Digital Economy, Marine Industrial Structure Upgrading, and the High-Quality Development of Marine Economy Based on the Static and Dynamic Spatial Durbin Model

Abstract

Achieving marine sustainable development goals requires improving the quality of the marine economy. In this study, we constructed a multi-dimensional evaluation index system to quantify the high-quality development of the marine economy (HDME) and digital economy for China’s 11 coastal regions between 2012 and 2020, and systematically explored the mechanisms between the digital economy, marine industrial structure upgrading, and high-quality development of the marine economy. We further empirically analyzed the influence of the digital economy and marine industrial structure upgrading on the HDME by using the static and dynamic spatial Durbin model and threshold model. We found that the digital economy improved the HDME of local and neighboring regions, and the sub-dimensional indicators had a significant heterogeneous effect. The marine industrial structure upgrading only positively affected the quality of the marine economy in neighboring areas. Furthermore, the digital economy and marine industrial structure upgrading also had synergistic effects on improving the marine economy, which mainly depended on digital infrastructure and digital industrialization. There also existed digital infrastructure and digital industrialization thresholds within marine industrial structure upgrading, affecting the HDME. Our results provide new insights for the green and sustainable development of the marine industry and marine economy from a broader technical perspective in the digital age. Governments should recognize the spatial spillover and synergistic effects of the digital economy and marine industrial structure between different regions and implement policies considering their impacts to achieve the marine sustainable development goals.

1. Introduction

“Blue growth” is the new engine for national economic development [1]. The China Marine Economic Development Report 2022 point outed that China’s gross ocean product (GOP) reached 9462.8 billion yuan in 2022, which accounted for 7.8% of the gross domestic product. However, the marine economy encounters several problems in the current development process, including unbalanced speed and quality of growth, low efficiency of marine innovation, inadequate allocation of sea-related resources and factors, and insufficient high-end marine industries. Thus, under the increasing resource and environmental constraints, how to promote the transformation of the marine economic growth mode to improve its “quality” has been highly valued by governments. Accompanied by the continuous advancement of information technology revolution, the digital economy provides new opportunities for optimizing the marine industrial structure and achieving the effective improvement of “quality” and the reasonable growth of “quantity” of the marine economy. With the characteristics of high innovation, strong permeability, and wide coverage and sharing, the digital economy plays positive roles in technological innovation, factor allocation, and green transformation [2,3]. This highly conforms with the new high-quality development concept of “innovation, coordination, green, openness, and sharing” of the economy.

In addition to technological innovation, marine industrial structure upgrading is also an important driving force for promoting the marine economy. Marine industrial structure upgrading is conducive to improving the efficiency of resource allocation, reducing energy consumption and carbon emission intensity, and enhancing the degree of industrial cooperation and synergy. Particularly, the continuous integration of the digital economy and marine industries accelerates the diffusion and application of digital technologies between various production sectors. This significantly improves the flexibility, precision, and synergy of marine production and manufacturing, and promotes production efficiency and the optimization of the marine industrial structure. Thus, can the digital economy and the marine industrial structure upgrading, respectively, enhance the high-quality development of the marine economy? What is the mechanism behind this? Are there synergistic effects between them? Are there spatial effects? Is there a heterogeneity impact of the digital economy in terms of its different dimensions? Clarifying and answering the above questions is of great significance in promoting the transformation of marine industries to green, and intelligent and improving the quality of marine economic development.

With the support of digital information technologies, data elements have widely penetrated all links of production and life, continuously promoting the economy toward digitalization, intelligence, and networking [4,5]. Early studies about the digital economy mainly focused on the concept definition, measurement, function path and development trend, etc. [6,7], paying attention to the effects of specific information technologies or information infrastructure, such as big data and broadband, on economic growth [8,9]. The impact of the digital economy on the marine economy is becoming the focus of scholars. At present, the relevant research on the digital economy and marine economy mainly concentrates on how the digital economy or specific digital technologies affect marine fisheries, marine innovation, marine transportation, and others, which belong to the sub-topic of high-quality development. For example, Jiang et al. (2024) found that the digital economy improved the quality of marine fisheries’ development by promoting marine green technology innovation [10]. Fang et al. (2024) explored the relationships between digital technologies and the sustainability of the marine economy, thus reaching similar conclusions [11]. Yao et al. (2023) pointed out that there existed a nonlinear relationship between the digital economy and marine low-carbon economy [12]. He et al. (2022) verified that digital technology positively affected marine equipment manufacturing industries and had obvious regional heterogeneity [13]. Liu et al. (2023) found that the coordination between the digital economy and marine economy presented an increasing trend, which was affected by digital infrastructure and other factors [14].

There exist complex relationships between marine industrial structure upgrading and the marine economy, although some studies have confirmed the positive promotional effect of marine industrial structure upgrading on the marine economy [15]. However, with the deepening of research, more and more scholars have found that the impact of marine industrial structure upgrading on the marine economy is closely related to the direction of marine industrial structure evolution. For example, Xie et al. (2019) pointed out that industrial structure evolution could be divided into advancement and rationalization, which exert completely different impacts on the marine economy [16]. Wang et al. (2021) found that the advancement evolution of the marine industrial structure restrained marine fisheries, while the rationalization evolution presented promoting effects [17]. In addition, some scholars believe that there exist nonlinear relationships between them. For example, Wei et al. (2021) found that there was an inverted U-shaped relationship between marine industrial structure upgrading and GTFP [18].

With the continuous integration of digital information technology and industries, the impact of the digital economy on the industrial structure has attracted the attention of scholars. The positive effect of the digital economy on industrial structure upgrading has been confirmed. Bai et al. (2023) found that the digital economy promoted the transformation and upgrading of the industrial structure to reduce carbon emissions [19]. Song and Jiang (2024) believed that the driving effect of the digital economy on both industrial structure advancement and rationalization mainly depended on the intermediary variable of technological innovation [20]. Chen and Zhou (2024) found that the human capital structure and independent innovation were the important pathways of the digital economy affecting the industrial structure [21]. As far as the marine economy field is concerned, relatively few studies have directly linked the digital economy to marine industrial structure upgrading, and more attention has been paid to the impact of digital technology on the single marine industries [22,23].

Overall, the existing studies about the digital economy, marine industrial structure, and the high-quality development of the marine economy mainly have the following limitations. First, few studies have systematically investigated the relationships between the digital economy, marine industrial structure upgrading, and the high-quality development of the marine economy under a unified analytical framework. They more often focused on the effect of the digital economy and marine industrial structure upgrading, respectively, ignoring the synergistic effects between them. Second, less attention has been paid to the spatial spillover effect and spatial interaction of the digital economy and marine industrial structure upgrading. Besides, they also ignored the heterogeneity impacts of different dimensions of the digital economy, such as digital infrastructure, digital industrialization, and industrial digitalization.

Based on this, we attempted to incorporate digital economy, marine industrial structure upgrading, and the high-quality development of the marine economy into a unified analysis framework to explore the internal mechanisms among them. We firstly constructed a multi-dimensional evaluation index system to measure the HDME index and digital economy index, respectively. We further applied the static and dynamic spatial Durbin model to examine the impacts of the digital economy, marine industrial structure upgrading, and their interaction on the HDME, which aims to clarify the synergistic effect and spatial spillover effect of the digital economy and marine industrial structure upgrading. Furthermore, we also analyzed the heterogeneous impacts and threshold effects of digital economy sub-dimensional indicators (digital infrastructure, digital industrialization, and industrial digitalization) to identify the focus of integration between the digital economy and marine industries. Our study helps to promote the green transformation of the marine industrial structure and improve the quality of marine economic development, thus achieving the sustainable development goals of the marine economy.

2. Theoretical Analysis

The high-quality development of the marine economy is the comprehensive concept of innovative, coordinated, green, open, and shared development, emphasizing the organic unity of the marine economy’s “quality” and “quantity”. We attempted to explore the intrinsic mechanism of the digital economy and marine industrial structure upgrading affecting the HDME within a unified framework.

Digital Economy and High-Quality Development of Marine Economy

The digital economy itself has basic characteristics, such as high innovation and strong permeability, creating favorable conditions for improving the marine economy’s quality. The new economic growth theory pointed out that endogenous technological progress is the crucial element of economic growth [24]. Especially, the digital economy represents the essential reflection of innovation. The digital economy depends on data elements, takes modern information networks as the carrier, and promotes digital industrialization and industrial digitalization through continuous innovation of digital technologies [25,26]. The process of digital industrialization and industrial digitalization is also the process of knowledge creation and technology diffusion. Combined with the technology-economic characteristics of data elements, such as non-competitiveness and low cost, this technology spillover has strong penetration and multiplier effects among regions and sectors [27], thus improving local and neighboring regions’ marine economy.

