The impact of chinese big tech on the traditional financial market: evidence from Ant Group

Abstract

Based on the actual situation of Chinese financial market, we logically deduce the risk spillover from Big Tech’s financial business to the traditional financial market. We combined the data of Ant Group to empirically analyze the impact of Chinese Big Tech’s financial business on the profitability of the traditional financial market. The results show that Chinese Big Tech’s financial business has an impact on the traditional financial market, but the degree is different. It has a greater impact on the stock and trust markets, followed by insurance and funds, and less on the banking industry. Secondly, impacts have significant time-varying characteristics and have both immediate and long-term effects. The impulse response in banking, insurance and trust markets fluctuated, while equity and fund markets continued to decline. The short-term volatility of each market is mostly positive, but the medium and long-term volatility is negative. Thirdly, the impacts at each major point are significantly different and heterogeneous. The decline of Ant Group’s ABS issuing scale has a greater impact on the banking and trust markets, while its listing turmoil has a greater impact on the stock and fund markets.

Introduction

Big Tech refers to large global technology companies with digital technology advantages. They usually provide services such as social media, Internet search, software, online retail, data storage and processing to individual users, or provide infrastructure services to other companies [1]. However, in recent years, relying on its users, scenarios and technology advantages, Big Tech has further developed its financial business, won a number of financial licenses, and gradually formed a financial holding company with distinctive features. Unlike GAFA in the United States, Chinese Big Tech, represented by BATJ1, binds traditional financial institutions, abuses consumer financial leverage, induces excessive borrowing, recycle multiple asset securitization, indulges in the savage growth of capital, and forms a dual monopoly of technology and finance, which is a serious threat to the traditional financial industry stability.

Firstly, Big Tech relies on its monopoly to extract bank funds and operate with high debt. It uses complex credit enhancement methods such as asset packaging and stratification to reduce the share of co-lending and create a subprime mortgage bubble, which can easily lead to banking crisis. Secondly, with strong network effects, Big Tech integrates the bank lending market and the asset securitization market, and centrally distributes debt assets or income rights, which is easy to form shadow banking risks. In addition, Big Tech relies on capital expansion, hyping Big Tech concept stocks, leading to the accumulation of leveraged funds. A large number of bubbles in equity pledge financing is easy to cause securities market crisis from capital channels. At the same time, Big Tech has formed a quasi-monopoly in various fields, and has speaking right that affects the national economy and people’s livelihood. It can control public opinion and hide real information, which is easy to cause securities market crisis from information channels. Chinese Big Tech has the functions of financial infrastructure under the new economic form, and its business scope is highly related to the traditional financial market. While promoting Chinese financial development, it has buried new hidden financial risks.

Big Tech in China has attracted the attention of the academic circles, but little literature studies the impact of Big Tech on traditional financial markets. The existing research lacks the combination of Big Tech and traditional financial markets and quantitatively analyzation the intertemporal and time-varying characteristics of impact. Therefore, our research questions are, will Big Tech’s financial business have an impact on Chinese traditional financial market? What is the strength of the impact? Are the shocks different for different types of financial institutions? Specifically, we incorporate Big Tech and traditional financial markets into a unified research framework, and analyze the risk spillover from Big Tech to traditional financial markets. Based on the cross-business relationship between Big Tech and traditional financial markets such as banks, stocks, insurance, trusts, and funds, we select Ant Group as the research object, and use the time-varying parameters structural vector autoregressive model with stochastic volatility (TVP-SVAR-SV) to quantify Chinese Big Tech’s impact on the profitability of traditional financial markets and analyze its dynamic time-varying characteristics.

The innovative findings in this paper are: (1) Starting from Big Tech financial risks, we have logically deduced the risk spillover from Big Tech’s financial business to traditional financial markets. (2) We use a Bayesian time-varying parameters structural vector autoregressive model with stochastic volatility to study the impact of Big Tech’s financial business on the profitability of traditional financial markets at different time points and events, so as to analyze the intertemporal time-varying characteristics of risk spillover effect. (3) Through the identification strategy of group test, we confirmed the heterogeneity of risk spillovers from Big Tech’s financial business to traditional financial institutions with different types and regions.

Literature review

Big Tech’s financial business stems from the rapid development of Chinese FinTech industry in the past ten years. FinTech is the integration of Financial and Technology [2]. Regarding the definition of financial technology, Schueffel [3] emphasized that FinTech can be defined as a new type of financial industry that applies technology to improve financial activities. Gomber et al. [4] believe that FinTech is not only a large number of digital financial services and products provided by FinTech companies, but also covers various financial software developed by them, as well as convenient customer interaction models. The characteristic of FinTech is that the front-end interface and the back-end system are interconnected and the efficiency of information processing is improved. Sakasonova and Kuzmina-Merlino [5] attribute the development of FinTech to globalization, which provides opportunities for small start-ups to develop financial services without the help of banks. Through the combination of finance and IT, such enterprises provide consumers with more efficient traditional banking services [6]. Ivashchenk et al. [7] propose that FinTech has transformed the traditional delivery of financial services by implementing new technologies. These technologies provide financial products and services directly to end users through online and mobile channels [8]. In recent years, Chinese FinTech has developed rapidly. It has evolved from the early Internet Finance to ABCD pattern, including Artificial Intelligence, Blockchain, Cloud Computing, and Big Data. In addition, FinTech also includes the use of R programming, remote sensing technique, GIS analysis for Spatial data analysis [9–15].

The development of FinTech is mainly composed of three stages: financial IT stage, Internet Finance stage, and FinTech stage [1, 16]. The first stage appeared in the 1980s. During that period, the overall IT level of the financial industry was significantly improved. Electronics became the core of financial industry. Electronic devices such as ATM and POS were widely used during this period, greatly promoting the business efficiency. The second stage occurred at the beginning of the new century. Internet finance has developed rapidly, and online business platforms have emerged. Many financial institutions have effectively integrated customer resources and sales channels based on the Internet [17]. Nowadays, FinTech has entered the 3.0 stage. Not only has there been a comprehensive innovation in key technologies, but the integration of finance and technology has also been greatly improved. The application of big data, cloud computing, artificial intelligence, and blockchain [18] has completely changed the original practices of traditional financial institutions in customer information collection, risk pricing models, and investment decision-making processes.

The FinTech has brought huge challenges to the banking industry. The emergence of Internet finance companies and third-party payment platforms has crowded out liability, intermediary and asset business of commercial banks [19].

First of all, FinTech has an impact on the liability business of commercial banks. Internet financial products are the important source of shock. It has low transaction costs, low information asymmetry, good customer experience and easy operation. Therefore, it can exert a strong “long tail effect”, gathering a large number of customers with small funds [20]. At the same time, compared with the liability business of commercial banks, Internet financial products have certain advantages in terms of liquidity and yield. This would be a huge draw for investors, diverting low-cost deposits from banks. In the long run, the Internet financial products will alienate the risk disposal model of bank assets and liabilities, increase the interest rate volatility of bank funding, and cause risks to accumulate within the banking system [21]. Obviously, the launch of Internet financial products has brought a negative impact on commercial banks, which will objectively reduce bank performance [22].

Secondly, FinTech has an impact on the asset business of commercial banks. Large banks focus on “hard information”, such as financial statements. In contrast, small banks accumulate “soft information” through person-to-person exchanges. This has advantages in small and micro enterprise loans. Even, this behavior has advantages over relying solely on computers in avoiding the principal-agent problem [23]. With the advancement of technology, the improvement of information transparency can help commercial banks optimize their asset business. “Relationship loan” may gradually disappear [24]. For example, in the 1990s, banks began to replace relationship loans with automated credit scoring systems when serving SMEs [25]. In addition, the use of IT to achieve information sharing also helps banks provide loans to long-tail customer groups [26].

Finally, the intermediary business of commercial banks will also be impacted by FinTech. Zheng and Liu [27] believe that the intermediary business is crucial to their survival, and payment settlement is an important part of intermediary business. Now it is being crowded out by third-party payment. Wang and Wang [28] pointed out that the squeeze of intermediary business will affect the profitability of banks. Zhao and Liu [29] analyzed the data of 94 banks in China from 2011 to 2015 and believed that the negative impact of FinTech and bank competition is greater than the positive impact of technology spillover. Competition will reduce the non-interest income of the latter, especially regional banks.

Logical deduction of risk spillover

Domestic and foreign literatures have discussed in detail the risk spillovers from Chinese Internet Finance to commercial banks. However, there are few related studies on Big Tech, which has been developing rapidly in the past two years. Furthermore, existing research does not analyze the impact of Big Tech on other financial markets. Therefore, according to the actual situation of Big Tech’s financial business, we deduce its risk spillover path.

Leaving hidden risks of internet finance

Big Tech agglomerate production elements and generates extreme economies of scale. The technological infrastructure with the nature of quasi-public goods is excessively claimed, and technological barriers are set up to obtain consumer surplus, resulting in the Digital Gap and monopoly. Big Tech pursues short-term gains and absolute profits, abuses consumer financial leverage, amplifies the scale of liabilities, distorts the interest rate premium, induces predatory lending, causing borrowers to be over-indebted. The group irrational behavior caused by violent market fluctuations can easily lead to the of public illegal financial activity. Big Tech’s cross-disciplinary financial business overlaps and penetrates. A large number of small and decentralized financial models appear. Traditional financial institutions rely on Big Tech in terms of network traffic direction and technological risk control, increasing the mutual transmission of risks in related institutions. The chain reaction leads to further contagion of risks, spreading individual risks into cross-industry and cross-market risks, and eventually leading to the cascading failure of multi-layer networks. Big Tech has a wide coverage and a similar business model. Affected by the nonlinear dynamic mechanism, the central node of the ‘small world’ network has strong adaptability with changes in the internal and external environment, so the financial risk contagion is more rapid.

In general, Big Tech has a huge volume and a wide range of stakeholders. Through the catalysis of complex networks and exponential effects, it has formed the characteristics of being too big to fail, too wide to fail, too large to fail, and too fast to fail. However, Big Tech prefers high-risk businesses. Even if there is a crisis, the government will set soft budget constraints for social stability, which will further encourage Big Tech’s risk-taking behavior and leaving hidden risks of Internet Finance.

Leading to crisis in the banking industry

Relying on its monopoly position, Big Tech extracts bank funds for high-debt operations, uses complex credit enhancement methods such as asset packaging and stratification to reduce the share of co-loans. This will transfer and conceal the credit risks of the original borrowers, and create a subprime mortgage bubble. Due to the insufficient of Big Tech’s risks retention, exogenous shocks such as Covid-19, cause high debt risks to be concentrated in the banking industry through the feedback of Liquidity Spirals and Devil Loop. Bank equity and debt risks are rising and exchange feedback with Big Tech through the capital chain. Under the influence of pro-cyclical effects, maturity, liquidity and credit mismatches occur, which can easily lead to banking crisis.

Formation of shadow banking risks

With its strong network effect, low threshold and popularity, Big Tech integrates the bank lending market and the asset securitization market, and uses Internet credit sales and settlement tools to centrally distribute debt assets or income rights. Asset management plan and its derivatives securitization split and multi-level nesting, arbitrage through credit sinking, so that high-risk products can be spread indiscriminately through Big Tech platform. Under the influence of market expectations, behavioral choices, and the currency and credit chain mechanism, the chain effect of outsourced investment accumulates and magnifies, and it is easy to form shadow banking risks.

Causing securities market crisis through funding channels

Big Tech relies on the capital expansion of disorderly operation and the speculation of Big Tech concept stocks to lead to the accumulation of leveraged funds, resulting in excess market liquidity, and a large number of bubbles in equity pledge financing market. However, Big Tech is greatly affected by the market and policies. Risk factors such as business, law, finance, and technology will lead to the capital withdrawal from concept stocks, tightening market liquidity, and forming a unilateral market with continuous spillover of risk. Once asset prices plummet, and liquidity breaks, the leveraged funds hit the warning line for liquidation. Investors have to withdraw high-quality assets from the market to forcibly liquidate their positions. Financing liquidity and market liquidity interact, forming a vicious cycle that accelerates market decline, leading to liquidity depletion of market, and triggering securities market crisis.

Causing securities market risks through Information channels

Big Tech has formed a monopoly or quasi-monopoly in various fields, and has speaking right that affects the national economy and people’s livelihood, which can control public opinion and hide real information. Big Tech’s service objects are mostly long-tailed people who lack professional financial knowledge and investment decision-making ability. Their herd mentality is serious. The polarization of the network group resulted from the risk exposure formed by asset volatility can easily lead to the market panic and rapid spread of long-tail risk, and eventually cause crisis in the securities market (Fig. 1).

Fig. 1.

Logical deduction of risk spillover

Models

The logical reasoning above theoretically affirms Chinese Big Tech’s impact on traditional financial markets, but we still need to analyze the impact from a quantitative aspect. Big Tech has variable adaptability and exponential effect, so the model should compare the time-varying and abrupt-changing characteristics of immediate and long-term impacts under lagged conditions. We adopt Bayesian time-varying parameters structural vector autoregressive model with stochastic volatility (TVP-SVAR-SV) proposed by Nakajima et al. [30]. This model assumes that the intercept term, coefficients, variance and covariance terms all change with time, which is more in line with the operation logic of Big Tech. We can obtain unstable correlations between variables by estimating time-varying parameter. At the same time, we can solve the model heteroskedasticity through time-varying volatility, thereby improving the accuracy of estimation.

Since the VAR model was put forward, it has been widely used in various fields of microeconomics. However, the assumption of constant coefficients cannot explain the nonlinear relationship between variables when there is a sudden change. To solve this problem, many improved models have emerged since the 1990s. A common paradigm is the nonlinear dynamic VAR model. In recent years, with the improvement of computing power, the nonlinear variable coefficient VAR model based on the Markov Chain Monte Carlo (MCMC) has become an important development direction. The nonlinear variable coefficient characteristic makes it more sensitive and more robust to grasp economic changes than the previous nonlinear dynamic VAR model. This article mainly analyzes the impact of the financial business of Big Tech represented by Alibaba on traditional financial institutions. During this period, Alibaba experienced important events such as the decline in the scale of Ant Group’s ABS and the Ant Group IPO turmoil. In comparison, the nonlinear VAR model is a better choice to represent the dynamic time-varying characteristics. At present, nonlinear VAR models mainly include TVAR, MS-VAR and TVP-SV-VAR three types. However, the former two have the problem of discretization of the conversion mechanism, while the latter has the problem of too many mathematical constraints. We will use variable coefficient VAR model for empirical analysis. Drawing on the form of Nakajima et al. [30], we assume that the parameters to be estimated obey the first-order random walk and adopt the form of random volatility. This can fully reflect the persistent changes in parameters brought about by structural mutations, and reduce the estimation bias caused by fluctuation differences.

The construction of the TVP-SVAR-SV is based on the SVAR. The SVAR can be expressed as Eq. (1).

$$A{y}_t=F_1{y}_{t-1}+\cdots +F_s{y}_{t-s}+{\mu }_t,t=s+1,\dots ,n$$ 1

of which, $y_t$ is a column vector with $k\times 1$ dimensions. $A$ and $F_1,\cdots ,F_s$ are coefficient matrices with $k\times k$ dimensions. ${\mu }_t$ is the structural impact term with $k\times 1$ dimensions. It is assumed that ${\mu }_t\sim \text{N}(0,\mathrm{\Sigma }\mathrm{\Sigma })$, and $\mathrm{\Sigma }$ can be represented by Eq. (2).

$$\mathrm{\Sigma }=\left(\begin{array}{c}\begin{array}{cc}{\sigma }_1& 0\end{array}\begin{array}{cc}\cdots & 0\end{array}\\ \begin{array}{cc}0& \ddots \end{array}\begin{array}{cc}\ddots & ⋮\end{array}\\ \begin{array}{cc}⋮& \ddots \end{array}\begin{array}{cc}\ddots & 0\end{array}\\ \begin{array}{cc}0& \cdots \end{array}\begin{array}{cc}0& {\sigma }_k\end{array}\end{array}\right)$$ 2

SVAR assumes that the coefficient matrix $A$ is a lower triangular matrix, which can be expressed as Eq. (3).

$$A=\left(\begin{array}{cccc}1& 0& \cdots & 0\\ a_{21}& \ddots & \ddots & ⋮\\ ⋮& \ddots & \ddots & 0\\ a_{k1}& \cdots & a_{k,k-1}& 1\end{array}\right)$$ 3

By integrating the Equations above, Eq. (1) can be rewritten into a simple VAR, as shown in Eq. (4).

$$y_t=B_1{y}_{t-1}+\cdots +B_s{y}_{t-s}+A^{-1}\mathrm{\Sigma }{\epsilon }_t,{\epsilon }_t\sim N(0,I_k)$$ 4

of which, $B_i=A^{-1}{F}_i$, $i=1,\cdots ,s$. We stack the elements in $B_i$ into a column vector $b$ with $k^{2}s\times 1$ dimensions, and define: $X_t=I_s\otimes (y_{t-1},\dots ,y_{t-s})$. $\otimes$ represents the Kronecker product. Equation (4) can be abbreviated as Eq. (5).

$$y_t=X_{t}b+A^{-1}\Sigma {ϵ}_t$$ 5

The parameters in Eq. (5) do not change with time. In order to construct the TVP-SVAR-SV, we need to rewrite Eq. (5) into Eq. (6).

$$y_t=X_t{b}_t+A_t^{-1}{\Sigma }_t{ϵ}_t,t=s+1,\dots ,n$$ 6

In Eq. (6), $b_t$, $A_t^{-1}$and ${\mathrm{\Sigma }}_t$are state variables that change with time. Primiceri [31] defines the non-zero element in the lower triangular matrix ${\text{A}}_t$ as $a_t=\left(a_{21},,,a_{31},,,a_{32},,,\cdots ,,,a_{k,k-1}\right)$, and defines $h_t=\left(h_{1t},,,\cdots ,,,h_{\mathit{kt}}\right)$ satisfying $h_{\mathit{jt}}=log{\sigma }_{\mathit{jt}}^2$, of which, $j=1,\cdots ,k$, $t=s+1,\cdots ,n$. In order to reduce the estimated parameters, we assume that the parameters follow a random walk. That is, $b_t$, $a_t$ and $h_t$ obey $b_{t+1}=b_t+{\mu }_{b_t}$, $a_{t+1}=a_t+{\mu }_{a_t}$, $h_{t+1}=h_t+{\mu }_{h_t}$, while satisfying the distribution conditions shown in Eq. (7).

$$\left(\begin{array}{c}\begin{array}{c}{ϵ}_t\\ {\mu }_{b_t}\\ {\mu }_{a_t}\end{array}\\ {\mu }_{h_t}\end{array}\right)\sim N\left[0,,,\left(\begin{array}{c}\begin{array}{cc}I& 0\end{array}\begin{array}{cc}0& 0\end{array}\\ \begin{array}{cc}0& {\Sigma }_b\end{array}\begin{array}{cc}0& 0\end{array}\\ \begin{array}{cc}0& 0\end{array}\begin{array}{cc}{\Sigma }_a& 0\end{array}\\ \begin{array}{cc}0& 0\end{array}\begin{array}{cc}0& {\Sigma }_h\end{array}\end{array}\right)\right]$$ 7

Since there are many parameters to be estimated, it is difficult to give exact expressions for the estimation of some parameters. It is very difficult to use the traditional maximum likelihood estimation (MLE) for parameter estimation. To address this difficulty, parameters are generally estimated by applying Markov Chain Monte Carlo (MCMC) in a Bayesian framework. Commonly used methods in MCMC include Metropolis, Metropolis-Hasting and Gibbs. Among them, Gibbs only requires to know the marginal probability density distribution of the parameters, so the posterior distribution can be obtained, while the cyclic sampling can be performed to obtain the Markov chain of the parameters. Generally speaking, the marginal distribution is relatively easy to obtained, so the Gibbs is widely used. Since Gibbs is based on the posterior distribution, this sampling is a method of Bayes inference. Applying this method can not only achieve accurate estimation of parameters, but also achieve consistent estimation of state variables.

According to the principle of the MCMC, if the prior distribution is set reasonably, the number of simulated samplings can be reduced, and the convergent parameter estimation results can be obtained faster. However, even if the prior distribution is set unreasonably, if the sampling times are large enough, it is possible to get a reasonable posterior distribution result. In order to apply the Gibbs, we need to assign the initial values for the parameters, and set the prior distribution of the time-varying parameters as: ${\mu }_{b_0}={\mu }_{a_0}={\mu }_{h_0}$, ${\mathrm{\Sigma }}_{b0}={\mathrm{\Sigma }}_{a0}={\mathrm{\Sigma }}_{h0}=10\times I$. For simplicity, assume ${\mathrm{\Sigma }}_b$ is a diagonal matrix and the prior distributions of the three covariance matrices are Eqs. (8), (9), (10):

$${\left({\Sigma }_b\right)}_i^{-2}\sim Gamma\left(\text{40,0.02}\right)$$ 8

$${\left({\mathrm{\Sigma }}_a\right)}_i^{-2}\sim Gamma\left(\text{40,0.02}\right)$$ 9

$${\left({\mathrm{\Sigma }}_h\right)}_i^{-2}\sim Gamma\left(\text{40,0.02}\right)$$ 10

Data

For Big Tech, we take Ant Group, a subsidiary of Alibaba, as our research object. Alibaba is the largest internet company in China, and it is the earliest to start financial business. Alibaba’s subsidiary Ant Group owns the Chinese largest financial business application, Alipay. Therefore, the Ant Group is the most typical. Specifically, we select the Wind Ant Group Index. The index includes companies that directly or indirectly hold shares in Ant Group, Ant Group shareholding companies, Ant Group holding companies, and Ant Group partners.

In the traditional financial market, we take banks, stocks, insurance, trusts, and funds as the objects. Similarly, we selected CSI 300 Bank Index, CSI 300 Stock Index, CSI Insurance Index, CSI Trust Index, and CSI Fund Index. At the same time, we select SWS Large Bank Index, SWS Joint Stock Bank Index, SWS Regional Bank Index for heterogeneity test. We select Wind Internet Finance Index as the control variable.

The impact of Big Tech is directly reflected in the decline of the profitability of traditional financial markets. Therefore, we take the logarithmic rate of return of each index, and make monthly adjustments to all indexes. We perform Z-score processing to obtain the standardized index rate of return as the surrogate of profitability. The range is from March 2018 to January 2022 and the data is from the Wind.

Empirical results

Markov chain Monte Carlo (MCMC)

We use MCMC to estimate the parameters and set to sample 10,000 times. When using the MCMC algorithm to sample, the results before convergence are not stable distribution. We discard the first 1000 results, that is, delete the results in the “burn-in” process. Due to space limitations, we select the impact of Ant Group on the banking industry as a representative to analyze the MCMC results.

In Table 1, the Gewekes are significantly lower than 1.96, so the null hypothesis that ‘sampling results converge to the posterior distribution’ cannot be rejected. It can be seen from the Inefficiency that only (${\mathrm{\Sigma }}_h$)₁ and (${\mathrm{\Sigma }}_h$)₂ have the higher invalid factors. But we can still get 241 uncorrelated samples (10,000/41.50) in the 10,000 samplings, which is enough to fit the posterior distribution. This verifies the rationality of the MCMC.

Table 1.

The results of parameter estimation by MCMC

Parameter	Mean	Stdev	95%L	95%U	Geweke	Inefficiency
(${\mathrm{\Sigma }}_b$)1	0.0227	0.0026	0.0184	0.0286	0.759	1.60
(${\mathrm{\Sigma }}_b$)2	0.0227	0.0026	0.0183	0.0283	0.105	2.95
(${\mathrm{\Sigma }}_a$)1	0.0673	0.0211	0.0391	0.1207	0.238	12.16
(${\mathrm{\Sigma }}_a$)2	0.0754	0.0271	0.0408	0.1462	0.538	22.18
(${\mathrm{\Sigma }}_h$)1	0.1059	0.0481	0.0482	0.2320	0.668	41.50
(${\mathrm{\Sigma }}_h$)2	0.1237	0.0726	0.0487	0.3175	0.062	37.26

Mean is the mean of the parameter posterior distribution. Stdev is the standard deviation. 95%L and 95%U are the lower and upper bounds of the 95% confidence interval. Geweke is convergent diagnostic statistic. Inefficiency is the number of ineffective factors. The values in (${\mathrm{\Sigma }}_b$) and (${\mathrm{\Sigma }}_a$) are the result of multiplying by 100

Figure 2 shows the distribution of some parameters in the MCMC. We also select the impact of Ant Group on the banking industry as a representative to analyze the distribution of some parameters in the MCMC. The first row of Fig. 2 is an autocorrelation diagram. The second row is a simulation path diagram. The third row is the fitted posterior distribution density function. It can be seen from the autocorrelation diagram that the autocorrelations of the parameters converge rapidly to 0. The autocorrelation coefficients of (${\mathrm{\Sigma }}_b$)₁ and (${\mathrm{\Sigma }}_b$)₂ decrease the most rapidly, while the convergence speed of autocorrelation coefficients in (${\mathrm{\Sigma }}_h$)₁ and (${\mathrm{\Sigma }}_h$)₂ is significantly slower than that of other variables, which corresponds to that the (${\mathrm{\Sigma }}_h$)₁ and (${\mathrm{\Sigma }}_h$)₂ have the largest invalid factors reflected in Table 1. It also shows that MCMC effectively simulates the distribution of parameters. It can be seen from the simulation path diagram in the second row that the concentration of the six sample simulation paths is high. That is to say, the frequency of extreme values in the simulation path of the six parameters is not high, reflecting that the simulation path of parameter is relatively stable. It can be seen from the posterior distribution density function graph in the third row that the density function of the 6 parameters are all spiky and thick-tailed distributions. Combined with other judgment conditions, this distribution does not affect the validity of the sampling results. That is, the density function can also be regarded as an approximate normal distribution.

Fig. 2.

Distribution of some parameters in MCMC

Index Movement and posterior volatility analysis

In the SV-TVP- SVAR, the parameters to be estimated at each moment have $k+k^{2}s$ coefficients, $k$ variance autoregressive coefficients, $k(k-1)/2$ elements in the structure matrix that describe the current relationship between variables, and $k$ variance volatility. $k$ is the number of variables entered into the model. $s$ is the lags for the SV-TVP- SVAR. We set $k=3,s=2$.

The upper part of Fig. 3 shows the movements of each index. It can be seen that the largest fluctuation of the logarithmic rate of return is the Ant Group Index, indicating that the Ant Group has relatively high risks. As a financial platform for Big Tech, Ant Group has diversified financial businesses. However, the Chinese government has not set up a special agency to regulate the financial risks of Big Tech, so the financial business of Big Tech has been in a state of savage growth and a regulatory vacuum. In recent years, with the standardized development of the Internet Finance, the scale of some of Big Tech’s financial businesses has declined, but Big Tech has not stopped the pace of financialization, so the risk is still high.

Fig. 3.

Index Movement and Posterior Volatility

The fluctuations of the logarithmic returns of the bank index and the trust index are similar, but the return of the trust index is more volatile, indicating that its overall risk is greater than that of the banking industry. As we all know, bank-trust cooperation is an important part of Chinese shadow banking. Their businesses are intertwined, also their interests. It is easy to induce cross-industry and cross-market risk contagion, so the fluctuation is similar [32]. But by contrast, trusts are in a regulatory grey area and have more risks.

The fluctuations of the logarithmic returns of the stock index and fund index are similar to that of Ant Group. The securities market represented by the stock index is still in infancy in China. Most of the investors are retail investors, who do not have professional investment ability, and are prone to group irrational behavior. The rapid spread of reputational risks and long-tail risks will also increase the stock market risks. Similarly, Chinese fund market is dominated by stock funds, so the risks are also higher.

The bottom half of Fig. 3 is the posterior volatility of each index. It can be seen from the figure that the variance of logarithmic returns of each index fluctuates over time, which further verifies the rationality of the TVP-SVAR-SV’s assuming that the variance fluctuates randomly. In addition, the variance fluctuation of the Ant Group Index is synchronized with that of the traditional financial market. This reflects that the financial business carried out by Ant Group has indeed had an impact on the traditional financial market, and this impact is more obvious during the period when Ant Group’s own risks are high. It can be said that the indexes we choose can reasonably refer to the impact of Big Tech on traditional financial markets.

Analysis of immediate impulse response of traditional financial market to Ant Group

Since the TVP-SVAR-SV can give different estimates of parameters at different time points, VAR analysis of constant coefficient can be performed at different time points. As one of the characteristics that distinguish the TVP-SVAR-SV from the traditional VAR model, the immediate impulse response uses the time-varying parameters estimated by the MCMC to fit the impulse response of the traditional financial market to Ant Group at all time points.

The immediate relationship is defined by the matrix A from Eqs. (1) to (4). The immediate impulse response of the traditional financial market to Ant Group are shown in Fig. 4. The solid line in the middle represents the posterior mean, and the two dashed lines represent the upper and lower bounds of the 95% confidence interval. It can be seen from the figure that the immediate impulse responses of the traditional financial market are all higher than 0, indicating that Ant Group has an impact on the traditional financial market at all times. Comparing Figs. 3 and 4, it can be seen that when Ant Group’s profitability reaches the highest, the immediate impulse responses in various traditional financial markets are all low. It shows that the volatilities of Ant Group and the traditional financial market has changed in opposite during the same period. As the profitability of Ant Group increases, the profitability of traditional financial markets gradually decreases. That is, the risk spillover of Ant Group gradually increases.

Fig. 4.

Immediate impulse response of traditional financial market to Ant Group

Analysis of time-varying parameter impulse response for different terms

Figure 5 depicts the time-varying parameter impulse response results for different terms. Since the data frequency is monthly, the lags we set are 3 periods, 6 periods, and 9 periods, corresponding to the first, second and third quarters of the Chinese fiscal year. However, Alibaba is a listed company in the United States, and the annual report issuance time is not consistent with the Chinese fiscal year, so we choose the time corresponding to Alibaba’s annual regular announcements, which is May, August and November. These three months correspond to the short- term, medium- term, and long-term of exogenous shocks.

Fig. 5.

Time-varying parameter impulse response for different terms

Figure 5 includes five traditional financial markets of banks, stocks, insurance, trusts, and funds. It can be seen from Fig. 5 that the time-varying parameter impulse response curves of different terms in the traditional financial market are different. Time-varying parameter impulse responses in the stock and fund markets showed a downward trend, while the rest were volatile. Most short-term volatility is positive, while medium and long-term volatility is negative. In addition, the volatility of the impulse response of the stock market is larger, and banks and trusts are smaller.

Specifically, after the bank industry was hit by Ant Group, the impulse response ${ϵ}_{\mathit{ant}}\uparrow \to bank$ fluctuated significantly. The impulse response of the short-term impact was positive, but it was negative in the medium and long term. The fluctuation is between ± 2 and the amplitude is small. The stock ${ϵ}_{\mathit{ant}}\uparrow \to stock$ shows a downward trend after the short-term impact, while the impulse response in the medium and long term fluctuates greatly. Unlike banks, the stock market fluctuates between ± 5, with a large amplitude. After the short-term, medium-term and long-term impact of Ant Group, ${ϵ}_{\mathit{ant}}\uparrow \to insurance$ fluctuated wildly and all were negative. Volatility ranges from − 3.5 to 2.5. ${ϵ}_{\mathit{ant}}\uparrow \to trust$ are similar to banks, with positive impulse responses to short-term impacts, but negative to mid-term and long-term impacts. The difference is that trusts fluctuate between ± 2. Similar to the stock market, the funds impulse response shows a downward trend. The short- term, medium- term and long-term impulse response are all positive, and the fluctuation range is between − 3 and 2.

According to the theoretical analysis above, Ant Group relies on its monopoly position in the Internet industry to extract funds from commercial banks and operate with high debt. Most of its financing channels are commercial banks and it is deeply tied to commercial banks. At the same time, Ant Group used complex credit enhancement methods such as asset packaging and stratification to reduce the share of co-lending, transfer and conceal the credit risk of the original borrowers, and create a subprime mortgage bubble. Due to the insufficient risk retention of Ant Group itself, exogenous shocks cause high debt risks to be concentrated in the banking industry through the feedback of liquidity spirals and devil loops. Bank equity and debt risks are rising. Correspondingly, the impulse response of banks profitability fluctuates wildly, reflecting increased risk.

Different from banks, the profitability of stock and fund markets shows a downward trend during short-term impacts, and the mid-term and long-term profitability fluctuates greatly. However, the impulse response coefficient of the stock market is higher than that of the fund market. Part of the reason is that the Ant Group Index includes companies that directly or indirectly hold shares in Ant Group, Ant Group shareholding companies, Ant Group holding companies, and Ant Group partners. These listed companies will directly or indirectly have an impact on the stock market, resulting decline in impulse response after a short-term impact. On the other hand, Ant Group, as a quasi-listed company with a market value of more than 300 billion dollars, has a profound impact on the stock market. Especially in the long run, Ant Group is listed in the Internet industry with a high PE. After the IPO, related concept stocks may rise, and the stock market will fluctuate greatly. Therefore, Ant Group will have a big risk spillover to the stock market.

In addition to the credit business, the insurance is also one of the main businesses of Ant Group. Up to now, Ant Group has contributed 7.5 billion dollars in premiums, and has cooperated with more than 90 insurance institutions. Ant Group’s involvement in the online insurance market has a great negative impact on the traditional offline insurance market. However, due to the relatively weak development of Chinese insurance market, the development of online insurance has positive feedback on the current insurance market in the long run. Therefore, the impact of Ant Group on the profitability of the insurance market has fluctuated. The long-term profitability has increased, and the risk spillover has gradually decreased.

Similar to insurance, the impulse response of the trust market also shows fluctuations, with large short-term fluctuations and a decline in the medium and long term. After the bank-trust cooperation was suspended, the ties between Ant Group and the trust market gradually deepened. Shadow banking became active again. Therefore, as the profitability of Ant Group increases, the profitability of the trust market is also gradually increasing. Until the end of 2019, the China Banking Regulatory Commission issued new regulations on asset management and extended the transition period to the end of 2021, effectively preventing the disorderly development of Chinese shadow banking. Therefore, the profitability of shadow banking declines after 2020. However, in the future, shadow banking will still develop with the support of Big Tech. Therefore, the profitability of bundled trust market will gradually increase.

Analysis of impulse response of major points of Ant Group

One of the characters in TVP-SVAR-SV is that it can analyze the impulse response at a certain point. Therefore, we will study the impact of major events of Ant Group on the traditional financial markets. Figure 6 shows the impulse response at three points in October 2018, August 2020, and January 2021. Corresponding to the decline in sales of Ant Group’s financial products related to consumer loans, Ant Group announcement of preparing for IPO, and Ant Group announcement IPO suspended.

Fig. 6.

Time-varying parameter impulse response of major points

Figure 6 is time-varying parameter impulse response at major events in banks, stocks, insurance, trusts and funds markets. Comparing the markets above, it can be seen that all the impulse responses of time-varying parameters are trending downward. The difference is that the stock, insurance and trust markets have an impulse response more than 5, followed by funds. The lower is banking industry. It proves that Ant Group has a negative impact on the profitability of the traditional financial market at the major points.

In the first fiscal quarter of 2019 (from May to August 2018), Ant Group’s consumer finance ABS issuance scale fell off a cliff. It was not issued in the first two months. Only two ABS products with ‘Huabei Consumer Loan’ were issued in the third month, with an issuance of only 600 million dollars, a year-on-year decrease of 3 billion dollars. This is because regulatory policies for online loans continued to be strengthened. As we all know, the small loan company under Ant Group has two products: Huabei and Jiebei2. Ant Group used 400 million dollars of self-owned capital to lending, and got 4 billion dollars by taking the right of credit formed by the lending into the revolving ABS. The leverage is as high as 100 times. The right of credit of 4 billion dollars come from more than 2,000 banks in China. Once a risk occurs, the impact on the banking industry is immeasurable. However, in the first fiscal quarter of 2019, the decline in sales of financial products related to consumer loans of Ant Group had a great impact on the traditional financial market, especially banks and trusts. The impulse response of bank and trust market declined rapidly. This is because banks and trusts that rely on Ant Group for profit have suddenly lost some of their revenue. However, in the long run, the impulse responses of banks and trusts have a tendency to pick up. In contrast, bank regulation is stricter. Firewalls are more effective, and defense against external risks is stronger. Ant Group has less impact on banks.

On July 20, 2020, Ant Group announced that it plans to IPO simultaneously both in Science and Technology Innovation Board of Shanghai Stock Exchange and the Main Board of The Stock Exchange of Hong Kong Limited. Ant Group claims it will further support the digital upgrade of the service industry to expand domestic demand and strengthen global cooperation to assistant sustainable development. Also, it will support Ant Group to increase technology research and innovation. On November 2 of the same year, the People’s Bank of China, the China Banking and Insurance Regulatory Commission, the China Securities Regulatory Commission, and the State Administration of Foreign Exchange conducted regulatory interviews with several senior executives of Ant Group. On the same day, the China Banking and Insurance Regulatory Commission and the People’s Bank of China issued the ‘Interim Measures for the Regulation of Online Small Loan Business’ to standardize the online small loan business and unify operating rules. On the evening of November 3, Ant Group issued a letter to investors stating that the Group received a notice from the Shanghai Stock Exchange that day to suspend its IPO. Affected by this, the simultaneous IPO in Hongkong will also be suspended.

In recent years, Chinese Fintech has developed rapidly, and it has played a major role in improving the efficiency and quality of financial services. China encourages Fintech innovation, but its regulatory goals are also very clear. It is necessary to keep the bottom line of financial risks, and to prevent behaviors that use the guise of fintech innovation to wander on the edge of regulation. As an Internet company with a high PE, Ant Group IPO’s valuation exceeded 300 billion dollars. However, Ant Group itself has great risks. The IPO of such a unicorn company will definitely have a severe impact on Chinese stock market. Therefore, the impulse response of stock and funds markets show a long-term downward trend, and the degree of risk spillover increases.

Analysis of 3D impulse response of Traditional Financial Markets

Figure 7 is the three-dimensional impulse response of the traditional financial markets, that is, the analysis of the overall trend. In addition to five traditional financial markets, we also conducted a heterogeneity analysis of commercial banks that are deeply tied to Ant Group. Commercial banks in China are divided into large banks (Industrial and Commercial Bank of China, Agricultural Bank of China, Bank of China, China Construction Bank), joint-stock banks (China Merchants Bank, Shanghai Pudong Development Bank, China CITIC Bank, China Everbright Bank, Hua Xia Bank, China Minsheng Bank, China Guangfa Bank, Industrial Bank, Ping An Bank, Zheshang Bank, Hengfeng Bank, Bohai Bank, etc.), regional commercial banks (all urban commercial banks and rural commercial banks in China).

Fig. 7.

3D Impulse Response of Traditional Financial Markets

We focus on the heterogeneity analysis of large commercial banks ${ϵ}_{\mathit{ant}}\uparrow \to lbank$, joint-stock commercial banks ${ϵ}_{\mathit{ant}}\uparrow \to jsbank$and regional commercial banks ${ϵ}_{\mathit{ant}}\uparrow \to rbank$ in Fig. 7. As can be seen from the last three panels in Fig. 7, the increase of Ant Group’s profitability has had an impact on three types of commercial banks, but with different degrees. The impulse curve fluctuation of large banks is smaller than that of joint-stock banks. The longer the lag is, the more significant the impulse response is. The joint- stock bank is similar to the banking industry. With the increase of the lag, the impulse response first falls and then rises. The long-term degree is higher than that of large banks. The immediate volatility of regional banks is weak, but the impulse of long-term is higher than that of large banks and joint-stock banks.

We choose different types of commercial banks for heterogeneity analysis because Ant Group, as the largest Internet financial platform in China, is deeply bundled with more than 2300 commercial banks of different types. The sources of funds for the consumer finance ABS issued by Ant Group are all from commercial banks. However, Chinese large commercial banks are larger in scale and have stricter regulation. The Internet Finance business does not significantly increase the bank’s profitability. Large commercial banks pay no attention to the cooperation with Big Tech. Therefore, there is less cooperation with Ant Group, and the relevant cooperation is limited to the regional branches. The scale of cooperation is small, so as the impulse response.

Because Joint-stock commercial banks are more independent, they cooperate with Ant Group more than big banks. As listed companies, joint-stock commercial banks account for a larger proportion of the banking industry, and are strongly motivated by interests so as to be good at seizing any financial innovation opportunities. From previous mobile banking to FinTech nowadays, joint-stock commercial banks can be seen cooperating with Big Tech. Compared with large banks, joint-stock banks have a higher degree of impulse response, and fluctuate wildly in the long-term.

The predecessor of regional commercial banks was Chinese Rural Credit Cooperatives. After the reform and opening up, they were gradually restructured and turned into urban and rural commercial banks. Regional commercial banks are small in size, and their coverage is limited to regional residents. The particularity of regional commercial banks is that they do not have the nested regulation of large banks, nor do they have various regulation of listed companies like joint-stock banks. Therefore, most of Ant Group’s funds come from regional banks. Because of the small scale of regional banks, cooperation with Big Tech has a strong positive impact on the bank’s profitability, so the impulse response is higher than that of large banks and joint-stock banks.

The impact of FinTech applications on bank risk

In recent years, in the face of the challenges brought about by Big Tech, traditional commercial banks have also begun to actively use FinTech to start the transformation. In this process, FinTech has played a role in improving customer acquisition capabilities, reducing operating costs, strengthening risk control, and optimizing customer service [33]. In fact, among the many achievements of FinTech applications, risk control and anti-fraud are the most concerned by commercial banks. The reason is that FinTech can help commercial banks solve the most critical information asymmetry [24]. On the one hand, commercial banks can collect more dimensions of customer information with big data. On the other hand, artificial intelligence, cloud computing and blockchain can centralize massive data processing, draw a complete customer portrait, and reduce information asymmetry.

Although the application of FinTech can improve the risk control ability of commercial banks, the degree will be different between large banks and regional banks. At present, large banks usually choose to build their own FinTech subsidiaries or cooperate with Big Tech by virtue of their strong capital, scale and human resources. Most regional banks usually choose to cooperate with external FinTech companies or form FinTech alliances. The reason is that most regional banks lack massive data support and compound talents, and the cost of building their own FinTech platforms is high. Therefore, it is more practical to choose the strategy of “borrowing a boat to go to sea”. The “Internet Finance Alliance of Banks” (IFAB) is a typical representative.

In addition to capacity building methods, commercial banks’ own scale differences, management models, and organizational structures will also affect the effect of FinTech applications. On the one hand, credit scoring models rely on statistical analysis of default risk. Larger databases mean more accurate predictions. Therefore, large banks have a significant advantage [25]. On the other hand, large banks usually face principal-agent conflicts and lack market constraints and incentives [34]. This affects the use of FinTech. In contrast, regional banks have inherent advantages, such as light historical burden, short decision-making radius, and strong autonomy. This is what the Chinese often say, “a small boat makes it easy to turn around”. Regional banks can adapt faster to changes in new things. Research shows that joint-stock banks have the strongest ability to absorb FinTech spillovers, followed by regional banks, and large banks are the weakest [35].

Conclusions, policy suggestions and future scope

Combined with the actual situation of Chinese financial market, we logically deduce the risk spillover from Big Tech’s financial business to the traditional financial market in China. Firstly, Big Tech relies on its monopoly to extract bank funds and operate with high debt. It uses complex credit enhancement methods such as asset packaging and stratification to reduce the share of co-lending and create a subprime mortgage bubble, which can easily lead to banking crisis. Secondly, with strong network effects, Big Tech integrates the bank lending market and the asset securitization market, and centrally distributes debt assets or income rights, which is easy to form shadow banking risks. In addition, Big Tech relies on capital expansion, hyping Big Tech concept stocks, leading to the accumulation of leveraged funds. A large number of bubbles in equity pledge financing is easy to cause securities market crisis from capital channels. At the same time, Big Tech has formed a quasi-monopoly in various fields, and has speaking right that affects the national economy and people’s livelihood. It can control public opinion and hide real information, which is easy to cause securities market crisis from information channels. We established the TVP-SVAR-SV and empirically analyzed the impact of Chinese Big Tech’s financial business on the traditional financial market based on the data of Ant Group. The results show that Chinese Big Tech’s financial business has an impact on the traditional financial market, but the impact is different. It has a greater impact on the stock and trust markets, followed by insurance and funds, and less on the banking industry. Secondly, impacts have significant time-varying characteristics and have both immediate and long-term effects. The impulse response in banking, insurance and trust markets fluctuated, while equity and fund markets continued to decline. The short-term volatility of each market is mostly positive, but the medium and long-term volatility is negative. Thirdly, the impacts at each major point are significantly different and heterogeneous. The decline of Ant Group’s ABS issuing scale has a greater impact on the banking and trust markets, while its listing turmoil has a greater impact on the stock and fund markets.

Regarding policy suggestions, we believe that: First, traditional financial institutions, especially commercial banks, should actively integrate into FinTech, rationally use AI to improve the quality of financial services, and obtain greater performance growth. Second, traditional financial institutions should gradually transform to “light assets”, such as using 5G, VR/AR to create a remote interactive, touchable, and body-aware financial scene environment. This can provide customers with multi-level online financial services, give full play to the customer agglomeration effect, and reduce the marginal cost of financial services. Third, traditional financial institutions should use FinTech to build an open cooperation platform, attract Big Tech to participate in scenario design, and create a comprehensive financial ecosystem. Fourth, traditional financial institutions should use FinTech to optimize the risk monitoring system and improve the ability to identify, warn and deal with financial risks. Fifth, the government should strengthen the top-level design of FinTech and the reform of the financial system to create a sustainable competitive environment for traditional financial institutions.

Regarding the future scope of Big Tech, we believe that: First, future research should focus on the digital divide. FinTech derivatives bring digital monopoly, create digital barriers, disrupt market balance, and go against the original intention to enhance inclusiveness. This is a key problem Big Tech has when it comes to financial operations. This is also the focus of regulation of FinTech. Second, future research should focus on RegTech (Regulatory Technology). RegTech is also developing in parallel with FinTech. However, the innovation speed of FinTech is generally ahead of RegTech, and the resulting regulatory arbitrage and game troubles may have an impact on the traditional financial system.

Funding

Funding was provided by National Social Science Fund (Grant No. 22&ZD120 and 19BJL033).