are crucial as they provide us with more information. Robust economic and statistical data are foundational for effective and sound economic choices.
Consider macroeconomic strategies, that is, fiscal and monetary policies, which gradually took shape after the Great Depression. Central banks worldwide often make the news for interest rate decisions. Essentially, their primary aim was to navigate economic cycles. But to steer these cycles, they must first discern the current economic trajectory – whether a downturn or an upswing -- as well as the pace of development, all based on high-quality economic data. Without clarity of data, informed economic decisions will become impossible.
Official statistical data in China used to face a lot of skepticism. Some was reasonable, some was undue. For most people, statistical data is akin to a black box, making it difficult to dissect it in the meticulous manner like Professor Xu and ultimately ascertain the accuracy and reliability of this data.
The reason I think Professor Xu's book is important is that I myself have benefited immensely from it. Statistical data is an integral part of many economic activities and decision-making we engage in today. I only mentioned examples related to the government, but it is equally crucial for every company and every investor to know the price trends and the speed of price changes, without which decision-making becomes precarious. This is my overall understanding that I wanted to share with you all.
Next, I would like to share three post-reading thoughts:
The NBS has made tremendous efforts to improve the quality of statistical data. Some foreign experts or investors often criticize our statistical data, especially regarding the GDP's seemingly stable year-on-year growth. While on the one hand, this could highlight the efficacy of China's policies in maintaining economic stability, skeptics often argue that the data is not fact-based and has been manipulated. This skepticism isn't baseless – manipulation of data did happen in the past. However, Professor Xu's book (I especially recommend the first chapter), discusses the significant work that has been done to improve the quality of statistical data since 2000, mainly in four areas: 1) conducting economic censuses, 2) establishing an online direct reporting system for enterprises, 3) conducting statistical surveys of the service industry, and 4) household surveys. Professor Xu's book provides detailed introductions, so I won't elaborate further.
Among these, online direct reporting is a key move in circumventing possible intervention by local officials, which is the main reason for untruthful data. Research shows China’s GDP-related promotion system incentivized local officials to manipulate statistical data. But ever since the NBS mandated that all companies report directly to the NBS, bypassing local governments, the manipulation problem has seen significant improvement. Many scholars and investors around the world have felt this improvement in data quality over the years. Of course, no country can claim absolute accuracy of its data due to myriad technical nuances, but the efforts made by the NBS in data enhancement should be acknowledged and the results appreciated.
As someone who uses statistical data on a regular basis, I sometimes find a significant discrepancy between official figures and our intuitive perceptions. Taking the Consumer Price Index (CPI) as an example, many would find that the prices of vegetables, meat, or rice have gone up by 10%, 20%, or even higher; yet still, the official statistics only show an increase of 3%.
This discrepancy isn't necessarily indicative of deliberate data manipulation, as many people conjecture, but could stem from individual biases. The public's predominant exposure to soaring prices of daily necessities like food and energy can distort their perception, whereas in fact, the CPI is based on a large basket of goods. People might be oblivious to price drops in products they buy less frequently, such as household appliances; nor are they aware of the actual extent of price decrease. This example is to illustrate that despite differences between personal perceptions and the statistical data, the official figures are not necessarily faulty.
When I worked outside the Chinese state apparatus, I observed varying viewpoints among different types of analysts in investment banking. Economic analysts were responsible for macro analysis, strategy analysts for microanalysis, and stock analysts for individual stock analysis. In most cases, micro analysts were more pessimistic than macro analysts, probably because they focus more on company details, which might lead them to a more pessimistic view of the entire market. That is to say, variance in feelings and outlooks often stem from differences in research methods and analytical perspectives.
To give another example, if someone wanted to understand the recent economic situation, they might choose talk to people in the Haidian District, Beijing. But if they only converse with those in a poor economic state, they will most likely hear complaints and concerns. It's not that they're intentionally skewing the statistics, but these individuals might have a stronger motivation to voice their grievances. In contrast, those in a better economic condition might stay silent because they lack strong motivation to express. Therefore, it is important to know those strongly-felt perspectives might not represent the entire market. It's necessary to form our own judgments and decisions by juxtaposing various sources and viewpoints.
