Does the Digital Economy Promote Domestic Non-Tradable Sectors?: Evidence from China

Abstract

The impact of the digital economy (DE) has become the important faction of the market volume of domestic non-tradable sectors (DNSs). As rising digitalization supersedes traditional market power as a driving force, there is increasing concern about the volume of trade and economy; however, the literature of how the DE procession changed the DNS’s are limited, although the Chinese government is eager to enlarge the scale of the domestic market to be consistent with the trend of digitalization. This paper addressed this issue by employing a series of data from prefecture-level cities between 2010 and 2019 in China. Using panel data methods under fixed effect, synthetic difference-in-differences (SDID), and temporal-spatial econometrics, the paper’s hypothesis sheds light on the positive impact of the DE on DNSs. The regression results showed a 14.84% of improvement for the effects of DE development on DNS growth. The policy impact effect increased the average treatment effect by 3.9% average treatment effect, accompanied by temporal and spatial correlations. Further analysis illustrated that a possible intermediary mechanism through which the DE promotes the development of DNSs is the enhancement of the local product market development. It was concluded that policy-makers of developing countries should be devoted to breaking down domestic trade barriers among different regions to enhance the benefits of digitalization.

1. Introduction

The digital economy (DE) is penetrating and integrating all sectors of society in terms of depth and breadth. During the past ten years, the proportion of the DE in the GDP of China has increased from 21.6% to 39.8%. The popularity of digital technology can directly contribute to international trade that is more accessible and feasible [1,2]. Therefore, with the convenience of the Internet, the DE spreads the online trade domain over temporal and spatial barriers, and an increasing number of companies are changing towards e-commerce, especially cross-border e-commerce [3]. Indeed, the DE has become an indispensable part of China’s efforts to transform its form of economic growth [4] and optimize its economic structure [5]. As a result, international trade has become increasingly prevalent and accounts for a large proportion of trade during the DE process. However, there are controversies among scholars over whether domestic non-tradable sectors (DNSs) benefit from the DE [6,7,8,9]. Of course, DNSs, such as local product and service providers, investments, or public infrastructures, are indispensable for improving a country’s economic strength and development quality [10]. Could DNSs benefit from the digital transformation trend just as the international trade sector does? This paper attempts to find the evidence to answer this question.

In 2020, China proposed gradually forming a new development pattern with domestic circulation as the main body and domestic and international circulation mutually reinforcing each other. As a “high potential digital trade” country among 59 countries, China has enormous potential for digital trade due to its large economic scale and population [11]. Therefore, there is an urgent need to study how domestic non-tradable economy can be promoted by the DE.

Present studies focus on the DE and openness to the global market. However, few studies have been conducted on the internal circulation of domestic trade in China [12]. The DE can not only smooth the connection to the international market but also help the domestic markets remove trade barriers and establish trade facilitation and liberalization [13]. The traditional perception holds that DE development will naturally drive an in-crease in bilateral trade between countries, but the improvement in the DE by a single country cannot change the level of its trade partners because of national differences in the development of digital infrastructures. The gap in DE development for backwards trade partners shapes the bottleneck of international trade. In contrast, the DE transformation within a country can quickly improve the scale of DNSs, realizing the facilitation of digital trade and digital production activities, thus opening up a large regional market. Furthermore, the wide application of digital technologies not only raises production efficiency but also provides information sharing and diversified products for customers [9]. On this basis, how to effectively utilize the DE to develop domestic trade is a crucial issue for every country. This paper tries to provide some empirical evidence in China to serve as a reference for developing countries.

Regarding other countries, recent studies have primarily focused on developing countries. Scholars have probed problems such as how to develop the DE in Indonesia [14], how MNCs conduct international trade in the digital era in Brazil [15], and the import substitution policy in Nigeria [16]. The role of the DE in Africa’s foreign trade and economic development is also a focus of academics [17]. Although there are many similar studies on how such low- and middle-income countries (LMICs) should formulate their own national strategies for the DE [18], they mainly focus on international trade. In addition, there are studies exploring the role of the DE in relatively developed regions such as continental Europe [19], but few of them have focused on how to make full use of the DE in the development of the scale of their domestic non-tradable economy. Therefore, this paper’s discussion of the promoting effect of the DE on DNS volume has great practical significance.

We employed a panel fixed effect model to identify the static impact effect of the DE on DNS volume. Unlike most studies, we utilized the entropy weight method (EWM) to construct and measure the explanatory variable, the DE index (DEI). First, the baseline regression results demonstrated that the DE can strongly promote the development of the national DNS volume. To further enhance the robustness of the results, we changed the DEI construction method to a global principal component analysis (GPCA). Then, we utilized the identification strategies of a synthetic difference-in-differences (SDID) model to evaluate the policy impact and use an instrumental variable (IV) regression model to eliminate possible endogeneity. Second, we ran a temporal and spatial correlation model to explore the impact mechanism of the DEI. Third, the mediation effect analysis showed that the DE development level can affect the growth of the local non-tradable sector by improving the degree of product market development. Finally, our heterogeneity analysis found that the promoting effect varied by region and that the promoting effect on eastern China is significantly higher than that on central and western China.

2. Literature Review and Research Hypotheses

2.1. Literature Review

There is significant research in the literature on the nexus between DE and trade activities [17,20,21,22]. Most of the papers discuss the topic from three aspects: (1) DE development change production activities and reorganized behaviors of firms; (2) the approaches and methods in which DE expands the width and depth of trade, and improves trade scale; and (3) how DNSs benefit from DE.

Table 1. Summary of Literature review.

Literature Statements	Influencing Factors of DE	Authors, (Year)
The DE improves production	Catalyzing new technology revolution [23]	Wu & Yu, (2022)
	Supervision measures for enterprises [24]	Sun et al., (2020)
	Accelerating rural economic development [25]	Couture et al., (2021)
	Lifting enterprise total factor productivity [26]	Hua et al., (2022)
	Boosting foreign trade exports [17]	Abendin & Duan, (2021)
DE expand trade volume	Enhancing the trade nexus between economies [21]	Ma et al., (2022)
	Expanding the openness of developing countries and lowered trade barriers [22]	Ahmedov, I., (2020)
The mechanism of DE improving trade	Reducing the search and communication costs [27]	Bakos, J.Y., (1997)
	Loosen the restrictions of geographical distance [28]	Blum & Goldfarb, (2006)
	Increasing the possibility of trade [29]	Fink et al., (2005)
	Reducing the uncertainty of trade policies [30]	Baker et al., (2016)
	Popularize of automation and intelligent trade [31]	Brynjolfsson et al., (2019)
Mechanism of the DE affects the DNS	More individualized and differentiated products for customers [32]	Gorodnichenko & Talavera, (2017)
	Cheaper the prices of intermediate goods [33]	We & Zhang, (2022)
	Raising the domestic value-added rate [20]	Ding et al., (2021)

shows the summary of the literature statements.

First, the digitalization trend motivates the firms with a new tech revolution and advanced productivities. Vigorously developing the DE, as represented by artificial intelligence, big data, 5G technology and cloud computing, is the major priority of economic development competition in all economies [34]. Facing to the new round of the technological and industrial revolution [23], application of big data and information technology could further improve supervision measures for enterprises [24], compensating for the deficiency of traditional economic indicators [35] and promoting rural economic development through e-commerce [25], which further narrows the gap in regional economies. From the microeconomic perspective, the digital transformation has significantly improved enterprise total factor productivity [26].

