1 Introduction

As global land resources diminish, coastal nations are increasingly recognizing the ocean’s economic potential (Jiang et al., 2014; Kildow & McIlgorm, 2010). According to the United Nations’ ‘Second World Ocean Assessment’, the ocean covers more than 70% of Earth’s surface and supports approximately 12% of the global population’s livelihoods (United Nations, 2020). With its extensive coastline, China has ample marine resources and considerable space for economic growth (Liu et al., 2017). Over the past decade, China’s marine economy has experienced substantial growth and strongly driven its national economy. In 2022, China’s marine gross production value (GOP) surpassed 9.4 trillion yuan, contributing 7.8% to its total gross domestic product (GDP; NDRC & MENR, 2023). However, China’s historically extensive approach to marine economic development has brought challenges such as increased marine environmental pollution, depletion of marine resources, and unbalanced marine industrial structures (An et al., 2022). Consequently, China is aiming urgently to achieve marine high-quality marine economic development (HQMED), which emphasizes minimizing marine resource use and pollution discharge while ensuring rapid economic growth.

Since the introduction of the high-quality development concept in 2017, the Chinese government has introduced several initiatives for HQMED. For example, the 14th Five-Year Plan for Marine Economy Development clearly outlines the promotion of HQMED and establishes designated leading marine economic demonstration zones and specialized Marine Innovation Areas (Guo et al., 2022). Coastal provinces such as Tianjin, Zhejiang, and Hainan have also integrated HQMED into their respective marine economy development strategies. Therefore, the government and academic community are both focusing on HQMED.

Scholars have recently explored various avenues to achieve HQMED, including technological innovation (Liu et al., 2021a, 2021b), financial advancement (Wang et al., 2021), and the digital economy (Jian et al., 2021). As a spatial economic activity, industrial agglomeration (IA) is considered a key factor influencing the realization of environmentally sustainable, green, and high-quality development (Guo & Ma, 2021; Han et al., 2018; Zheng & He, 2022). In Opinions on Promoting High-Quality Marine Economic Development (2018), the Ministry of Natural Resources of China emphasized the importance of nurturing and reinforcing emerging marine industries. The endogenous growth theory asserts that IA can spur innovation through knowledge spillover (KS), enhancing regional development capabilities (Wu et al., 2022). Consequently, the role of KS in the impact of MIA on HQMED has garnered attention (Wen et al., 2021; Xie & Li, 2021; Zhao & Bai, 2009). MIA can attract a substantial talent pool, thereby minimizing information loss during knowledge dissemination and promoting frequent KS (Liu et al., 2016). Furthermore, KS can significantly influence firms’ productivity and innovation, leading to improved production efficiency and reduced resource wastage and environmental pollution, ultimately driving HQMED (Wang & Wang, 2022).

This study uses panel data from China’s 11 coastal provinces from 2008 to 2020 to quantitatively assess marine IA (MIA)’s impact on HQMED and examines whether KS plays a mediating role. This study makes multifaceted contributions to the literature. It introduces a unified analytical framework including MIA, KS, and HQMED. This framework examines the linear and nonlinear impacts of MIA on HQMED, including the mediating effect of KS, thereby improving the marine economy research landscape and delivering critical empirical support for promoting HQMED. Incorporating spatial considerations, the study uses the spatial Durbin model (SDM) to examine MIA’s spatial spillover effect on HQMED and provide deeper insights into the interconnectedness of MIA and HQMED. Finally, from the KS perspective, the study uses mediating and threshold effect models to analyze the mediating and nonlinear impacts of MIA on HQMED, thereby offering valuable insights for developing targeted policies.

2 Literature review

Currently, research on HQMED mainly focuses on three key aspects: its measurement, regional disparities, and influencing factors. Regarding measurement, scholars have pursued two primary approaches: one uses total factor productivity (TFP) to measure HQMED (Liu et al., 2021a, 2021b), and the other argues that depending solely on TFP may not adequately capture factors such as resource allocation efficiency and output structure. Instead, it supports the construction of a comprehensive indicator system to evaluate the quality of economic development. For instance, Gao et al. (2022) proposed an evaluation index system that includes perspectives on the marine economy, resources, and environment. Their findings showed steady improvements in Jiangsu’s HQMED. Lu et al. (2019) evaluated HQMED based on five development concepts—innovation, coordination, greening, openness, and sharing—and identified Shanghai, Guangdong, and Jiangsu as frontrunners in HQMED. Researchers have also examined regional variations in HQMED. An et al. (2022) showed that HQMED in China’s eastern coastal provinces significantly outpaces that in its southern and northern counterparts. Li et al. (2020) highlighted the Yangtze River Delta region as a HQMED hotspot, while the Bohai Rim and Pearl River Delta regions show alternating fluctuations in HQMED. Further analyses revealed that multiple factors, including marine technology innovation (Liu et al., 2021a, 2021b), environmental regulations (Ren & Ji, 2021), and digital economic development (Jian et al., 2021), shape HQMED.