With the application of digital information technologies, the information asymmetry barrier between producers and consumers in the traditional economy has been gradually broken [28], thus promoting the factor allocation efficiency of marine enterprises. Meanwhile, the new input of data factors not only creates its own value, but also strengthens the synergy between different production factors and optimizes the combination of factor inputs, thereby promoting total factor productivity. Furthermore, the digital economy also prompts the optimization of the energy structure, thus reducing carbon emissions and transforming consumption patterns [29,30]. Particularly, resources under the influence of the digital economy have broken-down the traditional geographic limitations and strengthened the spatial correlation of economic activities between regions. Therefore, the digital economy may have spatial spillover effects. Thus, we proposed Hypothesis 1:

Hypothesis 1.

Digital economy directly promotes the local HDME by relying on its own advantages and affects that of neighboring regions through spatial spillover effects.

The marine industrial structure upgrading mainly impacts the quality of local and neighboring marine economic development through factor reallocation, technology spillover, and division of labor specialization. First, industrial structure upgrading could facilitate the transfer of production factors among sectors with different production efficiencies [31]. Industrial rationalization further improves the mismatch of resources and factors among industries, enhancing the rational allocation of factors. Second, with the upgrading of the industrial structure, industries with high added value and high production efficiency have gradually become the leading industries in the region [32], accelerating technology spillover and diffusion between different sectors. New knowledge and technologies are spreading from knowledge-intensive marine industries to traditional marine industries through personnel exchanges and technical guidance [33], and guiding traditional industries, such as marine fisheries, toward green and high-end development. Meanwhile, the linkage and spillover effects generated by the productive service industries, such as blue finance, further boost the marine economy’s quality. Furthermore, the division of labor and specialization in the process of marine industrial structure upgrading also enhance the scale efficiency for marine enterprises. Besides, the adjustment of the local marine industrial structure inevitably comes with the out-migration of backward industries and the spillovers of advanced technologies. Obviously, the cross-regional industrial transfer and technology diffusion resulting from the marine industrial structure upgrading would influence the marine industrial structure, marine technical progress, and the quality of the marine economy in neighboring areas, thus causing the spatial spillover effects. Thus, we proposed Hypothesis 2:

Hypothesis 2.

The marine industrial structure upgrading positively affects the HDME of local and neighboring areas.

The synergistic effects between the digital economy and marine industrial structure upgrading positively affect the quality of the marine economy. The continuous convergence of data elements with marine industries has accelerated the diffusion and application of digital technologies between various production sectors [34,35]. This effectively expanded the medium- and high-end supply and enhanced the supply and demand matching at the industrial level through factor allocation optimization and production efficiency improvement [36], thus promoting marine industrial structure upgrading. On the one hand, the construction of marine digital infrastructure, for instance, marine communication networks and marine big data platforms, provides backing for the effective combination of digital technologies with marine industries. Digital information technologies have driven emerging digital industries, such as electronic information manufacturing [37], which helps to achieve the intelligent, green, and high-end development of marine manufacturing. Developing marine information industries can improve the research innovation capacity of marine manufacturing industries and the independent manufacturing capability of high-end marine electronic equipment through innovation driven by digital technologies. This makes the marine high-end manufacturing industries more competitive.

On the other hand, marine industries can transform the traditional mode of production and develop toward the “intelligence plus marine industry” by integrating with digital technologies, thus advancing the overall intelligence level of the marine industry [38]. Owing to the diffusion of digital technology in various marine industries, the communication and cooperation between the upstream and downstream of the marine industry chain has been strengthened. This helps to improve its integrity and coordination, positively affecting marine industrial structure upgrading [39]. Obviously, the digital economy further amplifies the “structural dividend” of marine industrial structure upgrading, driving the “quantity” and “quality” dual growth of the marine economy. Namely, there exist positive synergy effects between the digital economy and marine industrial structure upgrading. Based on the above analysis, the digital economy and marine industrial structure upgrading may have spatial spillover effects on the regional marine economy, and thus the impact of their interaction may also have spatial spillover effects. Therefore, we proposed Hypothesis 3:

Hypothesis 3.

The interaction between the digital economy and marine industrial structure upgrading can exert impacts on the HDME in local and neighboring areas

3. Methodology and Data

3.1. Model Construction

We constructed a spatial econometric model to analyze the impact of the digital economy, marine industrial structure upgrading, and their interaction on the HDME. The general form of the spatial econometric model was constructed as follows:

$$\begin{array}{l}ln{\mathit{HDME}}_t=\mathit{\delta }\mathit{W}ln{\mathit{HDME}}_t+\mathit{\alpha }{\mathit{\iota }}_n+{\mathit{X}}_t\mathit{\beta }+\mathit{W}{\mathit{X}}_t\mathit{\theta }+\mathit{\mu }+{\mathit{\xi }}_t{\mathit{I}}_n+{\mathit{u}}_t\\ {\mathit{u}}_t=\mathit{\lambda }\mathit{W}{\mathit{u}}_t+{\mathit{\epsilon }}_t\end{array}$$

(1)

where $ln{\mathit{HDME}}_t={\left(ln{g}_{1t},\cdots ,ln{g}_{nt}\right)}^{\prime }$ represents the logarithmic observation vector of the high-quality development of the marine economy during the investigation period, and $\mathit{X}={\left(lndig,lnind,lndig\times lnind,lnfee,lnren,lninf,lnopen\right)}^{\prime }$ is the independent variables matrix, composed of various independent variables’ vectors. W is the spatial weight matrix, while $\mathit{W}ln{\mathit{HDME}}_t$ and $\mathit{W}{\mathit{X}}_t$ are the spatial lag of dependent and independent variables, respectively. $\mathit{\delta }$ is the spatial autoregressive coefficient vector, $\mathit{\theta }$ is the spatial lag coefficient of variables, $\mathit{\lambda }$ denotes the spatial autocorrelation coefficient vector, $\mathit{\beta }$ denotes the parameters to be estimated, $\mathit{\alpha }$ denotes the constant term matrix, $\mathit{\mu }=\left({\mu }_1,\cdots ,{\mu }_n\right)$ represents the individual effect, and ${\mathit{\xi }}_t$ represents the time effect, while ${\mathit{u}}_t$ and ${\mathit{\epsilon }}_t$ denote the error term matrix. When $\mathit{\lambda }=0$ and $\mathit{\theta }=0$, the model above would transform into the SAR model, when $\mathit{\delta }=0$ and $\mathit{\theta }=0$, it would transform into the SEM model, and when $\mathit{\lambda }=0$, it would transform into the SDM model.

Considering that marine economy development is a continuous systematic activity, it is affected by both the current factors and its development level in the previous period under the long-term effects of the cultural environment, political system, and other factors. Thus, we further introduced the temporal lag term of the HDME to construct the spatial panel model, including dynamic effects:

$$\begin{array}{c}ln{\mathit{HDME}}_t=\tau ln{\mathit{HDME}}_{t-1}+\delta \mathit{W}ln{\mathit{HDME}}_t+\mathbf{\alpha }{\mathbf{\iota }}_n+\\ {\mathit{X}}_t\mathit{\beta }+\mathit{W}{\mathit{X}}_t\mathit{\theta }+\mathit{\mu }+{\mathit{\xi }}_t{\mathit{I}}_n+{\mathit{\upsilon }}_t\end{array}$$

(2)

where $\mathit{\tau }$ and $\mathit{\delta }$ are the corresponding coefficients of the temporal lag and spatial lag for the dependent variable, and ${\mathit{\upsilon }}_t$ denotes the error term matrix.

Considering that there may be a nonlinear relationship between the marine industrial structure upgrading and the HDME under different dimensions of the digital economy, we further constructed the threshold model with digital infrastructure, digital industrialization, and industrial digitalization as threshold variables to analyze the threshold effect:

$$\begin{array}{c}lnHDM{E}_t={\phi }_0+{\phi }_{1}lnin{d}_{it}\times I(lndi{g}_{jit}\le \gamma )+{\phi }_{2}lnin{d}_{it}\times I(lndi{g}_{jit}>\gamma )+\\ {\phi }_c{X}_{it}+{\mu }_{it}\end{array}$$

(3)

where, i and t denote the province and year, respectively, j is the three dimensions of the digital economy, I(·) is the indicator function, lndig_jit represents different dimensions of the digital economy, namely, digital infrastructure (lndig₁), digital industrialization (lndig₂), and industrial digitalization (lndig₃), γ is the threshold, and μ_it denotes the disturbance term.

3.2. Data

3.2.1. Dependent Variable

The HDME is a comprehensive concept involving multiple aspects, such as economy, environment, and society, rather than only including the absolute size of marine economy growth. In terms of the measurement, some scholars used a single index, such as green total factor productivity, to measure it, which is not accurate enough. Other scholars have constructed a comprehensive index system from a multi-dimensional perspective to measure it. For example, An et al. (2022) measured the comprehensive development of the marine economy across five aspects, including economic efficiency and industrial coordination [40]. Gao et al. (2022) and Wu et al. (2023) constructed an index system from different dimensions, including innovative, coordinated, green, open, and shared development [41,42]. Thus, referring to the practice of Gao et al. (2022) [41], we constructed a comprehensive evaluation index system from the perspective of innovative, coordinated, green, open, and shared development to measure the HDME of China’s coastal regions during 2012–2020 by applying the panel entropy method (see

Table 1. The evaluation index system of the HDME.