Second, the empirical exploration of the trade expansion effect of the DE has become increasingly important. Numerous studies have found that the DE has boosted foreign trade exports, enhanced the trade nexus between economies, expanded the openness of developing countries and lowered trade barriers in the international market. Many studies in the literature focus on the mechanism through which the DE promotes trade expansion. It is easy to understand that the DE effectively promotes the trade volume by lowering costs. Earlier studies have found that the penetration of the DE and digital trade activities can effectively reduce the high search and communication costs incurred by enterprises [27] and break through the restrictions of geographical distance [28] to increase the possibility of trade [29]. Second, from the perspective of the trade structure, the DE has also raised the intensive margin and expansive margin of international trade. In addition, the application of digital technology brings many benefits, including curbing information costs, increasing the transparency and timeliness of information, avoiding many potential risks for enterprises, reducing the uncertainty of trade policies [30], and promoting the expansion of the trade scale [31].

Third, some studies have discussed the mechanisms of digital trade developments based on e-commerce. The DE provides a diversified and international market with cross-border e-commerce, rendering its products more individualized and differentiated for customers and no longer heavily relying on the scale effect [32]. Given the decreasing search costs that the DE brings, the quality of export products is upgraded because firms are able to purchase cheaper intermediate goods [33]. For small and medium-sized enterprises, areas with higher levels of Internet penetration and better DE development are trade priorities [34].

Existing empirical evidence from China shows that establishing a comprehensive experimental area of cross-border e-commerce effectively promotes an expansive margin but has no significant effect on the intensive margin [3]. Additionally, the evidence verifies that the DE not only expands the scale of trade imports and exports but also improves the quality and technical complexity of export products to achieve a high-quality development of foreign trade [36,37].

Another closely related stream of research in the literature investigates China’s degree and proportion of dependence in the domestic economic cycle. Based on the domestic and foreign demand and supply rates of final and intermediate goods. And Yang [38] concluded that China’s domestic economic cycle dependence is as high as 90%. He et al. [39] also found that China’s domestic economic cycle accounted for 81% in 2014. These studies basically conduct a theoretical discussion and an empirical analysis on the importance of domestic trade but lack empirical support for how to develop domestic trade facilitation and liberalization, which slows down the progress in economic practice of China’s efforts to build a large domestic market at this stage.

Although some scholars have paid attention to the trade potential of the domestic markets [40,41], the literature makes little effort to explore the change trend of the trade proportion structure in the domestic market from the perspective of DE development. Studies on cross-border e-commerce contribute sufficient evidence to enhance cross-border e-commerce and foreign trade with the development of digitalization and information technology, but they pay little attention to the independence of trade activities in the domestic markets. Under the background of the new pattern of “double circulation” of China’s economy due to the opening-up of large domestic markets, DE construction has become a national development strategy. Research on the trade-promoting effect of the DE on DNSs holds strong practical significance and can serve as a theoretical reference for policy-makers and theoretical researchers.

2.2. Hypotheses

For China, the proportion of the DE in the GDP is increasing year by year. The transformation of traditional trade into digital trade has become an established trend [42]. The digital transformation brings pros and cons to firms. On the one hand, it can reduce the trade risks caused by uncertainty and information asymmetry [27] and enable enterprises to reduce costs and improve their ability to operate profitably. On the other hand, it will also bring enterprise risks such as the failure of industrial chain and supply chain remodeling as well as the risk of choosing the wrong path for digital transformation technology [43].

First, the DE not only plays an important role in realizing international trade and openness but also strongly promotes trade liberalization in the domestic markets [44]. There are three kinds of approaches to enlarge the local trade volume: (i) digital technologies make information acquisition possible and easy, reducing the search costs incurred by both firms and customers. Thus, companies can receive imported immediate products to produce and process at a lower price, while customers are allowed to purchase at a cheap price, leading to greater provincial trade opportunities [45]; (ii) the appearance of emerging technologies, allowing for resource circulation and the provision of customized products between different regions, has become possible [46], contributing to breaking down market barriers [47], and further improving domestic integration [48]; and (iii) the DE ties provinces up by meeting the specialized needs of customers, upgrading the whole industrial chain and providing increasing opportunities for cooperation. Accordingly, this paper proposed the following hypothesis.

H1. 

DE development has significant trade creation effect on DNS.

Second, DE development and the regional differences in the domestic trade market have a continuous historical path, and the basis of early development has serial correlation with the level of subsequent development. Thinking about it from another dimension, China has a vast territory, there are significant regional differences in the quality of economic growth, and the spatial correlation of the domestic trade marketization level is obvious. Therefore, it is reasonable to infer that while the domestic markets are spatially converging, there is a strong temporal correlation between the digital economic infrastructure and economic activity entities in different regions. In view of this analysis, this paper proposes the following hypothesis.

H2. 

The expansion effect of DE development on DNSs has temporal and spatial correlations.

Third, the deepening revolution of the DE closes the trade links between the provinces, and every region can give full play to its competitive advantages and factor endowment to realize a unified domestic market [40]. Digital technologies allow them to realize the sharing of resources and information, bringing benefits to every participant [2]. On this basis, this paper proposes the following hypothesis.

H3. 

DE development can upgrade regional marketization to improve the local non-tradable sector.

Fourth, from the perspective of spatial differences, prefecture-level cities in mainland China can be divided into three regions based on the traditional geographical distribution of the eastern, central and western regions, with different degrees of development. Compared with the central and western regions, the eastern region has a better historical foundation for DE development, a developed economy, and a high degree of trade openness [38]. Therefore, the eastern region is more willing to open its market to other parts of China and develop domestic circulation. An analysis of inter-provincial samples shows that the development of the DE in China is dominated by the eastern region (represented by the Beijing-Tianjin-Hebei region, the Yangtze River Delta and the Pearl River Delta), supplemented by the central and western regions, and the degree of market openness in the eastern and western regions is also different. In light of this situation, this paper proposes the following hypothesis:

H4. 

The DE has more significant promotion effect on DNS in the east than the central and western regions.

3. Methodology, Data and Variables

3.1. Methodology

We followed the practice of Tang et al. [49]. First, this paper sets a panel effect regression model which shows the obvious advantages to control multidimensional fixed effects of sample characteristics. The DE of 285 cities was taken as the explanatory variable, domestic trade openness was taken as the explained variable, which refers to the practice of CHEN et al., [41], and total social retail sales, the financial deepening level and FDI were taken as the control variables using the selection criteria of Tu et al., and Lyu et al., for reference [50,51]. In addition, this paper refers to Yang and He’s practice to adopt the entropy weight method to calculate the urban DE (DEI) [52]. At the same time, in view of the possible temporal correlation of samples, this paper further adopted the dynamic panel GMM method for regression estimation as was done by Nepal et al. [53]. The model is constructed as follows:

$$Domestictrad{e}_{it}={\beta }_0+{\beta }_{1}DE{I}_{it}+{\beta }_2{X}_{it}+{\theta }_t+{\mu }_i+{\epsilon }_{it}$$

(1)

$i$ stands for cities; $t$ represents the years; $Domestictrad{e}_{it}$ means domestic trade openness (Ln_DomE); ${\beta }_0$ denotes the constant term; $DE{I}_{it}$ represents the development level of DE; $X_{it}$ represents the control variables at other levels; ${\beta }_1$, ${\beta }_2$ represent the marginal effect of each control variable on domestic trade openness, respectively; ${\mu }_i$ represents the fixed effects of the DE on domestic trade openness; ${\epsilon }_{it}$ is the random disturbance term [54].