Existing literature has systematically examined the relationship between IA and economic development. Some scholars contend that IA fuels economic progress. For example, Xie and Li (2021) demonstrated IA’s significant role in driving green development in China. Zheng and He (2022) showed IA’s positive impact in China’s Chengdu-Chongqing economic circle on high-quality development, with technology innovation serving as a mediating factor. However, various perspectives suggest a nonlinear relationship between IA and economic development. Guo et al. (2020) identified a U-shaped correlation between IA and economic development in 34 cities in northeast China. Xu et al. (2022a, 2022b) found that agricultural IA only stimulates agricultural development beyond a certain threshold. Other studies argue that IA’s association with economic development is influenced by other factors. Zhang et al. (2022a, 2022b) proposed that agricultural IA positively impacts the development of a sustainable agricultural economy in regions with lower per capita GDP. Guo and Ma (2021) used the System Generalized Method of Moments to reveal the moderating role of environmental regulations in the IA–economic development relationship.

According to the new economic geography theory and endogenous growth theory, KS is a key pivotal factor influencing spatial agglomeration, innovation, and regional economic development (Xu et al., 2022a, 2022b; Zhao & Bai, 2009). Arrow (1962) was the first to elucidate the role of spillover effects in economic development, highlighting how fresh investments enhance productivity through cumulative production experience. Romer (1987) underscored knowledge and technology R&D as the bedrock of economic development, characterized by noncompetitiveness and partial excludability. Studies have also explored the impact of KS on energy efficiency (Sun et al., 2021) and sustainable development (Aldieri et al., 2022; Huber, 2012). Concerning IA’s influence on economic development, Xie and Li (2021) and Chang and Oxley (2009) observed that IA fosters interindustry linkages, amplifies KS, and reduces knowledge diffusion costs, thereby improving green TFP. Wen et al. (2021) showed the mediating role of specialized KS in the relationship between port-IA and regional economic development. However, no such effects are found in diversified and competitive KSs.

In summary, although earlier literature has established a foundation, certain limitations need refinement. First, studies on the relationship between IA and economic development have focused on regional and manufacturing domains while ignoring the marine industry. Second, few studies have explored the relationship between IA and economic development from a KS perspective. This study aims to bridge this gap in the literature. Third, most studies have concentrated on the linear impact of IA on economic development, overlooking its nonlinear and spatial spillover effects, thereby lacking in-depth and comprehensive analysis. This study combines MIAs, KS, and HQMED into an integrated analytical framework to address these gaps. Using the SDM, mediating effect model, and threshold effect model, this study aims to improve comprehension of the connections between MIAs and HQMED in China and considers linear, nonlinear, and spatial spillover effects.

3 Mechanism analysis

3.1 Direct effect

The influence of MIA on HQMED is mainly realized through externalities, based on the Marshall–Arrow–Romer externalities theory (Marshall, 1920). This mechanism can be understood in several dimensions. First, the scale effect comes into play. According to the Marshall–Arrow–Romer theory of externalities, IA contributes to economies of scale (Marshall, 1920; Romer, 1986). Specifically, IA increases firms’ capacity for innovation, resource utilization, and pollution management efficiency, thereby creating an environment that fosters regional innovation (He et al., 2022). Second, the sharing effect involves the sharing of achievements and risks. Specifically, IA facilitates the sharing of infrastructure, information, and cleaner technologies, resulting in reduced production and transaction costs (Xu et al., 2022a, 2022b). Concurrently, IA reduces the risks associated with technological innovation for firms, thereby accelerating the generation and dissemination of environmentally friendly technologies. Third, the competition effect manifests itself. On the one hand, IA encourages firms to pursue technological and process innovations, which hastens the adoption of low-carbon and cleaner technologies and increases production efficiency across the industrial chain (Xu et al., 2022a, 2022b). On the other hand, excessive competition among firms might erode profit margins, potentially undermining incentives for energy-saving and emission-reduction initiatives (Pei et al., 2021). Finally, the congestion effect emerges. The concentration of marine industries can lead to increased resource consumption and environmental pollution, which in turn increases the cost of environmental management (Guo & Ma, 2021). Moreover, increased prices for production factors, including land and labor, result in additional production costs for firms (Lu et al., 2021a, 2021b). In agglomeration areas, traffic congestion can also lead to higher living expenses (Guo et al., 2020). Given the current relatively modest level of MIA development in China, the congestion effect has a limited impact.