Looking at the broader landscape of statistical data, it's crucial to tune into what the public and the market really need. Statistics bureaus should ramp up their communication efforts with both groups to clarify data and indicators. The discrepancy between highly technical and comprehensive indices and public perception is an indicator that the benefits of these indices have not been fully exploited. Therefore, it is necessary to engage more with the public and the market to better understand their needs and points of focus.
There are methods other than competition for the statistical bureaus to improve their engagement with the market, such as offering context and explanations when data is released to help with public understanding. To illustrate, if official statistics show a 3% rise in consumer prices, an accompanying breakdown could reveal that while food prices spiked by 10%, appliance costs decreased by 5%, hence the overall 3% increase. Even if such insights don't resonate with everyone, they can still significantly boost public comprehension and acceptance of data.
In conclusion, I believe the enhancement of data quality is going to be a long-term task due to myriad factors on the national level. But two primary factors stand out:
The first factor is tied to reforms. Determined efforts need to be made to drive valuable reforms. Some local government officials harbor biases and preferences regarding statistical data. I've visited several enterprises for survey, and a few insiders, including experts, have hinted at the government’s reluctance to highlight less-than-rosy economic news, often for fear of tarnishing their image or "performance measurement." Despite the Statistics Law of the People's Republic of China being in place, government officials still intervene, directly or indirectly, out of consideration for their promotional prospects or other factors, thereby resulting in data inaccuracies. If such circumstances prevail, how can we gradually mitigate them? I consider this to be a long-term endeavor. Facilitating an online direct reporting system for businesses could be an effective approach, yet it might not fully resolve the issue. While I'm not an expert in statistics, there certainly remains further work to be done in this sector.
The second factor is related to industrial development. Sometimes, adjustments are made to economic growth and GDP statistics that have already been released. I recall two instances where GDP statistics were adjusted, each time registering an increase. This was because previous statistical methods focused more on tangible production, while statistics for the service sector and emerging industries lagged behind. Upon rectifying the statistics, it was found that certain activities had not been included. The thriving development of the digital economy, for instance, calls for improvement in statistical practices. Statistical data holds vital significance for the government, businesses, and individuals alike, and there is always room for enhancing its quality.
There is much work to be done in the above two dimensions. [Former Deputy] Director-General Xu has undertaken extensive efforts in this regard. His new book also contributes to the improvement of statistical work, as more individuals, through careful study of his book, can gain a comprehensive understanding of how data is generated and why some seemingly unreliable data is logically sound. Progress in demand often propels progress in supply.
Xu Xianchun:
I'm deeply grateful for the insights provided by Professor Wu and Professor Huang. As Professor Wu mentioned, we've had a 25-year interaction. He identifies himself as a challenger of government statistics, and even though I have had a long-term career in the government statistical department, I don't think there are problems between us. In reality, I don't agree with all of Professor Wu's views, but I greatly appreciate his dedicated approach as a scholar. I'm also open to engaging with various challengers, including scholars who question China's government statistical data, because I believe in the principle of gaining insights from various perspectives.
Typically, I refrain from commenting on whether an individual's viewpoint is correct and instead put it within an overarching category for more collective interpretation. When scholars raise doubts, I carefully consider whether these doubts are valid. If they are, then efforts should be made to improve statistical work accordingly; if the doubts prove to be invalid, I reflect on where the issue lies and how explanations can be provided to help clarify the situation. I believe this approach is highly beneficial for enhancing government statistical work. If everyone publicly praises statistical work but privately criticizes the accuracy of the data, that would be unfortunate. I think different voices should be earnestly listened to to identify areas of concern and make necessary improvements.
While working at the NBS, I organized the writing and publication of a series of manuals for national economic accounting procedures. Some colleagues were concerned about scholars criticizing the manuals after publication. I reassured them that there was no need to worry as long as we worked diligently and put forth our best efforts. I also told my colleagues that it was a good thing for scholars to point out issues, as it would help us improve our work. Therefore, I never rejected criticism from scholars, including sharp critiques from Professor Wu. I apologize for not having read the article questioning China’s long-term economic figures mentioned by Professor Huang; if I had seen it, I would have seriously studied it.