To enhance the robustness of the core results, we introduce a policy shock variable, the National Big Data Comprehensive Pilot Zone (NBDCPZ) as the digital improvement factor, and employ the traditional DID model and latest expenditure synthetic difference-in-differences (SDID) model [55] to identify the causal relationship between DE development and DNS growth. As a reformative method of standard DID, the SDID model formula (2) was constructed as follows:

$$Domestictrad{e}_{it}={\beta }_0+{\beta }_{1}Trea{t}_i\times Perio{d}_t+{\beta }_2{X}_{it}+{\theta }_t+{\mu }_i+{\epsilon }_{it}$$

(2)

Based on the hypothesis that there is a temporal correlation between the main explanatory variables, the impact of the DE development level on the domestic market’s variables may have a time-lag effect, following methodology of Acemoglu et al. [56]. This paper further takes the lagged DE development level as the explanatory variable to observe whether the historical level of DEI has formed the scale expansion of the non-tradable sector in the current domestic markets [54]. Therefore, we constructed a lag model of the digital development index, as shown in formula (3).

$$Domestictrad{e}_{it}={\beta }_0+{\beta }_{1}L.DE{I}_{it}+{\beta }_2{X}_{it}+{\theta }_t+{\mu }_i+{\epsilon }_{it}$$

(3)

In addition to these, we considered the spatial correlations between the nearby cities in geographic distance to address these issues. Therefore, this paper constructed SAR, SDM, SAC and SEM spatial models to choose a suitable explanation for the spatial nexus of the key variables. Formulas (4)–(7) are established as follows:

$$Domestictrad{e}_{it}={\beta }_0+{\beta }_1{W}_{d}Domestictrad{e}_{it}+{\beta }_{2}DE{I}_{it}+{\beta }_3{X}_{it}+{\epsilon }_{it}$$

(4)

$$Domestictrad{e}_{it}\begin{array}{}\\ ={\beta }_0+{\beta }_1{W}_{d}Domestictrad{e}_{it}+{\beta }_2{W}_{d}DE{I}_{it}\\ +{\beta }_{3}DE{I}_{it}+{\beta }_4{W}_d{X}_{it}+{\epsilon }_{it }\end{array}$$

(5)

$$Domestictrad{e}_{it}={\beta }_0+{\beta }_{1}DE{I}_{it}+{\beta }_2{X}_{it}+{\theta }_t+{\mu }_i+W_d{\epsilon }_{it}$$

(6)

$$Domestictrad{e}_{it}={\beta }_0+{\beta }_1{W}_{d}Domestictrad{e}_{it}+{\beta }_{2}DE{I}_{it}+{\beta }_3{X}_{it}++W_d{\epsilon }_{it}$$

(7)

3.2. Data

This paper selected the national urban database from 2010 to 2019. When processing the data, we found that the variables of some cities in western China, such as Gansu Province and the Tibet Autonomous Region, were seriously missing; thus, referring to the article of He et al. [39], provincial city samples with incomplete data were removed. Finally, a panel dataset containing 285 cities at the prefecture level and above were used for analysis. Among them, the missing values of variables for some cities were filled in by the exponential smoothing method and moving average method. All the data in this paper were from the China Urban Statistical Yearbook.

Table 2. Descriptive statistics of the variables.

Variables	Indicators’ Full Name	Obs	Mean	Std. Dev.	Min	Max
Ln_DomE	Logged Domestic Economy	2668	25.5237	0.7932	23.4755	27.7435
DEI	Digital Economy Index	2668	0.0996	0.1011	1.92 × 10−9	1.0000
GPCA	DEI by General Principal Component Analyses	2613	0.0646	0.0846	−3.68 × 10−9	0.8660
DigiFin	Digital Finance Index	2357	0.4542	0.2055	0.0077	1.0000
LnUrbanPop	Logged Urban Population	1889	15.1025	0.6753	12.2165	18.4892
LnFix	Logged Fixed Investment	1889	25.2999	0.8456	22.0046	28.0607
LnFDI	Logged Foreign Investment	2521	21.0577	1.7903	12.2192	26.0447
Lnsaving	Logged Local Resident Saving	2668	25.9383	0.9657	23.7112	29.1557
LnAGDP	Logged Average GDP	2668	10.6089	0.5714	8.7729	12.4564
LnPopdensity	Logged Popular Density	1617	5.7171	0.9004	1.6094	7.8816
MPS	Marketization Process Score [57]	2668	6.7143	1.6754	−1.42	11.4
GM	Government market	2668	6.1376	1.4932	−14.29	9.22
NSE	Non-state Economic	2668	7.9654	2.1565	1.19	11.8
PMD	Product Market Development	2668	8.2572	0.9428	1.31	9.79
FMD	Factor Market Development	2668	5.6782	1.9439	−0.66	15.28
IAD	Intermediary Agent Development	2668	6.3762	4.2557	−0.41	24.33

presents the descriptive statistics of the original values of the main variables in this paper. The value of RMB is deflated by the GDP deflator in the 2010 base period, and the value of US dollars is denominated in 2010 constant US dollars and converted into RMB based on the exchange rate of that year. The results showed that the observed values of the variables are characterized by a large mean and large standard error, which indicates that there are great differences in foreign investment, the urbanization level, trade liberalization and other aspects. All the observed values of the numerical variables were logarithmically transformed before continuing the econometric analysis to maintain the stability of the data.

3.3. Variables and Data

3.3.1. Explanatory Variable

According to the method of Yang and He [52], this paper used the entropy weight method (EWM) to measure the DE and formulate the DE development level index at the prefecture level, denoted as the DEI. In the literature, most comprehensive evaluation indicators of the DE are based on provincial-level data. Some studies focus on differentiating between the efficiency of the DE development level and the regional structure. Other studies establish multi-level indicators for measurement based on digital infrastructure, digital industry development, the digital governance level, and other dimensions.

This paper considered the availability of relevant data at the city level, taking digital infrastructure as the main observation perspective, and used Zhang et al. [58] and Cao et al. [59] as references for the selection of indicators to construct a DE index system. Specifically, the number of broadband internet access users per 10,000 households, the number of computer service and software employees, the total revenue of telecom businesses in a city, the number of mobile phone users per 10,000 households at the end of the year and the digital financial inclusion index were used to investigate the DE development level [60,61].

Referring to Lu and Fang’s [62] processing methods for the main indicators of the DE, this paper adopted the entropy weight method for calculation, and the specific steps are as follows:

(1)

Standardization of the original data of each indicator.

Since there are certain differences in the nature, dimension, magnitude, and other characteristics of the indicators of the urban DE, the original data were standardized in order to eliminate the influence of these differences on the outcome variables. We developed the measurement of indicators around five aspects: regional trade, the digital financial inclusion index, the number of internet broadband access users per 10,000 households in a city, number of computer services and software employees, the city’s telecommunications business revenues and the number of mobile phone users per 10,000 households in the city at the end of the year. These five indicators are positive such that the higher the index value is, the better the regional trade performance is. Following the practice of Roller and Waverman [63], the value of index j of prefecture-level city i. Equation (8) was used to obtain the standardized index value, which is the original data of the positive index, normalized to:

$$X_{ij}=\frac{{\alpha }_{ij}-\min \left({\alpha }_{ij}\right)}{\max \left({\alpha }_{ij}\right)-\min \left({\alpha }_{ij}\right)}$$

(8)

The determination process of the index weight based on the entropy method is as follows:

(2)

Calculate the proportion of the index $j$ in city $i$:

$$\begin{array}{l}{P}_{ij}=\frac{x_{ij}}{{\displaystyle {\sum }_{i=1}^m{x}_{ij}}}\\ \end{array}$$

(9)

Calculate the entropy value and weight of the index $j$:

$$e_j=\frac{-1}{Inm{\displaystyle \sum _{i=1}^n{p}_{ij}}}$$

(10)

$${\omega }_j=\frac{1-e_j}{{\displaystyle {\sum }_{j=1}^m(1-e_j)}}$$

(11)

Carry out comprehensive score measurement:

$$s_j={\displaystyle \sum _{j=1}^n{\omega }_j{\alpha }_{ij}}$$

(12)

In Equation (8), is the original data of each index in the digital economic index system, and ${\omega }_j$ is the weight of each index given by the entropy value method. Clearly, the larger the value of $S_i$, the better the sample effect is, indicating the higher level of regional DE development.

After deleting the cities with incomplete data, this paper used the entropy weight method to calculate the digital economic development index of 285 cities at prefecture-level in China from 2010 to 2019. The data on the development level of the DE were from the Statistical Yearbook of Chinese Cities, and the moving average method was used to fill in the missing values.

To intuitively uncover the differences between regions, we also classified city samples based on the eastern, central, western, and north-eastern regions and made a scatter plot. As shown in the scatter plot below (Figure 1), China’s DE is mainly developed in the eastern region, supplemented by the central, western, and north-eastern regions. Most cities in the eastern coastal areas, such as Shanghai, Beijing, Dongguan, and Nanjing, are among the leading cities in DE development. In the central and western regions, Chongqing, Chengdu, Wuhan, Xi’an and other cities have great potential for DE development, with the provincial capital cities as the center, and the surrounding cities forming a positive DE interaction. The figure shows that the DE index in northeast China is lower than that in the rest of the regions, and the development power of the DE in northeast China needs to be strengthened.

3.3.2. Explained Variables

Existing studies often represent the openness of a region based on three dimensions: the economy, technology, and society. Referring to the practice of Hlatshwayo and Spence [64], this paper used the remaining proportion of the GDP to represent the amount of the non-tradable sector of a region, excluding the total volume of foreign trade.

3.3.3. Control Variables

To accurately measure the relationship between the DE and domestic trade liberalization, this paper refers to the practice of Li et al. [65] to set variables that may have an impact on domestic trade. The details are as follows: the level of financial deepening is represented by the ratio between the sum of the balance of deposits and loans of financial institutions at the end of the year and regional GDP; the urbanization level is measured as the ratio of the urban population to the total population; and FDI is expressed as the actual amount of foreign investment in each city. Other control variables include per capita GDP, fixed asset investment, and population density.

3.3.4. Mediator Variables

We employed the marketization process score (MPS) which was indexed by Fan and Wang [57], which could shed light on the potential moderate approach that the DE promotes DNSs through the lower barriers of trade and greater activation of local markets. It is composed of four sub-indicators that are listed at the bottom of

Table 3. The categories and data sources of variables.

Variables	Variables	Indicators’ Full Name	Meature Units	Data Resource
Explantory variables	DomE	Domestic Economy	index	Calculated by author
Explained variables	DEI	Digital Economy Index	index	Calculated by author
Substitutable explained variables	GPCA	DEI by GPCA method	index
	DigiFin	Digital Finance Index	index	Peking University
Control variables	LnUrbanPop	Logged Urban Population	people	Statistical Yearbook of Chinese Cities
	LnFix	Logged Fixed Investment	yuan
	LnFDI	Logged Foreign Investment	2521
	Lnsaving	Logged Local Resident Saving	yuan
	LnAGDP	Logged Average GDP	yuan
	LnPopdensity	Logged Popular Density	people/square kilometre
Mediator variables	MPS	Marketization Process Score	index	Wind database
Sub-indicators of MPS	GM	Government market	index
	NSE	Non-state Economic	index
	PMD	Product Market Development	index
	FMD	Factor Market Development	index
	IAD	Intermediary Agent Development	index

.

4. Empirical Results

4.1. Panel Fixed Effect Model and Baseline Regression

This paper first used the baseline regression estimation of the panel fixed effects model. The Hausman test shows that FE model is prefer than RE model, we accept that fixed effect model is prefer to this estimation. After controlling for multiple year, city, and province fixed effects,

Table 4. Baseline regression estimation of multidimensional fixed effect.

	(1)	(2)	(3)	(4)
Variables	Ln_DomE	Ln_DomE	Ln_DomE	Ln_DomE
DEI	0.4106 ***	0.1484 **
	(0.0972)	(0.0586)
GPCA			1.9821 **
			(0.8126)
DigiFin				−110.1385
				(72.0659)
LnUrbanPop		0.0080	−0.7098	−0.7079
		(0.0131)	(0.7565)	(0.7679)
LnFixInvest		0.1804 ***	−12.7926 *	−15.2648 **
		(0.0188)	(7.4283)	(6.8676)
LnFDI		−0.0080	1.4027	1.9062
		(0.0056)	(1.2002)	(1.6735)
Lnsaving		0.0253	−11.2728	−15.3181
		(0.0539)	(20.5953)	(28.5303)
LnAGDP		0.1826 ***	−10.5952	−8.3144
		(0.0278)	(7.0091)	(7.1689)
LnPopdensity		0.2586 **	−33.2252	−43.5929
		(0.1121)	(58.1288)	(75.6157)
City FE	Y	Y	Y	Y
Year FE	Y	Y	Y	Y
Province FE	Y	Y	Y	Y
Cluster	Province	Province	Province	Province
Constant	25.5585 ***	16.8735 ***	24.7614 ***	1141.5087
	(0.0110)	(1.4699)	(0.3680)	(1,288.3858)
Observations	2306	1403	2027	1316
R-squared	0.9640	0.9850	0.9650	0.5908

Robust standard errors in parentheses; *** p < 0.01, ** p < 0.05, * p < 0.1.

reports the results of panel regressions of the DEI on local trade sector growth. Table 4 column (1), as the reference, is the baseline regression result without adding any control variables, and its estimated coefficient is 0.4106, which is significantly positive at the 1% level, preliminarily indicating that an increase in the DE development level will promote the development of domestic trade. Column (2) shows the regression results of the control variable sets, including the local urbanization level, the gross domestic product at the end of the year (GDP), the resident population of a city at the end of the year (POP), fixed asset investment, and the financial deepening level. After controlling for other possible impact factors, the regression coefficients of the DE development level drop to 0.1484 and are significantly positive at the 5% level, meaning that the level of DE development has a promoting effect on domestic trade. Hypothesis 1 is proven. The digital economy development could be the new economic growth point, either the digital product industrial or the digitalization of traditional industries play a crucial role of local economy increment.

4.2. Robustness Test

4.2.1. Explanatory Variables Replace Test

Robustness tests verify the reliability of the basic estimation results by replacing the explanatory variables. To test whether measuring the core explanatory variable by employing the entropy weight method is reasonable, we first drew on Liu’s [66] research and used global principal component analysis (GPCA) to re-calculate the DE, and the results showed that the regression coefficient is positive and significant at the 5% level in column 3 of Table 4. Then, we also considered whether the popularity of digital finance contributes to the effect. Therefore, we adopted the digital finance index (DFI) of Peking University to replace the DEI constructed by the entropy weight method in column (4). The regression results showed that the impact of digital finance development on the level of the local economic volume is not significant.

The fluctuations of the above process indicate that there may be multiple first-level sub-index systems involving different dimensions in the formulation of the DE index instead of a synthesized index system under a single dimension. Therefore, it is more reasonable to construct the DEI based on pure digital infrastructure. In future research, this paper will use the pure digital infrastructure indicators of the DE index to carry out the analysis. Hypothesis 1 is verified.

4.2.2. Synthetic Difference-in-Differences (SDID)

In this paper, the difference-in differences (DID) method was further employed to solve the biased estimation caused by endogeneity. From 2016 to 2017, the National Big Data Comprehensive Pilot Zone (NBDCPZ) was launched in 10 provinces, autonomous regions, and municipalities: Guizhou, Inner Mongolia, Shanghai, Henan, Chongqing, Shenyang, the Beijing-Tianjin-Hebei region and the Pearl River Delta. The policy is taken as a quasi-natural experiment to identify the causality of DE and domestic trade development [67], the policy variables are set as BDPZ, the time node is 2017, and the sample cities in the treatment group consist of 26 cities above the prefecture level located in the administrative area of the NBDCPZ. The endogenous influence of the original fixed effect model can be mitigated by the policy shock. However, traditional DID and PSM-DID model estimations are nonsignificant in this case. To reduce the ATT distortion of the local average treatment effect (LATE) caused by the average weight distribution of the control group samples by the DID method, the SDID estimation was used to carry out the robustness test [55]. The weight distribution and the differences between groups are shown in Figure 1.

The SDID estimation results are shown in

Table 5. Synthetic Difference-in-Differences Estimator.

Ln_DomE	ATT	Std. Err.	t	p > |t|	95% Conf.	Interval
Treatment	0.03901	0.00864	4.51	0.000	0.02207	0.05595

. We observe that the ATT coefficient is positive at 0.039 and statistically significant at the 5% level, indicating that the DE development policy of the NBDCPZ can bring a 0.039 standard deviation increase in the level of domestic trade sector growth. The result above tries to eliminate the influence of endogeneity, and it more fairly reflects the improvement in domestic trade market liberalization brought by the development of the DE. The growth period of DNSs is lengthy, but the process is accelerated by the deepening of the DE. Not only the full-sample regression of all cities but also the NBDCPZ policy effect evaluation of pilot cities illustrates the significant improvement in the local non-tradable economy after the strategy of DE development is implemented.

4.2.3. Instrument Variable (IV) Method

To further alleviate the interference of endogeneity in the results of this paper, we employed external instrumental variables. Based on the treatment method of Chen Q. [68], the panel instrumental variable method with fixed effects was adopted to further eliminate the endogeneity problem. It is customary to perform the first difference and then 2SLS regression using instrumental variables (DigiFin, Ln_TelRevenue, Ln_IAP). The level of digital financial inclusion popularization, broadband Internet access and total revenue of telecom services in each sample city were used as instrumental variables to measure the real impact of the digitalization level on domestic trade activities. These instrumental variables pass the Sargan test with p value = 0.6721, rejecting the null hypothesis of overidentification, and p value > 0.1 in the C-statistic test obeying the chi-square distribution (0.4914; 0.5343; 0.6742, respectively), accepting the null hypothesis of the exogeneity of the variables. In the further panel overidentification test, the Hansen statistic = 1.449 and p value = 0.4845, indicating that the overidentification hypothesis, was rejected. The first-stage F-statistic = 34.888 is larger than the key value of the F-statistic of 12.83 suggested by Stock, Wright and Yogo [69], which excludes the problem of weak instrumental variables.

Table 6. Benchmark of IV estimation.

	(1)	(2)	(3)	(4)	(5)	(6)
Variables	OLS FE	OLS FD	FirstSLS	TSLS	LIML	GMM
DEI	0.417 ***	0.326 ***	0.428 ***	0.818 ***	0.818 ***	0.812 ***
	(0.097)	(0.073)	(0.042)	(0.236)	(0.236)	(0.235)
LnUrbanPop	0.008	0.005	0.005	−0.005	−0.005	−0.005
	(0.009)	(0.005)	(0.005)	(0.012)	(0.012)	(0.012)
LnFix	0.174 ***	0.099 ***	−0.001	0.135 ***	0.135 ***	0.135 ***
	(0.023)	(0.020)	(0.009)	(0.022)	(0.022)	(0.022)
LnFDI	−0.005	0.005	−0.001	−0.002	−0.002	−0.002
	(0.005)	(0.005)	(0.002)	(0.005)	(0.005)	(0.005)
Lnsaving	0.400 ***	0.349 ***	−0.068 **	0.225 ***	0.225 ***	0.226 ***
	(0.041)	(0.078)	(0.030)	(0.068)	(0.068)	(0.068)
LnAGDP	0.253 ***	0.149 ***	0.017	0.132 ***	0.132 ***	0.132 ***
	(0.072)	(0.039)	(0.011)	(0.027)	(0.027)	(0.027)
LnPopdensity	0.260 *	0.059	0.013	0.151	0.151	0.151
	(0.152)	(0.107)	(0.047)	(0.114)	(0.114)	(0.114)
FE	Y	N	N	N	N	N
FD	N	Y	Y	Y	Y	Y
Constant		0.033 ***	−0.035 ***	0.035 ***	0.035 ***	0.035 ***
		(0.010)	(0.004)	(0.009)	(0.009)	(0.009)
Observations	1540	1267	996	996	996	996
R-squared	0.808	0.177	0.024	0.024	0.024	0.026
Number of id1	266	266	259	259	259	259

Robust standard errors in parentheses; *** p < 0.01, ** p < 0.05, * p < 0.1.

shows the results of IV estimation. Column (1) successively illustrates the OLS regression of the DEI with fixed effects (FE), and Column (2) provides the first difference (FD) result of OLS to operate the panel IV regression process. Both estimations are significant at the 1% level, but the coefficient of the FD estimation is lower than that of FE estimation such that the FD method reduces the overidentification of panel IV. We use the FD mode to estimate the first stage result in Column (3) and the regression results of the two-stage least squares (2SLS) method in Column (4). Columns (5) and (6) show the different results of least information maximum likelihood (LIML) and generalized method of moments (GMM) estimation; the coefficients rise to approximately 0.818 with the expansion effect. The endogeneity test results showed that GMM estimation is better than the 2SLS method. After passing the Hausman test and DWH heteroscedasticity test, part of the endogeneity effect is reduced. The IV effectively identifies the causal effect of the DE on the domestic non-trade volume.

5. Further Analysis

To further disclose the local economic expansion effect of digitalization, this section analysed the temporal and spatial relationships of the DEI and DNSs. The possible mechanism can be explored by a mediation effect model, and the regional heterogeneity of different cities plays a crucial role in the impact of the digital transformation.

5.1. Lag Effect of the DEI

The possible bias to the validity of the estimates reported comes from the presence of time-varying former DE development factors that later affect domestic non-trade sectors, although the city and province fixed effects absorb the time-invariant factors. To tackle this issue, we use a lag of one or two phases of the DEI and separately estimate the effect. The results are illustrated in

Table 7. Lag Effect of DEI.

	(1)	(2)	(3)	(4)
Variables	Ln_DomE	Ln_DomE	Ln_DomE	Ln_DomE
DEI	0.1484 *			−0.0437
	(0.0780)			(0.1611)
DEI, first lag		0.1661 ***		0.1267
		(0.0617)		(0.1252)
DEI, second lag			0.1874 ***	0.1749 **
			(0.0575)	(0.0770)
LnUrbanPop	0.0080	0.0003	0.2403	0.2370
	(0.0115)	(0.0135)	(0.1459)	(0.1747)
LnFix	0.1804 ***	0.1572 ***	0.1873 ***	0.1880 ***
	(0.0454)	(0.0231)	(0.0248)	(0.0408)
LnFDI	−0.0080	−0.0064	−0.0039	−0.0039
	(0.0080)	(0.0064)	(0.0068)	(0.0069)
Lnsaving	0.0253	0.0464	0.2295 ***	0.2299 *
	(0.1281)	(0.0691)	(0.0821)	(0.1165)
LnAGDP	0.1826 *	0.2659 ***	0.1108 ***	0.1077
	(0.0914)	(0.0395)	(0.0399)	(0.0703)
LnPopdensity	0.2586	0.4499 ***	0.3209 **	0.3211 **
	(0.2054)	(0.1342)	(0.1421)	(0.1484)
City FE	Y	Y	Y	Y
Year FE	Y	Y	Y	Y
Province FE	Y	Y	Y	Y
Cluster	Province	Province	Province	Province
Constant	16.8735 ***	15.0384 ***	8.2344 ***	8.2764 **
	(2.9879)	(1.8327)	(2.7656)	(3.9903)
Observations	1403	1170	937	937
R-squared	0.9850	0.9853	0.9885	0.9885

Robust standard errors in parentheses; *** p < 0.01, ** p < 0.05, * p < 0.1.

. Columns (1)–(3) show the same estimation results with similar coefficients and significance levels, and they employ the original DEI, the first-lag DEI and the second-lag DEI, respectively. The prolongation of the DEI lag period leads to a gradually increasing influence coefficient from 0.1484 to 0.1874. Column (4) includes all lag DEI factors, and the estimation shows that the second-lag DEI plays an important role in the time-varying effect of local non-tradable sector expansion. Hypothesis 2 is verified. Such digital economy construction should benefit the future growth of economy, at least two period lagging effects were discovered by the dynamic analysis in our research.

5.2. Spatial Correlation

Spatial correlation should be considered because neighboring cities may affect the sample cities in terms of many macroeconomic factors [70]. Due to the geographic distance between nearby cities, they not only share access to public digital infrastructures but also participate in the local trade market of goods and production factors. In the cyber age, the spreading of digital content has a distinct regional synergy. It breeds a larger local consumer market, forms the economic value of big data, and creates possibilities for the input of data during the production process. Therefore, we constructed an inverse distance matrix of 283 cities to match the sample city dataset and employ a series of spatial correlation models to estimate the spatial effect of DNS growth and DE development.

We performed a global Moran’s I test on the panel spatial autocorrelation of the model, while the results of the yearly Moran’s I of the dependent variables and independent variables are listed in Appendix A

Table A1. Skewness and kurtosis tests for normality (Joint test).

Variable	Obs	Pr (Skewness)	Pr (Kurtosis)	Adj-Chi2 (2)	Prob > Chi2
Ln_DomE	2710	0.0000	0.0000	139.88	0.0000
DEI_N1	2829	0.0000	0.0000	.	.
LnUrbanPop	1979	0.0000	0.0000	95.25	0.0000
LnFix	1979	0.3902	0.1060	3.34	0.1881
LnFDI	2662	0.0000	0.0000	125.56	0.0000
Lnsaving	2828	0.0000	0.0012	149.50	0.0000
LnAGDP	2829	0.0067	0.9958	7.32	0.0258
LnPopdensity	1696	0.0000	0.0000	194.30	0.0000

and

Table A2. Shapiro–Wilk W test for normal data.

Variable	Obs	W	V	z	Prob > z
Ln_DomE	2710	0.98839	18.148	7.454	0.00000
DEI_N1	2829	0.64635	574.514	16.362	0.00000
LnUrbanPop	1979	0.97360	31.009	8.733	0.00000
LnFix	1979	0.99781	2.567	2.397	0.00826
LnFDI	2662	0.98439	24.004	8.168	0.00000
Lnsaving	2828	0.97086	47.325	9.933	0.00000
LnAGDP	2829	0.99795	3.336	3.103	0.00096
LnPopdensity	1696	0.94650	54.608	10.113	0.00000

Note: The normal approximation to the sampling distribution of W’ is valid for 4 ≤ n ≤ 2000.

. As shown, both types of variables have a spatial correlation in the Moran’s I test.

We carried out several tests to check the applicability of the spatial correlation models. The Hausman test finds that chi2(7) = 116.95 and p value > 0.0000, and the null hypothesis is strongly rejected; thus, a fixed effect model is chosen for spatial estimation. The Wald test report shows that the SAC model or SDM is more adaptive for this research. Furthermore, the AIC and BIC were used for testing, and it was found that the AIC and BIC values of the error lag terms decrease. Clearly, the SAC model was more suitable for the situation in this paper.

Table 8. Spatial model estimation.

	(1)		(2)	(3)	(4)
	SDM		SAR	SAC	SEM
Variables	Main	Wx	Main	Main	Main
L.Ln_DomE	0.8663 ***		0.9800 ***
	(0.0184)		(0.0183)
L.WLn_DomE	−0.8881 ***		−1.7056 ***
	(0.2860)		(0.1168)
DEI	0.1175	−0.0087	−0.3729 ***	0.2106***	0.2221 ***
	(0.0856)	(0.5920)	(0.0831)	(0.0665)	(0.0670)
LnUrbanPop	−0.0077	−0.0711	−0.0213	0.0190	0.0167
	(0.0177)	(0.2302)	(0.0175)	(0.0224)	(0.0226)
LnFix	−0.0209	0.2171 *	−0.5598 ***	0.1605 ***	0.1851 ***
	(0.0203)	(0.1108)	(0.0176)	(0.0193)	(0.0200)
LnFDI	−0.0105 **	0.1925 ***	−0.0670 ***	0.0072	0.0073
	(0.0044)	(0.0485)	(0.0043)	(0.0051)	(0.0052)
Lnsaving	0.0591	0.2914	−0.0280	0.0921 **	0.0935 **
	(0.0406)	(0.1879)	(0.0399)	(0.0464)	(0.0473)
LnAGDP	0.1382 ***	0.3121 *	−0.2223 ***	0.2708 ***	0.2883 ***
	(0.0226)	(0.1683)	(0.0209)	(0.0247)	(0.0254)
LnPopdensity	−0.0767	1.0289	−0.2043	0.1316	0.1819
	(0.1423)	(1.4763)	(0.1390)	(0.1551)	(0.1565)
Spatial rho	0.0286		9.3548 ***	0.6278 ***
	(0.2855)		(0.1513)	(0.1021)
sigma2_e		0.0294 ***	0.0276 ***	0.0497 ***	0.0451 ***
		(0.0007)	(0.0007)	(0.0012)	(0.0012)
lambda				−0.0914	0.5158 ***
				(0.2251)	(0.1225)
Observations	2538	2538	2538	2820	2820
R-squared	0.3564	0.3564	0.0135	0.1813	0.6506
Number of id2	282	282	282	282	282

Robust standard errors in parentheses; *** p < 0.01, ** p < 0.05, * p < 0.1

reports the overall results of the spatial econometric models. Columns (1)–(4) show the SDM, SAR, SAC and SEM estimates, respectively. We preferred to use Column (3) of the SAC model to explain the spatial correlation. It illustrated that the DEI coefficient is 0.2106 and significant at the 1% level, the spatial rho coefficient is 0.0.6278 and significant and the 1% level, and the spatial effect of the residual is significant. The SAC model estimation showed that the development of the DE, from either the city itself or neighboring cities, can improve the local expansion of the non-tradable sector scale. Hypothesis 2 is verified.

The temporal and spatial nexus of the DE and DNS’s is important for understanding the mechanism through which the DE drives domestic economic growth. On the one hand, temporal correlation analysis showed that the lagging DEI could play a significant role in DNS growth. Approximately two periods of lagging DE development should contribute to local market integration through the advancement of digital infrastructures and the application of digital content (Table 8). On the other hand, spatial correlation analysis explains the impact mechanism of neighboring cities for this topic. Nearby areas can share DE development goals through easier digital content access from the main city, which takes the lead in the DE. The externality of DE development obliges the central government to advocate that local governments should construct the DE as the national strategy.

5.3. Mediation Effect

We focused on the increase in the marketization level as the approach by which the DE level improves DNS growth. Referring to Fan et al. [71], we employed the marketization index score system with the total score and five sub-indexes, which are provided by Wang et al., to describe different perspectives on the local marketization of the sample cities [57]. Unfortunately, the total score and the other four sub-indexes did not report statistically significant results, although in this regard, the product market development index (PDM) is an exception.

Table 9. Moderate effects of marketization process.

	(1)	(2)	(3)
Variables	PDMKTI	Ln_DomE	Ln_DomE
DEI, second lag	0.0398 *		0.1987 ***
	(0.0227)		(0.0573)
PDM		0.3931 ***	0.2833 ***
		(0.0912)	(0.0966)
LnUrbanPop	0.0786	0.0073	0.2180
	(0.0575)	(0.0130)	(0.1453)
LnFix	0.0161 *	0.1740 ***	0.1827 ***
	(0.0098)	(0.0188)	(0.0247)
LnFDI	0.0029	−0.0093 *	−0.0047
	(0.0027)	(0.0056)	(0.0068)
Lnsaving	0.1793 ***	−0.0468	0.1787 **
	(0.0323)	(0.0551)	(0.0834)
LnAGDP	0.0716 ***	0.1711 ***	0.0905 **
	(0.0157)	(0.0279)	(0.0403)
LnPopdensity	0.0431	0.2716 **	0.3331 **
	(0.0560)	(0.1116)	(0.1414)
City FE	Y	Y	Y
Year FE	Y	Y	Y
Province FE	Y	Y	Y
Cluster	Province	Province	Province
Constant	6.0038 ***	18.6735 ***	9.9354 ***
	(1.0895)	(1.5005)	(2.8109)
Observations	937	1403	937
R-squared	0.9234	0.9852	0.9886

Robust standard errors in parentheses; *** p < 0.01, ** p < 0.05, * p < 0.1.

presents the mediating effect. Column (1) reports that the second-lagged DEI can improve the PDM with a 3.98% standard error and significance at the 10% level. Column (2) shows that the PDM promotes DNS growth with a coefficient of 0.3931, although overidentified, and the result is significant at the 1% level. Column (3) reports the regression estimation including the DEI lagged two periods and the PDMKTI, both of which are significant at the 1% level. Therefore, the DEI can improve non-tradable sector growth by increasing the marketization level. Hypothesis 3 is verified. The new trade methods, such as e-commerce, ae opening up the local market access by terrain barrier reduction, and the digitalization of production activities has already shown the advantages during the organization of the product factors cross the regions.

Our mediation effect analysis provides a new answer by choosing the marketization level as the mediating variable to observe whether the DE can push China to form market integration and whether high-level market integration can further promote the non-tradable sector of the domestic economy. Market integration is a mediating method for releasing the power of the DE in the field of local non-tradable sector growth. Although the economic coverage of digital research has covered a wide range of disciplines and social fields, the empirical evidence on how to improve the domestic non-tradable economy is too limited for the DE. The previous literature discussed various perspectives of the mechanism of the issue, such as the impact of the DE on the environment and sustainable development [60,61,72], the impact of the DE on the efficiency of international trade and the types of import and export products [73,74], the digital transformation of enterprises and upgrading of the industrial structure [26,49], and the impact of digitalization on the flow of labor and capital [75,76,77].

Above all, product market development can become an important intermediary through which the DEI promotes the domestic nontrade sector (Table 8). Unfortunately, the other four digital sub-indexes and the total score were not significant. This means that deepening DE development still has a long way to go to reach the ideal status, and DNS growth should have great potential in the future. The heterogeneity analysis confirmed the findings above by disclosing that the cities belonging to the central and west regions are far away from eastern cities in terms of the tournament of DE development and DNS growth. The digital divide hinders the pace of market integration from the central and western regions to the eastern region.

5.4. Heterogeneity Analysis

With its vast territory, China has very different levels of local development. Therefore, it is necessary to analyze the regionally different effects of the DEI on DNSs. Following the existing research [62], we used the eastern-central-western standard to regroup the data categories and perform estimation with a panel FE regression model. As

Table 10. Heterogeneity analysis.

	(1)	(2)	(3)
	East	Middle	West
Variables	Ln_DomE	Ln_DomE	Ln_DomE
DEI	0.4082 **	0.3070	−0.0353
	(0.1517)	(0.6611)	(0.0968)
LnUrbanPop	0.0359 **	0.8478 ***	−0.0021
	(0.0119)	(0.1814)	(0.0125)
LnFix	0.0016	0.2281 **	0.1521 **
	(0.0746)	(0.1031)	(0.0531)
LnFDI	−0.0234	−0.1054 ***	−0.0038
	(0.0223)	(0.0312)	(0.0052)
Lnsaving	−0.2060 **	−0.1878	0.1321
	(0.0813)	(0.1306)	(0.2008)
LnAGDP	0.2102	0.9218 ***	0.2250
	(0.1207)	(0.1596)	(0.1355)
LnPopdensity	−0.1554	0.2412 ***	0.4542
	(0.1078)	(0.0675)	(0.3570)
Constant	20.5391 ***	−6.2845 ***	4.1335
	(3.1876)	(1.5425)	(3.5625)
City FE	Y	Y	Y
Year FE	Y	Y	Y
Province FE	Y	Y	Y
Cluster	Province	Province	Province
Observations	456	84	1018
R-squared	0.9750	0.9489	0.9818

Robust standard errors in parentheses; *** p < 0.01, ** p < 0.05.

shows, Columns (1)–(3) report the results for the cities belonging to the eastern, central, and western regions. Column (1) shows a positive effect with a coefficient of 0.4082 that is significant at the 5% level; in comparison, the other two groups are not statistically significant. It can be seen that only cities in the eastern regions benefit from DE development in the local non-tradable sector. Cities in the central and western regions cannot yet improve the non-tradable sector based on digital transformation. Hypothesis 4 is verified. The large gap between eastern cities and central-western cities brings the enough space for the latter, to play late-mover advantage in digital economy construction. On the other hand, the eastern cities could help the developing areas promote the digital economy in order to penetrate the local market.

5.5. Discussion

In this paper, we probed how DE development affects the scale of the non-tradable sector in the domestic market, which is an indispensable part of China’s aim to transform the enormous economic volume into a large market with internal circulation under the new “double circulation” pattern of development.

China is eager to enlarge total domestic market demand under the counter-global challenge. Policymakers hope that the domestic economy, especially the non-tradable sector, will benefit from the recent national policy of a unified market system. On 10 April 2022, an “opinion” clearly proposed to speed up the establishment of the national unified market system rules and break regional protectionism and market segmentation to speed up the construction of efficient, fair competition, and fully open up to a unified national market. Our findings provide a new approach to upgrade the inner demands of the local economy in developing areas.

The traditional perception holds that DE development will naturally drive an increase in bilateral trade between countries, but the improvement in the DE by a single country cannot change the level of its trade partners because of the national differences in the development of digital infrastructures. The gap in DE development for backwards trade partners shapes the bottleneck of international trade. In contrast, the DE transformation within a country can quickly improve the scale of DNSs, realizing the facilitation of digital trade and digital production activities, thus opening up a large regional market.

Our findings provide new evidence on the relationship between the DE process and the marketization of the domestic non-tradable economy. Prior to this paper, existing studies mainly focused on the effect of international trade openness. Some scholars have found that the Internet can intensify global trade, including the service trade and goods trade [78]. The Internet economy, as an earlier form of the DE, has pushed the traditional market to gradually turn to the online market [79]. Recently, many studies have focused on international trade in the cyber age [22]. For a powerful and prosperous nation, the traditional economy should be equally important to global trade. However, an insufficient number of papers offer clues from the domestic economic perspective, even though it naturally plays a crucial role in practice. Many policymakers in developing countries treat the DE as a strong development strategy to revitalize national strength. Currently, certain papers have considered both national trade and international trade when discussing the opportunities and challenges that the DE brings [80], and evaluations concentrating on the national level are still limited. Few studies concentrate on the impacts of the DE on the national non-tradable economy [81].

In this paper, we make several contributions to this field. First, the data selection, which is at the prefecture level, is more precise. The majority of the literature conducts explorations based on provincial data, which are more readily available; however, the provincial mean value usually causes less diversity in the data traits. In contrast, prefecture-level data reflect more characteristics and details in diverse regions with different development levels. Therefore, utilizing prefecture-level data can better capture the specific imperceptible differences in development. Second, our findings target how developing the DE can affect DNSs. According to our empirical results, the DE actively stimulates domestic economic activities by deepening marketization [82] and breaking internal trade barriers, which enriches the literature from a new perspective. Finally, we updated the identification strategy by employing the entropy weight method to measure the DE and by further changing the empirical methods, including SDID and IV models, to verify the robustness of the results. Furthermore, our research applies to not only China but also nearly all developing countries that have similar unbalanced levels of economic development in different regions and pay high attention to the DE.

Generally, narrowing the gap in the development of the DE and bridging the digital divide are important for central policymakers to consolidate the inner circulation of the economy because of the various differences in regional development. Neither the DE nor DNSs display the same differentiation level in eastern, central, and western China. Cities in the eastern region have distinguished abilities in digital transformation promoting DNSs. The digital transformation process sharply separates the traditional economy into different parts. However, DNSs still have many benefits from DE development in the cyber age. Those areas that adapt to digital development rapidly have more possibilities to capture opportunities for growth [83]. The promoting effect is more significant in the eastern regions than in the central and western areas. If cities in central and western regions intend to keep in touch with the tide of digital transformation, the central government and local municipalities should invest more funds and resources in developing the DE, helping local enterprises realize the digital transformation and improving the digital qualities of the local population.

6. Conclusions

In this paper, we attempted to determine the impact of the DE on DNSs based on prefecture-level city data from 2010 to 2019. We view the paper as the first step in calculating DEI adoption from the basic digital infrastructure indicators in China to observe the DE development level in 283 cities of China. We then chose DNSs as an indicator to estimate the local market scale without the international tradable sector and designed a methodology to investigate the impact of the DE on DNSs. Our research proved that DE development is significant for promoting DNS growth by utilizing a panel FE model, conducting a causal inference with the SDID method, and eliminating the endogeneity bias with the IV method. This paper further established mechanism models to analyse the lagged influence of the DEI and spatial correlation of the sample cities, and it was found that the mediation approach improved product market development to make the DE drive the growth in DNSs. Above all, our empirical work draws the following conclusions:

First, our primary results show the growth in DNSs affected by deepening of the DE in the past decade in China. Our findings showed that DE development plays a crucial role in DNS growth in China. Policymakers who proposed the “dual-circulation” new development pattern pay attention to implementing the digitalization strategy to establish a new engine of economic operation. One of the advantages of the DE is that enables local governments to integrate the domestic markets by breaking down barriers and providing more trade opportunities. The digitalization transformation should be an externality trend which breaks the barrier of local powers efficiently. The local incumbent has almost nothing to do except join the digitalization competition [84]. Our results enabled us to evaluate the positive effect of this endeavor, and the digitalization of the social economy provides a reasonable approach to achieve the target of economic inner circulation.

Second, we disclosed the temporal and spatial correlations between nearby cities in developing the DE and other macroeconomic factors that can improve DNSs. The spillover effect of the DE enables bordering cities to share the improvement effect from each other in DNSs. Municipalities should construct a comprehensive plan to develop the DE and assist neighboring cities in vigorously strengthening the development of the DE to realize win–win interprovincial trade.

Regions with a developed DE should also play a radiating and driving role, create broader spatial and temporal spillover effects, expand the scope of DE externalities, and seek new competitive advantages by enhancing regional marketization. The advanced digital economy cities have the natural preference of expending the digital coverage and deepening the degree to which they can promote their digital trade and production activities with the borders [85]. Policymakers should utilize the preference to reduce the gap in different areas.

How do the differences in DE development between regions change dynamically, and how does this change affect the growth of DNSs? Can the “low-end lock-in” risk in FDI [62] be avoided during the process of DE development? All these problems need further study and discussion.

Finally, we have provided the following suggestions:

(1)

Policymakers must pay attention to the impact effect of local non-tradable sectors under the digital economy development pattern, especially the product marketization level;

(2)

It is not the digital industry, but the digital economy infrastructure construction that is important to eliminate the gap of digital infrastructure between different areas, which are the responsibility of local government;

(3)

The advanced digital economy cities should help underdeveloped areas to promote the digital economy to establish the “win-win” situation, which needs the organization of local regimes, formally or informally;

(4)

The local government should construct one or two highly advanced digital economy cities to radiate out to neighboring cities when the public expenditure is limited to cover the whole construction duties of the region;

(5)

The urban digital economy benefits from the urbanization rate, so policymakers should synthetically consider digitalization and urbanization.