First Grade Index	Second Grade Index	Calculation Method
Marine innovative development	Number of marine scientific research institutions	Number of marine scientific research institutions
	Marine scientific research personnel investment	Marine scientific research institutions’ number of R&D personnel
	Marine scientific research funding investment	Marine scientific research institutions’ internal R&D expenditure
	Number of projects of marine research institutions	Number of projects of marine research institutions
	Marine science and technology outputs	Sum of the number of invention patents granted and papers and works published of marine scientific research institutions
Marine coordinated development	Marine economy contribution rate	Gross ocean product/gross domestic product
	Growth elasticity coefficient of marine tertiary industry	Marine tertiary industry growth rate/GOP growth rate
	Urban–rural coordination	Ratio of urban–rural per capita income
	Marine industrial structure diversification	Marine industrial structure diversification index
	Proportion of marine-related emerging industries	Added value of marine scientific research and education/GOP
Marine green development	Marine ecological protection	Area of marine natural reserves
	Investment in marine environmental control	Investment in industrial pollution control × (GOP/GDP)
	Degree of seawater pollution	Chemical oxygen demand in wastewater
	Marine industrial solid waste discharge	Industrial solid waste production/GOP
	Ocean disaster losses	Direct economic loss of storm surge disasters
Marineopen development	Foreign trade dependence	Total export–import volume of coastal areas/GDP
	Cargo throughput of port	Cargo throughput of port
	Passenger throughput of port	Passenger throughput of port
	International container throughput of port	International standard container throughput of port
Marine shared development	Per capita disposable income	Per capita urban disposable income
	Per capita disposable consumption	Per capita urban disposable consumption
	Per capita gross ocean product	GOP/Total population
	Urban unemployment rate	Urban unemployment rate
	Engel’s coefficient	Food consumption expenditure/Aggregate consumption expenditure

Note: marine industrial structure diversification index = ${\displaystyle \sum I_i\times ln{I}_i}$, where I_i is the ratio of the output value of marine industry i to GOP.

).

3.2.2. Independent Variables

(1) Digital economy (dig). Some scholars measured digital economy using the Internet broadband and other single indicators [42], which did not fully reflect the multiple dimensions’ information of the digital economy. Thus, based on the connotation of the digital economy, some scholars chose representative indicators to build a comprehensive index system from different dimensions, and then measured the digital economy indicators. For example, Zhang et al. (2022) constructed an index system from four different dimensions, including digital economy development carrier, digital industrialization, industrial digitalization, and digital economy development environment [43]. Pan et al. (2022) measured the digital economy from three aspects, including infrastructure, industrial scale, and spillover value [44].

Thus, considering the existing research and data availability, we constructed a multi-dimensional evaluation index system for the digital economy based on the three levels of digital infrastructure, digital industrialization, and industrial digitalization (see

Table 2. The evaluation index system of the digital economy.

First Grade Index	Second Grade Index	Calculation Method
Digital infrastructure	Mobile phone penetration rate	Mobile phone penetration rate
	Internet broadband access port	Number of Internet broadband subscribers port
	Internet domain name	Number of Internet domain names
	Length of distance of optical cable lines	Length of distance of optical cable lines
	Broadband subscribers of Internet	Number of broadband subscribers of Internet
	Capacity of mobile telephone exchanges	Capacity of mobile telephone exchanges
Digital industrialization	Proportion of digital industry business income	Computer, communication, and other electronic equipment manufacturing main business income/industrial enterprises above designated size of main business income
	Proportion of digital industry employees	Proportion of personnel employed in information transmission, software, and information technology services to urban employment
	Proportion of software-related business income	Software-related business income/added value of tertiary industry
	Per capita business volume of telecommunication services	Business volume of telecommunication services/Total population
	Digital TV subscribers	Number of digital TV subscribers
	Number of digital enterprises	Number of impersonal entities of information transmission, software, and information technology services
Industrial digitalization	Enterprise informatization	Proportion of enterprises with information management
	Rate of coverage of corporate website	Proportion of website enterprises
	Digital financial inclusion	E-commerce sales
	Scale of express business	Express delivery volumes
	Scale of e-commerce enterprise	Ratio of enterprises with e-commerce trading activities
	Digital financial inclusion	Peking University Digital Financial Inclusion Index of China

). We further applied the panel entropy method to measure the comprehensive index and its sub-dimensional index. Specifically, digital infrastructure is the basis of the digital economy, involving industrial Internet, artificial intelligence, and other fields. We mainly used the mobile phone penetration rate and other indicators to measure it. Digital industrialization includes digital technologies innovation and digital industrial production, involving electronic equipment manufacturing, etc. We measured it using the proportion of digital industry business income and other indicators. Industrial digitalization focuses more on the combination of digital technologies with other industries, which is measured by enterprise informatization and other indicators.

(2) Marine industrial structure upgrading (ind). The optimization of the marine industrial structure prompts production efficiency and economic growth [45]. We applied the proportion of marine tertiary industry output value to the gross ocean product to measure this indicator.

3.2.3. Control Variables

(1) Marine environmental regulation (fee). Marine environmental regulation is measured by the expropriated maritime royalties. (2) Marine research and development investment (ren). We measured this variable by the ratio of marine scientific and technological personnel to the employees in marine scientific and technological institutions. (3) Transportation infrastructure (inf). We used the per capita urban road area to measure the regional transportation infrastructure. (4) Trade openness (open). The proportion of total imports and exports to GDP was used to measure this variable.

3.3. Spatial Weight Matrix

We constructed a geographical distance matrix (W₁), human capital geographic spatial weight matrix (W₂), and economic geographic spatial weight matrix (W₃), respectively.

(1) Geographical distance matrix (W₁). This matrix was constructed to effectively reflect the geographical spatial relationships between provinces. The formula is shown below:

$$W_1=\left\{\begin{array}{cc}1/d_{ij}& i\ne j\\ 0& i=j\end{array}\right.$$

(4)

where d_ij represents the geographical distance between province i and province j, which was measured by the provincial capital cities’ longitude and latitude.

(2) Human capital geographic weight matrix (W₂). It is assumed that provinces with high levels of human capital could have greater spatial impacts on the adjacent areas’ marine economy. The matrix was composed of the product of the geographical distance matrix and the diagonal matrix formed by the average value of the proportion of human capital in coastal regions. The formula is shown below:

$$W_2=W_1\times diag\left({\displaystyle \frac{{\overline{h}}_1}{\overline{h}}}\right.,{\displaystyle \frac{{\overline{h}}_2}{\overline{h}}},\cdots ,\left.{\displaystyle \frac{{\overline{h}}_n}{\overline{h}}}\right)$$

(5)

where ${\overline{h}}_i={\displaystyle \frac{1}{t_1-t_0+1}}{\displaystyle \sum _{t=t_0}^{t_1}{h}_{it}}$ represents the average human capital level of province i during 2012–2020. $\overline{h}={\displaystyle \frac{1}{n(t_1-t_0+1)}}{\displaystyle \sum _{t=t_0}^{t_1}{\displaystyle \sum _{i=1}^n{h}_{it}}}$ represents the average human capital level of sample regions during 2012–2020. The human capital was calculated by the average years of education, which was calculated by the following formula: (number of population with primary school education × 6 + number of population with junior high school education × 9 + number of population with senior high school education × 12 + number of population with junior college and above × 16)/total population aged 6 years and above.

(3) Economic geographic weight matrix (W₃). The matrix was composed of the product of the geographical distance matrix and the diagonal matrix formed by the average value of the proportion of per capita GDP of coastal provinces. The formula is expressed as follows:

$$W_3=W_1\times diag\left({\displaystyle \frac{{\overline{y}}_1}{\overline{y}}}\right.,{\displaystyle \frac{{\overline{y}}_2}{\overline{y}}},\cdots ,\left.{\displaystyle \frac{{\overline{y}}_n}{\overline{y}}}\right)$$

(6)

where ${\overline{y}}_i={\displaystyle \frac{1}{t_1-t_0+1}}{\displaystyle \sum _{t=t_0}^{t_1}{y}_{it}}$ represents the average per capita GDP of province i from 2012 to 2020. $\overline{y}={\displaystyle \frac{1}{n(t_1-t_0+1)}}{\displaystyle \sum _{t=t_0}^{t_1}{\displaystyle \sum _{i=1}^n{y}_{it}}}$ represents the average per capita GDP of sample regions during 2012–2020. Besides, the above spatial weight matrix was normalized by rows to ensure the sum of each row was 1.

3.4. Data Sources

Considering the China Marine Statistical Yearbook is updated to 2020, we used the panel data of China’s 11 coastal regions from 2012 to 2020. All original data were from the China Marine Statistical Yearbook, provincial statistical yearbook, and China’s National Statistics official websites. Besides, some of the missing data were interpolated.

4. Analysis of Measurement Results

Taking 2012 and 2020 as representative years, we visualized and analyzed the measurement results of the HDME index and digital economy comprehensive index in China’s coastal regions, as listed in Figure 1 and Figure 2, respectively.

4.1. Results of HDME Index

The high-quality development of China’s marine economy presented a steady upward trend during 2012–2020, the overall average of which increased from 0.205 to 0.291, and the average annual growth rate was 4.51%. In 2012, regions with the highest and lowest ranking were Guangdong (0.363) and Guangxi (0.071), respectively. Compared with 2012, the magnitude of the HDME index in 2020 obviously improved. However, the relative ranking of coastal provinces only made a slight change. Among them, the top-three-ranked provinces were Guangdong (0.576), Shanghai (0.429), and Shandong (0.416), which was consistent with the ranking in 2012. Provinces with the middle level of the HDME index were Zhejiang, Jiangsu, Tianjin, Fujian, and Hainan, which always remained in the second echelon. Moreover, the quality of the marine economy in Hebei and Guangxi was relatively low and always in the third echelon. Liaoning’s marine economy fluctuated between the second and third echelons. On the whole, the HDME during 2012–2020 had obvious spatial disequilibrium characteristics and a multi-polarization trend. This was possibly related to the differences in the marine industry, land economic development, and marine resource endowment among the coastal regions. Particularly, Shanghai, Guangdong, and Shandong relied on their own advantages in policies, location, and financially to accelerate the improvement of marine independent innovation capacity and vigorously develop emerging marine industries, such as marine electronic information and marine engineering equipment.

4.2. Results of Digital Economy Index

Intuitively, the comprehensive index of the digital economy for coastal regions generally increased from 2012 to 2020. Its average value increased from 0.158 to 0.385, with an average annual growth rate of 11.77%. Significantly, the average annual growth rate of the digital economy was greater than that of the HDME during the study period, which indicates that the digital economy is becoming the key driving force in the marine economy. The development of the digital economy had obvious regional differences in the coastal regions. In 2012, provinces with a high level were Guangdong, Jiangsu, and Zhejiang, which were in the leading positions. Shandong, Shanghai, and Fujian were in the second echelon. Compared with other coastal regions, Liaoning, Hebei, and other provinces were in the third echelon, with a low level. In 2020, the top-three-ranked coastal provinces were Guangdong (0.842), Zhejiang (0.573), and Jiangsu (0.526), while the bottom ranked were Hainan (0.155), Tianjin (0.193), and Liaoning (0.226). This is basically consistent with the ranking of coastal areas in the Digital China Development Report 2020, which reflects the reasonability of the digital economy evaluation index system constructed and the measurement method used in this study, to some extent. Among them, Guangdong, with its own unique advantages in digital technology innovation and digital industrialization, gradually expanded the disparity with other provinces, and it ranked first in coastal regions.

5. Empirical Results and Discussion

5.1. Results of Spatial Autocorrelation Test

We conducted, respectively, preliminary tests of spatial correlation for the HDME, digital economy, and marine industrial structure upgrading between regions by using the Moran’s I index (see

Table 3. The results of the global spatial autocorrelation test.

	Variables	Spatial Weight Matrix	Moran’s I	Sd(I)	Z-Value
Moran’s I	lnG	W1	0.373	0.062	6.210 ***
		W2	0.368	0.063	6.029 ***
		W3	0.347	0.064	5.583 ***
	lndig	W1	0.496	0.062	8.200 ***
		W2	0.497	0.063	8.077 ***
		W3	0.492	0.064	7.850 ***
	lndig1	W1	0.702	0.062	11.538 ***
		W2	0.705	0.063	11.406 ***
		W3	0.730	0.064	11.577 ***
	lndig2	W1	0.433	0.062	7.196 ***
		W2	0.432	0.063	7.071 ***
		W3	0.448	0.064	7.172 ***
	lndig3	W1	0.323	0.062	5.397 ***
		W2	0.322	0.063	5.292 ***
		W3	0.286	0.064	4.620 ***
	lnind	W1	0.420	0.062	6.986 ***
		W2	0.423	0.063	6.910 ***
		W3	0.395	0.064	6.338 ***

Note: *** indicates 1% significance level.

). There are two methods to test the spatial correlation of panel data. One is to calculate the Moran’s I values of variables in different years based on the cross-sectional data during the sample period, while the other is to merge the data of variables in all years during the sample period into a mixed panel to calculate the Moran’s I values with the help of the spatial weight matrix, with N × T. Referring to Zhu et al. (2011) [46], we applied the second method, which merged the data of 11 provinces for 9 years during 2012–2020 into a mixed panel, to conduct the spatial correlation test for core independent variables and dependent variables. Under three spatial weight matrices, the Moran’s I values of the HDME index, digital economy comprehensive index, digital infrastructure sub-index, digital industrialization sub-index, industrial digitalization sub-index, and marine industrial structure upgrading were all significantly positive. This suggested that the HDME, digital economy, and marine industrial structure upgrading among coastal regions during the study period showed obviously positive spatial autocorrelation. Additionally, we also conducted a spatial autocorrelation test based on the residuals of the ordinary least squares (OLS) linear estimation results. The Moran’s I values of the OLS regression residuals under different spatial weight matrices were also significantly positive (see

Table 4. The results of spatial model selection tests.

Test Parameters	W1		W2		W3
	Statistics	p-Value	Statistics	p-Value	Statistics	p-Value
LM error	63.475	0.000	64.473	0.000	60.756	0.000
Robust LM error	42.841	0.000	45.098	0.000	43.717	0.000
LM lag	21.286	0.000	20.420	0.000	18.179	0.000
Robust LM lag	0.651	0.420	1.046	0.307	1.139	0.286
Moran’s I	9.990	0.000	10.042	0.000	9.855	0.000
Wald lag	43.95	0.000	42.79	0.000	34.01	0.000
Wald error	29.65	0.000	29.09	0.000	25.34	0.000
LR lag	35.13	0.000	34.80	0.000	28.58	0.000
LR error	28.16	0.000	27.88	0.000	22.11	0.002
Spatial fixed effect LR test	43.43	0.000	42.30	0.000	32.23	0.000
Time fixed effect LR test	132.84	0.000	140.48	0.000	167.81	0.000

). Hence, spatial correlation should be considered when exploring the relationships between digital economy, marine industrial structure upgrading, and the HDME.

5.2. Results of Spatial Model Selection Tests

A series of statistical tests were conducted to decide the suitable form of the spatial econometric model. The selection of the model mainly included the following ideas: one was from the special to the general, namely, to select the SAR model or SEM model using the results of the Lagrange multiplier (LM) test, while the other was from the general to the special; that is, if the model with spatial effects was not rejected, the Wald test or likelihood ratio (LR) test was further applied to determine whether the SDM model could be simplified to the SAR model or SEM model [47] (LeSage, 2014). As shown in Table 4, under W₁, W₂, and W₃, the values of LM lag and LM error were statistically significant, implying that both spatial lag and spatial error existed in the model. Thus, the SDM model should be selected. Although Robust LM lag was insignificant, the results of the Wald test and LR test both showed that the spatial Durbin model could not be degenerated into the SAR model and the SEM model. Hence, adopting the SDM model was reasonable. Moreover, under three spatial weight matrices, the LR joint significance test of the time fixed effect and spatial fixed effect was significantly positive, which suggested that a model with both spatial and time fixed effects should be selected. Thus, we applied the SDM model with spatial and time fixed effects to estimate.

5.3. Results of Static Spatial Durbin Model

Table 5. The results of static spatial Durbin model with three spatial weight matrices.

Variables	W1		W2		W3
lndig	0.448 ***	0.461 ***	0.451 ***	0.459 ***	0.439 ***	0.440 ***
	(5.33)	(5.71)	(5.31)	(5.58)	(4.84)	(4.83)
lnind	0.094	0.344 **	0.109	0.340 **	0.139	0.333 *
	(0.66)	(2.06)	(0.77)	(2.04)	(0.99)	(1.97)
lndig × lnind		0.394 ***		0.361 **		0.312 **
		(2.61)		(2.39)		(2.07)
lnfee	0.013	0.000	0.013	0.001	0.016	0.005
	(1.16)	(0.01)	(1.15)	(0.09)	(1.29)	(0.41)
lnren	0.407 ***	0.477 ***	0.403 ***	0.467***	0.340 ***	0.386 ***
	(3.71)	(4.41)	(3.68)	(4.30)	(2.99)	(3.39)
lninf	−0.365 ***	−0.372 ***	−0.377 ***	−0.378 ***	−0.318 **	−0.318 **
	(−2.62)	(−2.79)	(−2.73)	(−2.84)	(−2.38)	(−2.40)
lnopen	0.044	0.132 *	0.045	0.125 *	0.074	0.142 *
	(0.62)	(1.75)	(0.64)	(1.67)	(1.15)	(1.99)
W × lnHDME	−0.951 ***	−1.090 ***	−0.909 ***	−1.027 ***	−0.734 ***	−0.763 ***
	(−4.28)	(−4.73)	(−4.09)	(−4.43)	(−3.16)	(−3.03)
W × lndig	2.052 ***	1.945 ***	1.981 ***	1.849 ***	1.912 ***	1.782 ***
	(4.70)	(4.65)	(4.59)	(4.41)	(3.98)	(3.66)
W × lnind	0.287	1.613 **	0.210	1.427 **	0.171	1.063
	(0.64)	(2.24)	(0.49)	(2.01)	(0.40)	(1.51)
W × lndig × lnind		1.118 *		1.039 *		0.698
		(1.93)		(1.82)		(1.28)
W × lnfee	−0.009	−0.037	−0.000	−0.025	−0.000	−0.012
	(−0.18)	(−0.76)	(−0.00)	(−0.54)	(−0.01)	(−0.25)
W × lnren	1.577 ***	1.820 ***	1.600 ***	1.815 ***	1.218 **	1.337 **
	(3.18)	(3.67)	(3.12)	(3.55)	(2.03)	(2.22)
W × lninf	−1.037 **	−1.047 **	−1.046 **	−1.058 **	−0.962 **	−0.986 **
	(−2.04)	(−2.16)	(−2.12)	(−2.22)	(−2.01)	(−2.07)
W × lnopen	−0.709 **	−0.402	−0.740 **	−0.444	−0.968 ***	−0.693 *
	(−2.36)	(−1.26)	(−2.45)	(−1.38)	(−2.90)	(−1.92)
R2	0.7206	0.7174	0.7219	0.7197	0.7147	0.7161

Note: z-statistics are in parentheses. ***, **, and * indicate 1%, 5%, and 10% significance levels, respectively.

reports the estimation results of the static spatial Durbin model with the interaction term under three weight matrices. We focused on the impacts of the digital economy, marine industrial structure upgrading, and their interaction on the HDME. Meanwhile, we also reported the results of the model without the interaction term for comparative analysis. Under three spatial weight matrices, the coefficients of digital economy, marine industrial structure upgrading, and their interaction term were significantly positive. In the spatial Durbin model without the interaction term, only the digital economy was significantly positive, whereas marine industrial structure upgrading was not significant. We centralized the core variables and constructed the interaction term. After introducing the interaction term, the coefficients of the digital economy increased and remained significantly positive. The significance of the marine industrial structure upgrading coefficient was obviously improved and significant at the level of at least 10%. The interaction term coefficient was also significantly positive. This suggested digital economy and marine industrial structure upgrading had positive synergistic effects on the marine economy’s high-quality growth.

Since the spatial spillover effects measured by the spatial Durbin model were global effects rather than local effects, the point estimation results of the model were biased and did not effectively reflect the marginal impacts of the independent variables [48,49] (LeSage and Pace, 2009; Elhorst, 2014). Based on this, the direct, indirect, and total effects of variables on the HDME were estimated using the partial differential estimation method. Thus, we further decomposed the direct, indirect, and total effects of variables in the static spatial Durbin model with the interaction term, as shown in

Table 6. The decomposition results of the static spatial Durbin model.

	Effects	lndig	lnind	lndig × lnind	lnfee	lnren	lninf	lnopen
W1	Direct	0.244 **	0.125	0.288 **	0.006	0.278 **	−0.277 **	0.217 ***
		(2.11)	(0.62)	(2.05)	(0.47)	(2.49)	(−2.44)	(2.66)
	Indirect	0.911 ***	0.838 **	0.458	−0.026	0.840 ***	−0.406	−0.341 *
		(3.62)	(2.02)	(1.62)	(−0.90)	(2.76)	(−1.42)	(−1.87)
	Total	1.155 ***	0.964 ***	0.746 **	−0.019	1.118 ***	−0.684 **	−0.124
		(5.00)	(2.79)	(2.45)	(−0.77)	(3.59)	(−2.09)	(−0.65)
W2	Direct	0.263 **	0.157	0.263 *	0.005	0.273 **	−0.284 **	0.210 ***
		(2.33)	(0.80)	(1.89)	(0.40)	(2.44)	(−2.48)	(2.60)
	Indirect	0.880 ***	0.743 *	0.451	−0.019	0.874 ***	−0.430	−0.361 *
		(3.41)	(1.79)	(1.57)	(−0.68)	(2.70)	(−1.47)	(−1.90)
	Total	1.143 ***	0.900 **	0.714 **	−0.014	1.147 ***	−0.714 **	−0.151
		(4.75)	(2.53)	(2.29)	(−0.55)	(3.45)	(−2.12)	(−0.76)
W3	Direct	0.303 ***	0.235	0.269 *	0.007	0.284 ***	−0.246 **	0.222 ***
		(2.82)	(1.36)	(1.97)	(0.53)	(2.63)	(−2.09)	(2.83)
	Indirect	0.971 ***	0.590	0.329	−0.012	0.719*	−0.505	−0.528 **
		(2.86)	(1.34)	(1.08)	(−0.41)	(1.77)	(−1.48)	(−2.17)
	Total	1.274 ***	0.825 *	0.598 *	−0.006	1.003**	−0.751 *	−0.306
		(3.70)	(1.95)	(1.70)	(−0.19)	(2.36)	(−1.87)	(−1.21)

Note: z-statistics are in parentheses. ***, **, and * indicate 1%, 5%, and 10% significance levels, respectively.

.

Overall, the estimation results of variables under various weight matrices, especially the core explanatory variables, were basically in line with each other. This indicated that the estimation results were robust. By comparing the goodness of fit (R²) of models under various weight matrices, we found that the R² value of the static spatial Durbin model with the human capital geographic matrix, W₂, was the largest (0.7197). Therefore, we stressed on analyzing the results under the human capital geographic matrix, W₂. Table 6 shows that the direct, indirect, and total effects of the digital economy were significantly positive, which were 0.263, 0.880, and 1.143, respectively. The indirect effect of the digital economy was greater than its direct effect. Relying on its advantages in technological innovation and factor allocation, the digital economy directly promoted local marine economic growth and produced positive spatial spillover effects on neighboring regions through the demonstration–imitation and diffusion–spillover effects.

The total and indirect effects of marine industrial structure upgrading were significantly positive, whereas the direct effect was positive but insignificant. This implied that marine industrial structure upgrading did not effectively strengthen the local marine economy’s quality as expected, but positively affected that of neighboring areas. The likely reasons were as follows: The dividends of marine industrial structure upgrading had obvious skewing. At the initial stage, the structural dividends of industrial structure adjustment had not been fully released. The positive benefits from industrial structure adjustment may be offset by the negative costs to a certain extent. Moreover, the optimization of the marine industrial structure for improving the local marine economy may result in the free-riding behavior of neighboring areas, thus producing positive spillover effects. For the interaction term, the total and direct effects were significantly positive, whereas the indirect effect was statistically insignificant. Especially, the total and direct effects of the interaction term were much greater than those of the digital economy and marine industrial structure upgrading under corresponding effects. Obviously, the interaction term had positive regulatory effects but had not yet produced a significant spatial spillover effect. Namely, the synergy effect between the digital economy and marine industrial structure upgrading only significantly promoted the local marine economy. Owing to the great differences in the digital economy and marine industrial structure between regions, the efficiency, speed, and degree of integration between the digital economy and marine industries were also different. Moreover, these synergistic effects may have the hysteresis quality in the process of external spillover. Therefore, although the digital economy could cooperate with the marine industrial structure upgrading to affect the marine economy in the surrounding areas to a certain extent, the estimated coefficient was not significant.

The direct, indirect, and total effects of marine environmental regulation were not statistically significant, implying marine environmental regulation had not played positive roles in the regional marine high-quality economic development at this stage. For marine research and development investment, the increase in marine science and technology R&D personnel had positive effects on the marine economy in local and surrounding areas. The direct effect of transportation infrastructure was significantly negative, and the indirect effect was statistically insignificant. This may be due to the nonlinear characteristics of road and railway density with diminishing marginal effects, which had negative impacts on improving the marine economy’s quality. This is also in keeping with the findings of Coşar and Demir (2016) [50], who concluded that transport infrastructure had nonlinear characteristics and its contribution to trade and economic growth declined in the later period. Besides, trade openness enhanced the local marine economy and inhibited the quality of the marine economy in neighboring areas.

5.4. Results of Dynamic Spatial Durbin Model

We further introduced the temporal lag of the HDME into the models, estimated by the dynamic spatial Durbin model (see

Table 7. The results of the dynamic spatial Durbin model with three spatial weight matrices.

Variables	W1	W2	W3
L.lnHDME	0.217 **	0.218 **	0.252 ***
	(2.27)	(2.26)	(2.64)
W × lnHDME	1.010 ***	0.965 ***	0.784 ***
	(3.99)	(3.85)	(3.01)
lndig	0.379 ***	0.375 ***	0.352 ***
	(3.52)	(3.45)	(3.18)
lnind	0.303	0.305	0.271
	(1.60)	(1.62)	(1.52)
lndig × lnind	0.430 **	0.418 **	0.343 *
	(2.34)	(2.29)	(1.95)
W × lndig	1.503 ***	1.402 ***	1.324 **
	(2.98)	(2.80)	(2.46)
W × lnind	1.914 **	1.794 **	1.357 *
	(2.43)	(2.31)	(1.83)
W × lndig × lnind	1.503 **	1.489 **	1.145 *
	(2.09)	(2.12)	(1.77)
control variables	Yes	Yes	Yes
R2	0.6455	0.6481	0.6492

Note: z-statistics are in parentheses. ***, **, and * indicate 1%, 5%, and 10% significance levels, respectively.

). Compared with the static model, the dynamic spatial Durbin model includes both the spatial lag and temporal lag effects of explained variables. Under W₁, W₂, and W₃, the temporal lag coefficients of the HDME were all significantly positive. This suggested that the marine economy had obvious temporal inertia, namely, the previous marine economy’s quality exerted positive impacts on its current development. Its spatial lag coefficients were also positive and statistically significant, indicating that the regional marine economy’s quality had obvious positive spatial interactions. This was in line with the results of the spatial autocorrelation analysis above. Regions with high levels of the HDME could exert positive spillover effects on the marine economy in neighboring regions through demonstration externality. Consequently, the estimation results of the dynamic spatial Durbin model were more reasonable.

Owing to the fact that the dynamic model included the temporal lag and spatial lag of the HMDE, the direct and indirect effects could be divided into the short-term and the long-term effects, respectively.

Table 8. The decomposition results of the dynamic spatial Durbin model.

Weight Matrices	Variables	Short Term			Long Term
		Direct Effect	Indirect Effect	Total Effect	Direct Effect	Indirect Effect	Total Effect
W1	lndig	0.213	0.741 **	0.954 ***	0.142	0.931	1.074 ***
		(1.60)	(2.34)	(3.25)	(0.12)	(0.73)	(3.15)
	lnind	0.087	1.044 **	1.130 ***	−0.036	1.306	1.270 ***
		(0.38)	(2.27)	(2.82)	(−0.03)	(0.94)	(2.80)
	lndig × lnind	0.279 *	0.698 **	0.977 **	0.283	0.814	1.097 **
		(1.64)	(1.96)	(2.46)	(0.25)	(0.67)	(2.45)
W2	lndig	0.225 *	0.697 **	0.922 ***	0.216	0.826	1.042 ***
		(1.72)	(2.18)	(3.10)	(0.55)	(1.55)	(3.00)
	lnind	0.110	0.986 **	1.096 ***	−0.021	1.257	1.236 ***
		(0.50)	(2.15)	(2.70)	(−0.02)	(1.23)	(2.67)
	lndig × lnind	0.270	0.717 **	0.987 **	0.212	0.900	1.112 **
		(1.59)	(2.01)	(2.46)	(0.20)	(0.80)	(2.45)
W3	lndig	0.239 **	0.734 *	0.973 ***	0.270	0.875 *	1.144 **
		(1.96)	(1.94)	(2.60)	(1.26)	(1.74)	(2.42)
	lnind	0.168	0.781 *	0.949 **	0.157	0.956	1.113 **
		(0.91)	(1.70)	(2.13)	(0.47)	(1.54)	(2.09)
	lndig × lnind	0.255	0.595 *	0.850 **	0.296	0.699	0.995 **
		(1.56)	(1.69)	(2.05)	(1.15)	(1.55)	(2.03)

Note: z-statistics are in parentheses. ***, **, and * indicate 1%, 5%, and 10% significance levels, respectively.

reports the decomposition results of dynamic Durbin models with three spatial weight matrices. The results of independent variables were basically consistent with those of the static model, which indicated that the results were robust. Moreover, the sign and significance of core variables’ coefficients were basically consistent in the dynamic model with three weight matrices, again indicating that the estimation results had strong robustness. As mentioned above, we took the results under human capital geographic matrix, W₂, as an example for the analysis. In the short term, the direct, indirect, and total effects of the digital economy were significantly positive, and the indirect effect was greater than the direct effect. The total and indirect effects of marine industrial structure upgrading were significantly positive, whereas the direct effect was insignificant. Besides, the total and indirect effects of the interaction term were also significantly positive, while the direct effect was only significantly positive under W₁. This implied that the synergistic effects between the digital economy and marine industrial structure were still obvious when introducing the dynamic effects of the HDME. In the long term, the total effects of the digital economy, marine industrial structure upgrading, and their interaction were significantly positive, which implies that the positive and synergistic effects of the digital economy and marine industrial structure upgrading were only reflected in the total impact, without distinguishing between regions.

5.5. Results of Heterogeneity Test

Based on the different dimensions of digital infrastructure (lndig₁), digital industrialization (lndig₂), and industrial digitalization (lndig₃), we further examined the influence of the digital economy sub-dimensions index, marine industrial structure upgrading, and their interaction on the HDME, aiming to further grasp the relationships among them from the perspective of heterogeneity.

Table 9. The decomposition results of the dynamic spatial Durbin model with digital infrastructure.

Weight Matrices	Variables	Short Term			Long Term
		Direct Effect	Indirect Effect	Total Effect	Direct Effect	Indirect Effect	Total Effect
W1	lndig1	0.148	0.605 *	0.753 **	0.152	0.719	0.871 **
		(1.40)	(1.78)	(2.31)	(0.46)	(1.38)	(2.21)
	lnind	0.179	1.003 *	1.182 **	0.149	1.214	1.362 **
		(0.81)	(1.81)	(2.32)	(0.17)	(1.13)	(2.28)
	lndig1 × lnind	0.264 *	0.490 **	0.754 **	0.314	0.555	0.869 **
		(1.75)	(2.09)	(2.43)	(0.97)	(1.49)	(2.39)
W2	lndig1	0.149	0.571 *	0.720 **	−0.122	0.962	0.841 **
		(1.41)	(1.65)	(2.16)	(−0.02)	(0.15)	(2.06)
	lnind	0.196	0.979 *	1.175 **	−0.480	1.845	1.365 **
		(0.89)	(1.75)	(2.27)	(−0.03)	(0.12)	(2.21)
	lndig1 × lnind	0.269 *	0.509 **	0.778 **	0.130	0.773	0.904 **
		(1.77)	(2.15)	(2.49)	(0.03)	(0.18)	(2.44)
W3	lndig1	0.132	0.734 *	0.867 **	0.143	0.918 *	1.061 *
		(1.31)	(1.87)	(2.13)	(0.86)	(1.68)	(1.89)
	lnind	0.226	1.000 *	1.226 *	0.248	1.240	1.488 *
		(1.14)	(1.66)	(2.06)	(0.75)	(1.52)	(1.92)
	lndig1 × lnind	0.230	0.580 **	0.810 **	0.285	0.697 *	0.981 **
		(1.60)	(2.20)	(2.39)	(1.37)	(1.96)	(2.23)

Note: z-statistics are in parentheses. ** and * indicate 5% and 10% significance levels, respectively.

,

Table 10. The decomposition results of the dynamic spatial Durbin model with digital industrialization.

Weight Matrices	Variables	Short Term			Long Term
		Direct Effect	Indirect Effect	Total Effect	Direct Effect	Indirect Effect	Total Effect
W1	lndig2	0.151 ***	0.025	0.175 *	0.175 ***	0.014	0.188 *
		(3.01)	(0.20)	(1.74)	(2.82)	(0.10)	(1.74)
	lnind	0.355	0.424	0.779 *	0.398	0.438	0.836 *
		(1.55)	(1.07)	(1.94)	(1.46)	(0.98)	(1.93)
	lndig2 × lnind	0.306	1.005 *	1.311 **	0.324	1.084 *	1.408 **
		(1.51)	(1.93)	(2.33)	(1.34)	(1.87)	(2.31)
W2	lndig2	0.148 ***	0.026	0.174 *	0.172 ***	0.015	0.187 *
		(2.93)	(0.20)	(1.67)	(2.74)	(0.10)	(1.66)
	lnind	0.374 *	0.355	0.729 *	0.425	0.361	0.785 *
		(1.65)	(0.92)	(1.83)	(1.57)	(0.82)	(1.82)
	lndig2 × lnind	0.283	1.042*	1.325 **	0.297	1.130 *	1.427 **
		(1.40)	(1.93)	(2.29)	(1.23)	(1.88)	(2.27)
W3	lndig2	0.133 **	0.022	0.155	0.163 **	0.008	0.170
		(2.42)	(0.13)	(1.13)	(2.15)	(0.04)	(1.13)
	lnind	0.386 *	0.183	0.569	0.469 *	0.157	0.626
		(1.77)	(0.51)	(1.54)	(1.68)	(0.37)	(1.53)
	lndig2 × lnind	0.138	1.154 *	1.292 **	0.111	1.311 *	1.422 **
		(0.68)	(1.96)	(2.19)	(0.41)	(1.92)	(2.17)

Note: z-statistics are in parentheses. ***, **, and * indicate 1%, 5%, and 10% significance levels, respectively.

and

Table 11. The decomposition results of the dynamic spatial Durbin model with industrial digitalization.

Weight Matrices	Variables	Short Term			Long Term
		Direct Effect	Indirect Effect	Total Effect	Direct Effect	Indirect Effect	Total Effect
W1	lndig3	−0.172	−0.402	−0.573	−0.213	−0.493	−0.706
		(−1.40)	(−1.05)	(−1.29)	(−0.71)	(−0.89)	(−1.25)
	lnind	0.333	0.481	0.813 *	0.516	0.475	0.990 *
		(1.60)	(0.93)	(1.66)	(0.32)	(0.27)	(1.64)
	lndig3 × lnind	0.063	0.629	0.692	0.078	0.763	0.841
		(0.35)	(1.44)	(1.47)	(0.03)	(0.31)	(1.46)
W2	lndig3	−0.158	−0.339	−0.497	−0.224	−0.398	−0.623
		(−1.26)	(−0.89)	(−1.11)	(−0.70)	(−0.69)	(−1.08)
	lnind	0.331	0.496	0.828 *	0.479	0.545	1.024 *
		(1.62)	(0.95)	(1.67)	(0.79)	(0.61)	(1.65)
	lndig3 × lnind	0.072	0.671	0.743	0.072	0.847	0.919
		(0.41)	(1.53)	(1.58)	(0.06)	(0.65)	(1.57)
W3	lndig3	−0.120	−0.159	−0.279	−0.207	−0.178	−0.385
		(−0.99)	(−0.46)	(−0.68)	(−0.84)	(−0.33)	(−0.66)
	lnind	0.268	0.572	0.840 *	0.389	0.746	1.135 *
		(1.39)	(1.13)	(1.71)	(0.69)	(0.82)	(1.64)
	lndig3 × lnind	0.112	0.676 *	0.788 *	0.082	0.978	1.060 *
		(0.67)	(1.72)	(1.87)	(0.19)	(1.46)	(1.81)

Note: z-statistics are in parentheses. * indicates 10% significance level.

report the decomposition results of dynamic spatial Durbin models, with digital infrastructure, digital industrialization, and industrial digitalization as independent variables, respectively. Under three weight matrices, the decomposition results were basically consistent. Therefore, we also took the estimation results under the human capital geographic matrix, W₂, as an example for the analysis, as mentioned above.

As shown in Table 9, for digital infrastructure and marine industrial structure upgrading, both the total and indirect effects in the short term were significantly positive, implying that the improvement of digital infrastructure and marine industrial structure upgrading exerted positive effects on the marine economy. For the interaction term of digital infrastructure and marine industrial structure upgrading, the direct, indirect, and total effects were both significantly positive, which suggested that the synergistic effect between digital infrastructure and marine industrial structure upgrading helped to enhance the quality of the marine economy in local and surrounding regions. In addition, the positive impacts of digital infrastructure, marine industrial structure upgrading, and their interaction in the long term did not rely on the spatial spillover, but it was reflected in the total effect. The construction of digital infrastructure is the premise of digital industrialization and industrial digitalization. With advantages in narrowing the information gap, reducing enterprise transaction costs, and expanding the market scope, digital infrastructure can effectively accelerate the flow of key factors between various regions and industrial sectors. Meanwhile, digital infrastructure can enhance the diffusion of new technologies. This is conducive to guiding marine industries, especially marine manufacturing industries, to develop in the direction of intelligence and high quality, thus generating synergy with the marine industrial structure to the regional marine economy.

Table 10 reported that, in terms of digital industrialization, both the short-term direct and total effects were significantly positive, whereas the short-term indirect effect was insignificant. This indicated that digital industrialization only promoted the quality of the local marine economy. Furthermore, the long-term direct and total effects for digital industrialization were still significantly positive and were larger than the corresponding effects in the short term. Digital industrialization prompts the agglomeration of high-quality factors to digital information industries, thus developing the emerging digital industries, for instance, the electronic manufacturing industry, and forming digital industrial clusters. With the scale of local digital industries, the unbalanced flow of factors between regions may be intensified. Especially combined with competition and siphon effects, this could further weaken the spillover impact of digital industrialization. For the interaction term, both the total and indirect effects were significantly positive in the short and long term. This showed that digital industrialization played positive regulating roles in the process of marine industrial structure upgrading, affecting the HDME. Digital industrialization not only improves the innovation ability of new digital technologies in the industries by exerting economies of scale effects, but also spreads and penetrates technology to other industries through industry correlation. This accelerates the intelligent and digital transformation of traditional marine industries. Obviously, the structural dividends from marine industrial structure upgrading could be further released under the digital industrialization, thus driving the marine economy.

As listed in Table 11, the estimation coefficient of industrial digitalization was insignificant, which indicated that its positive impact on the marine economy was very limited at this stage. Under three weight matrices, marine industrial structure upgrading positively affected the marine economy. Although the interaction term was not significant under the human capital geographic matrix, W₂, the total effect of the interaction term was significantly positive under the economic geography matrix, W₃. This suggests that the regulatory role of industrial digitalization is affected by the regional land economy and geographical factors. One possible explanation is that although industrial digitalization plays positive roles in improving enterprise’s production efficiency, management, and innovation ability, marine enterprises may be confronted with difficulties in the process of digital transformation, such as high investment costs. Especially in the early stage, when the productivity effect of digital technology has not yet fully emerged, the positive benefits of digital transformation may not be able to make up for its cost, which may inhibit the marine enterprises’ digital transformation, or even lead to interruption. Besides, industrial digitalization may have the characteristics of industry heterogeneity. The economic effects of digital technology on different factor-intensive marine industries are different. Compared with technology-intensive marine industries, labor-intensive marine industries are relatively less dependent on advanced technology and production equipment. This implies that the digital transformation of these traditional marine industries is hard to accomplish in the short term. Furthermore, the agglomeration effect of industrial digitalization may widen the gap of marine industries vertically and increase the unbalanced development of the marine economy between regions laterally, resulting in the Matthew effect between industries and regions. Obviously, the positive effect of digital technology on the marine industrial structure still needs to be further improved.

Overall, there existed differences in the effects of the digital economy sub-dimensional indicators on the marine economy’s quality. Specifically, the positive promotion effect of digital industrialization was the largest, followed by the digital infrastructure, while the industrial digitalization effect was not obvious. The interaction terms of marine industrial structure upgrading with digital infrastructure and digital industrialization were all significantly positive under three weight matrices, whereas the interaction term between industrial digitalization and marine industrial structure upgrading was only significantly positive under W₃. This indicated that, at the present stage, the synergy effect between the digital economy and marine industrial structure upgrading mainly depended on digital infrastructure and digital industrialization.

5.6. Results of the Threshold Model

Given that the different levels of the digital economy sub-dimensional index may lead to different impacts in the process of marine industrial structure upgrading, affecting the HDME, we further used the threshold model, taking digital infrastructure, digital industrialization, and industrial digitalization as the threshold variables. As listed in

Table 12. Testing for the threshold effects of the digital economy sub-dimensional index.

Threshold Variables	Threshold Effect	Threshold Values	F-Statistics	p-Value	1%	5%	10%
lndig1	Single threshold	−3.3845 *	21.87	0.0700	19.5323	22.2182	69.8407
	Double threshold	−2.6232	13.12	0.1867	16.6990	20.1769	24.7591
lndig2	Single threshold	−3.7025 **	27.88	0.0300	18.1631	23.7130	33.8902
	Double threshold	−2.2427	9.11	0.3067	14.6810	17.6436	27.4584
lndig3	Single threshold	−3.6812	10.13	0.4200	19.4033	20.9675	39.7501
	Double threshold	−2.1729	8.48	0.4167	13.3880	17.8130	26.2443

Note: ** and * indicate 5% and 10% significance levels, respectively, and 300 bootstrap replications were employed for each of the three bootstrap tests.

, both digital infrastructure and digital industrialization passed the single-threshold test, whereas industrial digitalization failed to pass the test. Therefore, there existed a single threshold for digital infrastructure and digital industrialization in the process of marine industrial structure upgrading, affecting the HDME, the threshold values of which were −3.3845 and −3.7025, respectively.

The threshold estimation results are shown in

Table 13. The results of the panel threshold model.

Variables	lnHDME	lnHDME
lnind (lndig1 ≤ −3.3845)	0.776 ***
	(5.07)
lnind (lndig1 > −3.3845)	1.236 ***
	(7.92)
lnind (lndig2 ≤ −3.7025)		1.081 ***
		(8.74)
lnind (lndig2 > −3.7025)		0.718 ***
		(5.43)
control variables	Yes	Yes
R2	0.6372	0.7020

Note: *** indicates 1% significance level.

. When lndig₁ ≤ −3.3845, the coefficient of marine industrial structure upgrading was positive and statistically significant. Namely, marine industrial structure upgrading positively affected the HDME. When lndig₁ > −3.3845, the marine industrial upgrading was still significantly positive, and the estimated coefficient increased from 0.776 to 1.236. This indicates that with the advancement of digital infrastructure, the promotional effect of marine industrial structure upgrading on the HDME showed the characteristics of marginally increasing. For the threshold of digital industrialization, the coefficients of marine industrial structure upgrading were always significantly positive when digital industrialization was below and above the threshold value, respectively. This suggests that under the influence of digital industrialization, the marine industrial structure always had positive promotional effects on the marine economy’s quality.

5.7. Robustness Tests

For testing the robustness of the estimation results, we simultaneously constructed three spatial weight matrices, namely, the geographical distance weight matrix, W₁, the human capital geographic matrix, W₂, and the economic geographic matrix, W₃, to estimate the models. Furthermore, considering that the dynamic estimation of the model helped to tackle the endogeneity to some degree, we reported the results of both static and dynamic spatial Durbin models. The signs and significance of core independent variables and the spatial lag terms were basically consistent in the different spatial models with three weight matrices, implying our results were robust.

Based on this, we also further conducted the robustness tests by replacing the core explanatory variable. We used the proportion of marine tertiary industry output value to marine secondary industry to measure marine industrial structure upgrading (lninh).

Table 14. Robustness test, replacing the core explanatory variable.

Variables	W1		W2		W3
	Static	Dynamic	Static	Dynamic	Static	Dynamic
L.lnG		0.202 **		0.202 **		0.246 ***
		(2.14)		(2.12)		(2.59)
W × lnG	−1.129 ***	1.043 ***	−1.068 ***	1.000 ***	−0.792 ***	0.810 ***
	(−4.99)	(4.20)	(−4.70)	(4.06)	(−3.14)	(3.14)
lndig	0.354 ***	0.281 **	0.362 ***	0.284 **	0.388 ***	0.305 ***
	(4.11)	(2.51)	(4.14)	(2.54)	(4.06)	(2.69)
lninh	0.185 ***	0.145 *	0.180 **	0.142 *	0.166 **	0.119
	(2.58)	(1.81)	(2.51)	(1.78)	(2.27)	(1.57)
lndig × lninh	0.122 **	0.149 **	0.109 **	0.141 **	0.075	0.106 *
	(2.34)	(2.34)	(2.08)	(2.24)	(1.39)	(1.71)
W × lndig	1.355 ***	0.947 *	1.310 ***	0.882	1.486 ***	1.030 *
	(2.86)	(1.70)	(2.76)	(1.59)	(2.83)	(1.79)
W × lninh	0.546 **	0.663 **	0.487 *	0.619 **	0.290	0.431
	(2.07)	(2.34)	(1.87)	(2.22)	(1.07)	(1.56)
W × lndig × lninh	0.488 **	0.576 **	0.463 **	0.567 **	0.267	0.392 *
	(2.37)	(2.29)	(2.28)	(2.32)	(1.36)	(1.75)
Control variables	Yes	Yes	Yes	Yes	Yes	Yes
R2	0.7288	0.6654	0.7311	0.6680	0.7220	0.6612

Note: z-statistics are in parentheses. ***, **, and * indicate 1%, 5%, and 10% significance levels, respectively.

shows that the coefficients of lndig and lninh were also basically consistent with the regression results above (see Table 5 and Table 7), verifying the robustness of the results. Particularly, under W₁ and W₂, the significance of the digital economy, marine industrial structure upgrading, and the interaction term coefficient improved.

6. Conclusions and Implications

This study incorporated the digital economy, marine industrial structure upgrading, and high-quality development of the marine economy into a unified analysis framework for the first time to systematically explore the internal mechanisms among them. We measured the HDME index and digital economy comprehensive index, and its sub-dimensional index, by constructing multi-dimensional evaluation index systems during 2012–2020, respectively. We further comprehensively applied the static and dynamic spatial Durbin model and threshold model to investigate the impact of the digital economy and marine industrial structure upgrading on the HDME. The main conclusions are as follows:

(1)

The long-term effects of the digital economy and marine industrial structure upgrading were greater than its short-term effects. Specifically, the digital economy positively affected the HDME of both local and neighboring regions. Marine industrial structure upgrading improved the quality of the marine economy in neighboring areas through the spatial spillover effect. The synergistic effect between the digital economy and marine industrial structure upgrading promoted the regional HDME.

(2)

The sub-dimensional indicators of the digital economy had different impacts on the HDME. The positive impact of digital industrialization was the largest, followed by the digital infrastructure effect, whereas the industrial digitalization effect was not obvious. At this stage, the synergistic effect between the digital economy and marine industrial structure upgrading mainly depended on digital infrastructure and digital industrialization. The industrial digitalization should be improved through promoting the integration between the digital economy and marine industries, thus enhancing the synergistic effect.

(3)

There were digital infrastructure and digital industrialization thresholds in the process of marine industrial structure upgrading, affecting the HDME. Under the influence of digital infrastructure and digital industrialization, marine industrial structure upgrading always positively affected the HDME.

On the basis of this study, we will attempt to carry out the following explorations in the future: First, we will further explore how to promote the deep integration between the digital economy and marine industry by enhancing industrial digitalization. What is the mechanism behind industrial digitalization that affects the synergistic effect of the digital economy and marine industrial structure upgrading? Does the effect of industrial digitalization have regional and spatial heterogeneity? Second, we will take specific marine industries as an example, such as marine fishery and marine manufacturing, to explore whether the digital economy will promote the high-quality development of specific industries, and whether this effect still has the characteristics of the spatial spillover effect.

This study provided new insights for improving the marine economy. First, we could accelerate the depth convergence of the digital economy and marine economy to fully display the multiplication and superposition effects of digital information technology. Therefore, on the one hand, provincial governments should continue to increase investment in the digital infrastructure construction and accelerate the construction of platforms, such as marine big data centers and industrial Internet. On the other hand, all regions must adhere to the innovation-driven strategy and improve the level of digital industrialization and digital transformation. Strengthening the dominant position of enterprises in the innovation process and enhancing the linkage between enterprises, universities, and the market should be considered, which would help to improve the efficiency of joint innovation between enterprises, universities, and research institutes and the transformation efficiency of scientific and technological achievements. Besides, by actively cultivating emerging digital industries and supporting measures such as taxation, local governments can help to build the digital industrial clusters and industrial chains with marine characteristics, and further actively guide the digital transformation of traditional marine industries.

Second, making targeted marine industrial policies could accelerate the optimization of the marine industrial structure, fully exerting its structural dividend effect. Taking into account the actual situations of resource endowment and the industrial base, regions should enhance the independent research and innovation ability and cultivate leading enterprises to develop the emerging marine industries, for example, marine high-end manufacturing, electronic information, and offshore wind power. Moreover, the specialized and high-end development of marine production services, such as blue finance, is also necessary and should be highly valued. This is conducive to strengthening the benign interaction between various sectors of marine industries for technology, thus promoting the high-quality coordinated effects of marine manufacturing and production service industries. Additionally, to avoid the blind imitation and low-quality competition in the marine industry among regions, the interaction and cooperation between regional marine industrial structures should be stressed.

Third, strengthening the synergistic effects between the digital economy and marine industrial structure upgrading could maximize the intensifying function of the digital economy on the structural dividends from marine industrial upgrading. The digital economy and marine industrial structure upgrading have long-term and interactive influences on the marine economy, which will help to boost its quality in the future. Therefore, regions must identify the focus of the effective integration of the digital economy with the local marine industries and increase personnel training and technological investment in digital technology and industrial applications. It is also important to improve the utilization of data, algorithms, and other elements in the marine industry.

Particularly, the depth convergence of digital information technologies and marine industries can not only greatly improve the independent innovation and production capability of marine manufacturing core technology and equipment but can also facilitate the resource sharing between various sectors in the industry chain. The government also needs to accelerate the construction of “smart plus marine industries”, such as smart shipping and smart fishery, to comprehensively improve the productivity of traditional marine industries from all aspects of production and management.