Both professors mentioned a viewpoint I strongly agree with: the significance of government statistics. Whether it's for analyzing economic and social development trends, policy formulation, or academic research, government statistical data holds immense significance. Without them, many academic and policy studies would face significant limitations and lack of foundation for decision-making.
I also highly commend the recognition from both professors regarding the statistical reform efforts undertaken by government statistical departments. The NBS has implemented a series of reforms to improve the quality of statistical data. For instance, as Professor Huang mentioned, the reform that put in place the online reporting system for enterprises. The NBS adopted such a system for surveys on companies above the designated size (with a revenue of more than RMB 20 million per annum), qualified construction industry enterprises, wholesale enterprises above the designated size (with a revenue of more than RMB 20 million per annum), retail enterprises above the designated size (with a revenue of more than RMB 5 million per annum), accommodation and catering enterprises above the designated size (with a revenue of more than RMB 2 million per annum), and real estate development enterprises.
Through the online reporting system, survey forms are sent directly to enterprises, who fill them out and submit them directly back to the NBS. Regional statistical departments can access data from enterprises in their own regions but not data from other regions. If data discrepancies are found, regional statistical departments cannot alter them; corrections must be made at the enterprise level. If enterprises find errors in their reported data, they can make changes, but no intermediary steps can alter the data. This kind of online direct reporting method can, to a certain extent, curb the interference in statistical data by intermediate parties, and I believe it is an effective way to improve the quality of statistical data.
Regarding the accuracy of statistical data, as Professor Huang mentioned, no one can guarantee 100% accuracy. Statistical processes are complex, and various factors influence data accuracy, making it indeed challenging to ensure complete accuracy in statistics.
There are numerous reasons why government statistical data faces skepticism, one of which is the differing perceptions of the same data among different groups. Allow me to share a story. In 2003, I accompanied the then Commissioner of the NBS on a visit to the Italian National Institute of Statistics. The head of the institute asked us, "Do you know why people are protesting on the streets?" We said we didn't know. The director then informed us that they had just released economic growth figures for Italy, indicating a return to growth. But the protesters were saying, "How can the economy be growing when we don't even have jobs?" They believed the statistics were fabricated. On the other hand, the Ministry of Economy and Finance of Italy was complaining, "The economy has long been in growth; the statistical data is outdated." This illustrates how different groups perceive the same statistical data differently. Therefore, certain statistics, whether released by China's NBS or other countries' statistical departments, are prone to skepticism.
Professor Huang mentioned significant price increases for certain products like grain, pork, vegetables, and fruit, yet the CPI increase isn't proportionately high. This is because while some prices rise, others fall. People tend to notice price increases more readily but might overlook price decreases. Consequently, there can be a disparity between public perceptions and CPI increase figures.
Another example is the common assertion that the NBS’ reported per capita disposable income for residents is too high. Per capita disposable income is easily influenced by high-income households, as the income of one high-income household could be many times that of a middle- to low-income household. Whether due to the expansion of high-income groups or an increase in the income of pre-existing high-income groups, the per capita disposable income of residents could be inflated. This might make some low-income groups feel their income is being "averaged out." Therefore, the average per capita disposable income does have certain limitations. To address this, the NBS also publishes the median disposable income, which is arranged by ascending order of income and is unaffected by high-income households. Hence, there can be a discrepancy between per capita disposable income and the perceptions of certain groups, but this doesn't necessarily mean the statistical department is falsifying data.
The above represents my personal views, which might not be entirely correct. I invite criticism and correction from all of you. As a statistician, I believe in maintaining an open-minded attitude and being willing to face challenges. If the challengers' viewpoints are correct, we should promptly make corrections; if not, we need to provide clear explanations.
I'm truly grateful for the objective evaluations and support from Professors Wu and Huang. This also serves as an encouragement for my future endeavors in statistical teaching and research.
The East is Read posted an interview with Xu Xianchun, author of A Probe into China’s Official Statistics: Comprehension and Utilization:
Other posts on The East is Read:
Check out Pekingnology too:
