Abstract: In this paper, the author analyses the challenges facing data governance including the determination of data rights, and how to strike a balance between rights protection and value creation. He argues that data governance requires new thinking and should focus on supporting innovation and regulating behavior, with a particular emphasis on common prosperity.
Data governance is a new yet rather important topic. We have run into many problems, and there will be many more in the future. Here, I have three reflections on the issue.
I. The innovation and challenges brought about by big data
Big data plays a huge role in the fourth industrial revolution and functions as a critical driving force of the development of digital economy. There are already many successful cases of big data application like health QR code and credit risk assessment based on big data.
If applied properly, big data could bring huge economic benefits in five areas, i.e., increasing scale, boosting efficiency, improving user experience, reducing cost and controlling risk. If not, it will have serious negative impacts.
These five positive impacts might change some of the laws of economic operation. Take the 80-20 rule as an example, with the support of big data analysis, the boundary between the 20% and the 80% have become blurred, and financial institutions could provide services for the mass 80% at low cost and high efficiency. Since the marginal cost of tech platforms is very low, sometimes even close to zero, the law of diminishing returns to input might be changed because of the application of big data and big tech platforms. These changes might revolutionize the economy and finance.
The Chinese regulator has defined data as a new type of production factor. In economics, according to the production function, output is determined by several production factors which traditionally include land, capital and labor. If data is added to the function, the resulting influence is worth studying. It might change the marginal return of each factor as well as economy of scale, which could lead to changes in some fundamental features of the production function.
For latecomer countries, data might present a new opportunity for leapfrogging. The amount of traditional production factors like land, capital and labor can hardly change in the short term. Some factors like labor and capital need time to accumulate, and will only reach a certain level when the economy develops to a certain stage. But data, if well collected and analyzed, might be a factor which latecomers could use to catch up with and leapfrog advanced countries. It may even bring changes to the development pattern. In daily life, there are many cases in point.
But there are major differences between data and traditional factors, e.g. the clarification of rights. The clarification of data rights is complicated since data as a factor of production is not scarce and the use of it is not exclusive, which might be a good thing but will also make transactions and pricing difficult.
The second difference between data and traditional factors lies in allocation. Land, labor and capital can be allocated in the same scenario. Land is fixed, but labor and capital can move to any place and form new production units. As for data, some can be easily allocated but some are not since they are useful in a certain place, but useless in another place. Data governance needs to solve a series of problems, including rights clarification, transaction, allocation, pricing, usage, etc.
Proper use of data might lead to revolutionary changes in economy. However, it is not easy to do. Despite some successful cases, there are many complex issues to be solved.
II. Balance between rights protection and value creation
Data application involves the collection and analysis of data. The key issue here is how to strike a balance between rights protection and value creation. Good governance can help protect rights, break data silos, and help realize data sharing, reasonable pricing and sound allocation, thus creating maximum economic benefits. In practice, however, there might be many difficulties. A pragmatic strategy should be adopted in order to balance between privacy protection / data security and data sharing / value creation.
First, balance between security and innovation. Data protection in a broader sense includes the protection of national security and personal privacy, and varies greatly among regions around the world. Europe has done the best in this respect, but as a result, it does not have any particularly successful tech platforms or digital economy; China’s digital economy faced little regulatory constraints in its initial stage of development, which allowed the emergence of several industries in this sector; the US is somewhere in between, but we cannot tell whether it strikes the best balance.
This picture shows that we must strengthen data protection but should not overprotect like Europe has done, otherwise digital economy could not grow. Data protection should not go to extremes and should differentiate between various types of data. Some data that concern “private rights” can be strictly regulated, while other data can be regulated less strictly. Since the ultimate goal is to protect both data rights and data security while leveraging the power of big data analysis, any extreme measures would not be the optimal plan.
Second, balance between data sharing and efficiency. Big data is not small isolated units of information. Most of the “big data” that we see and discuss toady is not big data in real sense, but rather data from a particular ecosystem or platform. Data can only create value when further integrated and analyzed, which is why we must break data barriers and achieve data sharing.
In practice, some things are easy to do but some are not. In particular, there are two challenges.
First, data integration faces institutional constraints. Many financial institutions find it hard to provide services to micro and small enterprises because they lack enough information to assess the credit condition of these enterprises.
Guangzhou, Zhejiang, Shandong and some of the northern regions have made good attempts to support financial services via data integration. They have established local financial information platforms to integrate local “static information” and “hard information” such as social security, tax, judicial, and utility data, so as to help improve financial services for micro-, small- and medium-sized enterprises.
Technically, security issues could be solved through methods like federated learning and secure computation. On such platforms, companies or individuals authorize the platform to collect data and use them for credit assessment. This attempt has achieved good results and thus can be promoted as a successful case.
One of the problems is that the compartmentalization of administrative departments has hindered the maximum sharing of data. To solve the problem, we need a top-level design that can enable data sharing while ensuring security. Only in this way can we provide better financial services.
Second, big data credit reporting needs to deal with challenges like data iteration and profit distribution. China has two big data credit reporting companies but they are facing great challenges. We conducted study with the IMF and BIS on whether big data can be applied to credit risk evaluation. The answer is yes but with certain conditions. Such application shows good results in risk control when it’s limited to small, short-term loans. Once the loan amount is raised, the evaluation method might no longer work.
More importantly, the success of big data in risk control rests on two critical premises.
First, constant iterations. Many data are unconventional data. Transaction-based information and related data need to be constantly iterated in order to reflect people’s credit status, ability and willingness to repay. Without iterations, such data alone may not reflect anything useful.
This also means that big data credit reporting system should adapt to continuous and dynamic iterations of data. This is different from traditional credit reporting. FICO Scores in the US are static data, which can help financial institutions assess credit risks. In contrast, big data credit reporting needs dynamic iterations and real-time verification to make reliable assessment of credit risk.
Second, big data credit reporting relies on a comprehensive and sound ecosystem. Credit risk management is a systematic project. Big data credit risk evaluation is only a part of it. Big tech platforms and their ecosystems that generate digital footprints are just as important, especially for repayment management. Once separated from the platforms, big data alone can hardly perform effective credit risk assessment, let alone guarantee a relatively low default rate.
Currently, China’s big data credit reporting companies are not giving large tech firms enough incentives. Tech firms not only need to share data but also provide analytical support, yet they are only minority shareholders of these credit reporting companies, making this kind of sharing model hard to succeed.
The above two examples show that financial data sharing needs to be based on data’s own features. For data suited for sharing, we should create conditions to share them to maximize the benefit. For data not suited for direct sharing, we should find a mechanism to amplify their economic and social benefits.
For example, there is a credit service provided by big tech firms that treats individuals with no credit history as reliable borrowers. In this way, borrowing information is formed and included the credit reporting system of the central bank, thus allowing banks to lend money to these individuals. This could be seen as one way of big data sharing.
III. New thinking and measures of data governance in the era of digital economy
China’s digital economy is still in its infancy. Globally, digital economy could be divided into three markets: the US, China and the rest of the world. Among them, American companies dominate the third market. Of the top 20 digital tech unicorns, Chinese companies account for nearly half of them, which is remarkable.
However, if we dig deeper, we can find that China’s strengths are demographic dividend, separation dividend, and innovation opportunities due to insufficient data protection. It remains to be seen whether these strengths can keep supporting the development of China’s digital economy.
Putting China’s development into a global perspective, when we adopt regulatory and control measures, we should take into account the fact that China is leading in scale instead of technology. If we were to compete with other countries, how should we support and encourage innovation, especially on the premise of sufficient rights protection? This is our biggest challenge. This also means that we cannot simply follow American and European practices in the short term.
Another interesting observation is that in economic history, public antipathy towards big companies often emerges at the stage when income distribution continuously worsens. In the context of income inequality, people tend to be discontent with big companies and the super-rich. Improper and unfair economic practices could only exacerbate such hostility.
From this standpoint, big data governance should aim mainly towards supporting innovation and regulating behavior, but special emphasis should be placed on common prosperity, since innovation should not just result in thousands of billionaires.
Therefore, big data governance requires new thinking modes, as it might change economic behavior or rules of economic operation, like economy of scale, economy of scope and diminishing marginal returns. Governance measures should also change accordingly. Following traditional ways or copying US and European experiences may not be appropriate.
Take anti-trust action as an example, traditional criteria for monopoly include market share and price. In the past, a company could be defined as monopolistic as long as its market share reaches a certain level. But today, we may need to adopt a new mindset. Since the basic feature of platform economy is economy of scale, a successful company will certainly become large in scale. But the key factor is contestability, or whether there is market competition. To identify monopoly, contestability might be a more reliable criterion than market share.
Overall, as the development of digital economy has many new features and patterns, governance of digital economy needs new thinking and measures.