From this analysis, we derive the following hypothesis:

  • H1: MIA has a positive and facilitating impact on HQMED.

3.2 Spatial spillover effect

MIA, as a spatial economic phenomenon, impacts beyond their local boundaries, producing spatial spillover effects that echo and shape the surrounding HQMED (He et al., 2022; Lu et al., 2021a, 2021b; Zheng & He, 2022). This phenomenon can be understood from economic development and environmental pollution perspectives. From an economic development standpoint, local IA can stimulate the creation and dissemination of clean technologies and novel knowledge, triggering a cascade effect that advances HQMED in neighboring regions (Chen et al., 2020). Moreover, local IA attracts labor, capital, and other production factors from adjacent areas, potentially worsening regional disparities and yielding polarization effects (Song et al., 2023). Notably, such effects are more pronounced in marine ecosystems than in terrestrial economies. From an environmental pollution perspective, pollutants generated within the boundaries of the local MIA can easily disperse into surrounding regions, curbing HQMED in those areas and leading to a phenomenon known as the ‘pollution paradise’ effect (Hao et al., 2022). Conversely, if the local MIA increases the efficiency of pollution treatment, reduces the costs of pollution control, or establishes centralized pollution management systems, it could reduce pollution emissions in neighboring regions, thereby leading to a ‘pollution halo’ effect (Wang et al., 2019).

Based on this analysis, the following hypothesis is posited:

  • H2: Local MIA contributes to a spatial spillover effect, influencing the surrounding HQMED.

3.3 Mediating effect

MIA can expedite the flow and metamorphosis of knowledge, thereby indirectly advancing HQMED. This phenomenon can be analyzed from three distinct perspectives. First, from an industrial chain viewpoint, MIAs can distribute innovative elements and production factors across the agglomerated space through the industrial chain (Huang et al., 2022). Essentially, agglomerated firms can share resources such as clean technologies, expert technical personnel, and service sales platforms. This collaborative sharing reduces information costs, optimizes resource allocation efficiency and production performance, and ultimately enhances HQMED (Xu et al., 2022a, 2022b). Second, when viewed from the perspective of knowledge accumulation, MIA helps reduce information loss during the process of knowledge dissemination. This acceleration of industrial information and resource integration reduces information asymmetry faced by firms (Yuan et al., 2022). Moreover, technology alliances catalyzed by MIA foster the generation of novel knowledge, thereby enhancing TFP (Song et al., 2023). Finally, from a technological advancement perspective, Romer (1987) argued that the spillover effects of knowledge catalyze economic growth via technological progress. Geographical constraints, however, often restrict the scope of these KS effects (Wang & Wang, 2022). MIA assumes the role of augmenting channels for KS. For instance, they escalate the demand for talent while concurrently expanding avenues for information and KS. This holistic ecosystem promotes the research and development of sustainable technologies, thereby strengthening HQMED.

Thus, we propose the following hypothesis:

  • H3: MIA fosters KS, thereby indirectly driving HQMED.

The theoretical mechanism analysis framework is displayed in Fig. 1.

Fig. 1
figure 1

Influence mechanism of MIA on high-quality marine economic development

Full size image

4 Model, variables, and data

4.1 The entropy method

The entropy method is often used to establish the weight of the indicator. The main steps are as follows:

Step 1: Data standardization. The original data are standardized using the following formula:

$${Y_{hij}} = \left\{ \begin{gathered} \frac{{{X_{hij}} - \min \left\{ {X{}_j} \right\}}}{{\max \left\{ {X_j} \right\} - \min \left\{ {X_j} \right\}}}\left( {{X_{hij}}{\text{ is positive index}}} \right) \hfill \\ \frac{{\max \left\{ {X_j} \right\} - {X_{hij}}}}{{\max \left\{ {X_j} \right\} - \min \left\{ {X_j} \right\}}}\left( {{X_{hij}} {\text{is negative index}}} \right) \hfill \\ \end{gathered} \right.$$
(1)

where Xhij (h = 1,2,…,n; i = 1,2,…, r; j = 1,2,…, m) represents the value of indicator j for region i in year h; n, r, and m denote the number of years, sample size, and number of evaluation indicators, respectively; min {Xj} and max {Xj} denote the minimum and maximum values of indicator j across all years, respectively.

Step 2: Calculation of the information entropy (Ej):

$${E_j} = - \frac{1}{\ln (nr)}\sum\limits_{i = 1}^r {\left( {\frac{{{Y_{ij}}}}{{\sum\limits_{i = 1}^m {{Y_{ij}}} }}\ln \frac{{{Y_{ij}}}}{{\sum\limits_{i = 1}^m {{Y_{ij}}} }}} \right)}$$
(2)

Step 3: Calculation of the coefficient of variation (Fj):

$${F}_{j}=1-{E}_{j}$$
(3)

Step 4: Calculation of the weight coefficient (Wj):

$${W_j} = \frac{{F_j}}{{\sum\limits_{j = 1}^m {F_j} }}$$
(4)

Step 5: Calculation of HQMED:

$$HQME{D_{hi}} = \sum\limits_{j = 1}^m {{W_j}X{}_{hij}}$$
(5)

4.2 Benchmark model

This study uses spatial econometric models as the benchmark model, which mainly include three forms: spatial lag model (SLM), spatial error model (SEM), and SDM. Its general expression is as follows:

$$\begin{array}{c}HQME{D}_{it}={\alpha }_{0}+{\rho }_{1}{W}_{ij}HQME{D}_{it}+{\alpha }_{1}MI{A}_{it}+{\alpha }_{c}{X}_{it}+{\theta }_{1}{W}_{ij}MI{A}_{it}+{\theta }_{c}{W}_{ij}{X}_{it}+{\mu }_{it}\\ {\mu }_{it}=\lambda {W}_{ij}{\mu }_{it}+{\varepsilon }_{it}\end{array}$$
(6)

where HQMEDit denotes the level of HQMED of province i in year t, MIAit refers to the MIA level of province i in year t, and Xit denotes the control variables. \({\rho }_{1}\) represents the spatial autoregressive coefficient of HQMED. \({\mu }_{it}\) and \({\varepsilon }_{it}\) represent the random disturbance terms. If \({\rho }_{1}\ne 0,\theta =0,\) and \(\lambda =0\), the above model degenerates to the SLM. If \({\rho }_{1}=0,\theta =0,\) and \(\lambda \ne 0\), the above model degenerates to the SEM. If \({\rho }_{1}\ne 0,\theta \ne 0,\) and \(\lambda =0\), the above model is the SDM. The spatial weight matrix (Wij) is a key element in spatial econometric models. With the development of modern economic society, the factors affecting spatial correlation between variables are no longer only limited to geographical distance but also include economic development. Hence, we construct an economic–distance spatial weight matrix, considering the difference in economic development among regions. The formulas are given as follows:

$${W_{ij}} = \left\{ \begin{array} {ll}0,i = j \hfill \\ \frac{1}{{{d_{ij}}}} \times diag({\overline Y_1}/\overline Y ,\overline {Y_2} /\overline Y , \ldots ,\overline {Y_r} /\overline Y ),i \ne j \hfill \\ \end{array} \right.$$
(7)

where dij denotes the distance between cities i and j, which is calculated using the latitude and longitude coordinates of the provincial capital cities.\({\overline Y_i}\) and \(\overline Y\) represent the average per capita GDP of city i and the average per capita GDP of all sample cities during the study period, respectively.

4.3 Mediating effect

The mediating effect model is adopted to examine whether MIA has an impact on HQMED through KS. Specifically, we take HQMED as the explained variable, MIA as the core explanatory variable, and KS as the mediator. The mediating effect model is as follows:

$${KS}_{it}= {\beta }_{0} +{p}_{2}{W}_{ij}{KS}_{it} + {\beta }_{1}{MIA}_{it} + {\beta }_{c}{X}_{ir}+ {\varphi }_{1}{W}_{ij}{MIA}_{it}+ {\varphi }_{1}{W}_{ij}{X}_{it} + {\mu }_{it}$$
(8)
$${HQMED}_{it}={\gamma }_{0}+{\rho }_{3}{W}_{ij}{HQMED}_{it}+{\gamma }_{1}{MIA}_{it}+{\gamma }_{2}{KS}_{it}+{\gamma }_{c}{X}_{it}+{\varphi }_{1}{W}_{ij}{MIA}_{it}+{\varphi }_{2}{W}_{ij}{KS}_{it}+{\varphi }_{{\text{c}}}{W}_{ij}{X}_{it}+{\mu }_{it}$$
(9)

where \({\alpha }_{1},{\gamma }_{1},\) and \({\beta }_{1}\times {\gamma }_{2}\) represent the total effect, direct effect, and mediating effect of MIA on HQMED, respectively. If \({\alpha }_{1},\)\({\beta }_{1}\), \({\gamma }_{1},\) and \({\gamma }_{2}\) are all significant, there exists a partial mediating effect. If \({\alpha }_{1},{\beta }_{1}\), and \({\gamma }_{2}\) are significant, and \({\gamma }_{1}\) is not significant, there exists a complete mediating effect. If at least one of \({\beta }_{1}\) and \({\gamma }_{2}\) is not significant, there is no mediating effect.

4.4 Variable selection

4.4.1 Explained variable

Due to data availability, the research sample covers China’s 11 coastal provinces, including municipalities, from 2008 to 2020 (Fig. 2). The dependent variable examined is HQMED. Based on the works of Jian et al. (2021) and Lu et al. (2019), an evaluation index system for HQMED is created. This system is based on five core dimensions: innovation, coordination, greening, openness, and sharing. It consists of 5 primary indicators and 25 secondary indicators, as shown in Table 1.

Fig. 2
figure 2

Geographical distribution of 11 coastal provinces

Full size image
Table 1 Evaluation index system of high-quality marine economic development
Full size table

In this framework, the upgrading index of the marine industrial structure is measured by calculating the ratio of the output value of the ith marine industry to the total marine gross product value. Meanwhile, the level of high-level marine industrial structure is quantified by establishing the proportion of the output value of the marine tertiary industry relative to the output value of the marine secondary industry. Furthermore, the land–sea economic correlation is determined by applying the gray correlation degree between GDP and GOP. The expression can be written as follows:

$$\xi \left[{y}_{1}\left(k\right), {y}_{2}\left(k\right)\right]=\frac{{\text{max}}|{y}_{1}\left(k\right), {y}_{2}\left(k\right)|}{{\text{max}}\left|{y}_{1}\left(k\right)-{y}_{2}\left(k\right)\right|+2|{y}_{1}\left(k\right)-{y}_{2}\left(k\right)|}$$
(10)

where y1(k) and y2(k) denote GDP and GOP, respectively. Subsequently, the HQMED index for China’s 11 coastal provinces during the 2008–2020 timeframe is derived using formulas (1)–(5).

4.4.2 Core explanatory variable

The core explanatory variable is MIA. The location entropy index is one of the commonly used measures of IA, which can eliminate endogenous effects arising from regional differences (Wu et al., 2022). Therefore, we use the location entropy index to estimate MIA as follows:

$${MIA}_{it}=\frac{{Q}_{it}/{Q}_{i}}{{Q}_{kt}/{Q}_{k}}$$
(11)

where Qit denotes the GOP in province i in year t, Qi denotes the GDP of province i, Qkt denotes the total GOP in 11 coastal provinces in year t, and Qk denotes the total GDP of the 11 coastal provinces in the same year.

4.4.3 Mediating variable

The mediating variable is KS. With reference to Liu et al. (2016) and Wang and Wang (2022), the logarithm of the number of scientific and technological personnel engaged in marine research activities in each province is selected as a proxy variable for KS.

4.4.4 Control variables

To address endogeneity issues arising from omitted variables, five control variables are included in the benchmark model:

(1) The infrastructure level (INF)

INF plays a crucial role in aggregating labor and capital and promoting the dissemination of new knowledge and technologies, all of which have a positive impact on marine economy development (Zhang & Wang, 2021). However, it is important to consider the potential negative impacts of large-scale infrastructure construction on the marine environment, such as increased wastewater and solid waste discharge due to the consumption of gravel, soil, and energy (Ren et al., 2023). In this study, we use the metric of kilometers of road per square kilometer to measure INF.

(2) Urbanization level (UR)

UR is a key factor that influences environmental quality because the promotion of urbanization encourages residents to replace fossil fuels with clean energy, thereby reducing energy usage and pollution emissions (Zhao & Wang, 2022). However, the urbanization process significantly reduces the area of wetlands, forests, and grasslands, which weakens the atmospheric dispersion capacity of cities. We follow Wu et al. (2022) to use the urbanization rate as a measure of UR.

(3) Marine economic development level (PGOP)

PGOP represents the overall level of economic development related to the ocean and is considered a factor influencing HQMED (Wang et al., 2021; Yang et al., 2021). Following the study by Liu et al., (2021a, 2021b), we use per capita marine GDP to measure PGOP.

(4) International trade (OPEN)

OPEN is an important channel for promoting technological progress and a key driver for achieving high-quality development (Zhao et al., 2022). International trade facilitates the diffusion of technology among trading partners, resulting in the technology spillover effect. However, the pollution haven hypothesis suggests that slack trade openness policies can harm the environment. With reference to Ren and Ji (2021), we adopt the ratio of total imports and exports to provincial GDP as a measure of OPEN.

(5) Government fiscal expenditure (GOV)

GOV drives rapid economic development as it can affect the mobility of production factors, such as labor, capital, and technology, and promote environmental quality (Zhang et al., 2019). Referring to Guo and Ma (2021), we use the ratio of government fiscal expenditure to provincial GDP to measure GOV.

4.5 Data sources

The original data for the aforementioned variables are sourced from the China Ocean Economic Statistical Yearbook, China Statistical Yearbook, China Energy Statistical Yearbook, China Environmental Statistical Yearbook, and various provincial and municipal statistical yearbooks. To address heteroscedasticity, some variables are converted to their logarithmic forms. Additionally, all price-related variables are converted into constant price data, with 2008 as the base year to eliminate the impact of price fluctuations. The descriptive statistics of the aforementioned variables are presented in Table 2.

Table 2 Descriptive statistics of selected variables
Full size table

5 Empirical results and discussions

5.1 The spatiotemporal features of HQMED

To understand the dynamic evolution characteristics of HQMED across China’s 11 coastal provinces between 2008 and 2020, we present its kernel density curve (KDC) in Fig. 3. There is a distinct trend as the KDC gradually shifts to the right. This shift is accompanied by an increase in the center of density function, from 0.240 in 2008 to 0.254 in 2020, highlighting the progressive improvement of HQMED in China’s coastal provinces. Furthermore, the width of KDC substantially decreases while the height of the main peak shows a significant increase. These dual trends collectively suggest a gradual convergence of HQMED disparities among the coastal provinces. For further analysis, we present the spatial trajectories of HQMED in China’s 11 coastal provinces for the years 2008, 2012, 2016, and 2020 in Fig. 4. This temporal analysis exposes a fluctuating upward trajectory of HQMED over the 2008–2020 interval, with the metric increasing from 0.188 to 0.266. This longitudinal pattern clearly shows the progressive improvement in China’s marine economic development quality. This improvement aligns seamlessly with the conclusions drawn from the KDC analysis.

Fig. 3
figure 3

Kernel density curve of high-quality marine economic developent from 2008 to 2020

Full size image
Fig. 4
figure 4

Distribution of high-quality marine economic development in China’s 11 coastal provinces in 2008, 2012, 2016, and 2020

Full size image

Furthermore, we found substantive discrepancies in HQMED among various provinces, with the hierarchy being as follows: Shanghai > Guangdong > Tianjin > Shandong > Jiangsu > Hainan > Fujian > Zhejiang > Liaoning > Hebei > Guangxi. During the sample period, Hainan, Shanghai, and Guangdong stand out as regions demonstrating rapid advancements in HQMED. These findings highlight the dynamic interplay of HQMED across China’s coastal provinces over the examined timeframe. The combination of KDC insights and spatial analysis provides a comprehensive perspective on the evolving landscape of marine economic development quality.

5.2 Spatial correlation test of HQMED

Using the economic–distance spatial weight matrix, we examined HQMED using the Global Moran’s I index. The findings of this analysis are presented in Table 3. The Moran’s index consistently shows positive values across the sample period. Importantly, a significant number of these indices reject the null hypothesis of lacking spatial autocorrelation at a 10% significance level. This convincing observation highlights the significant spatial dependence inherent in HQMED. This spatial dependence indicates two clustering characteristics: a ‘High–High’ adjacency, which indicates that regions with higher HQMED values tend to cluster together, and a ‘low–low’ clustering, in which regions showing lower HQMED values tend to cluster. This observation supports the rationale for using spatial econometric models for subsequent empirical analysis. In essence, these results not only validate the appropriateness of using spatial econometric techniques but also explain the pronounced spatial patterns that support HQMED’s distribution across the examined regions.

Table 3 Global Moran’s I index of high-quality marine economic development
Full size table

Global Moran’s I index reflects the overall spatial correlation, which may ignore the atypical characteristics of local areas. Next, the local Moran’s I index of HQMED is calculated, and the Local Indicators of Spatial Association maps for the years 2008, 2012, 2016, and 2020 are plotted, as shown in Fig. 5. Most coastal provinces exhibit spatial clustering patterns of either ‘High–High’ or ‘Low–Low’ for HQMED.

Fig. 5
figure 5

Spatial agglomeration map of HQMED in 11 coastal provinces in 2008, 2012, 2016, and 2020

Full size image

5.3 Benchmark regression results

Before proceeding with benchmark regression, selecting a suitable spatial econometric model is essential. The outcomes of the Lagrange Multiplier test, Wald test, and Likelihood Ratio test all lead to rejecting the null hypothesis, which posits that the SDM degrades to either the SLM or the SEM. Additionally, the findings from the Hausman test indicate that the random effects model is more suitable for this analysis. Therefore, the SDM based on random effects is adopted as the benchmark model. The results of this benchmark model are summarized in Table 4.

Table 4 Results of the benchmark model
Full size table

The analysis shows that the spatial autocorrelation coefficient (ρ) assumes a significantly positive value. This finding aligns with the work of Zheng and He (2022) and implies a positive spatial correlation of HQMED across China’s coastal provinces. Furthermore, the coefficient for MIA is 0.061, indicating statistical significance at the 1% level. This outcome highlights MIA’s influence on significantly fostering HQMED. Thus, this finding supports the validity of H1. Moreover, the coefficient of W*MIA is 0.041, indicating statistical significance at the 1% level. This result shows a significant positive spatial spillover effect arising from the impact of MIA on HQMED. This intriguing insight underscores a ‘trickle-down’ mechanism in which local MIAs significantly improve the HQMED of their proximate regions. Consequently, this finding confirms the validity of H2. The reasons may be that MIA facilitates the spillover of several environmental protection technologies and that the scale effect of pollution control may play a role. With the spillover of technology and the transfer of industries, the surrounding region’s economy will also develop.

In terms of control variables, the coefficients for UR and OPEN show significant positive associations with local HQMED. Interestingly, the coefficients for W*UR and W*OPEN, representing their spatial spillover effects, are significantly negative. This intriguing trend implies that although urbanization and international trade positively impact HQMED within their immediate regions, they adversely affect the HQMED of neighboring areas. This aligns with the findings of Zheng et al. (2023). The underlying mechanism appears to involve a ‘siphoning effect’, in which increased UR and OPEN levels in local provinces attract labor, capital, and production factors from surrounding provinces, thereby impeding their own HQMED. The coefficient related to PGOP assumes a significant negative value, indicating that marine economy development does not inherently facilitate HQMED. This suggests a prevailing trend of extensive marine economy development that relies heavily on resource utilization and pollution emissions at a lower-quality developmental stage, which confirms the observations of Liu et al. (2021a, 2021b). Moreover, the coefficient of INF is significantly positive, suggesting that reinforcing infrastructure construction contributes to HQMED. This relationship is likely due to the high-level infrastructure requirements of various marine industries, such as the development and use of marine resources, shipping logistics, and marine tourism. This conclusion aligns with the findings of Guo et al. (2022). Finally, the coefficient of GOV lacks statistical significance, suggesting that its role, as described in the model, does not significantly influence the variations in HQMED.

Following the framework proposed by LeSage and Pace (2008), we present the benchmark results describing the effects as direct (related to MIA’s impact on local HQMED), indirect (related to MIA’s influence on surrounding HQMED), and total effects (the aggregate of direct and indirect effects). In columns (2)–(4) of Table 4, the coefficients for the direct, indirect, and total effects of MIA all exhibit statistical significance at the 1% level. Remarkably, the coefficient for the indirect effect (0.087) surpasses that of the direct effect (0.069).

This intricate pattern signifies that MIA not only improves local HQMED but also creates a significant positive spatial spillover effect on proximate provinces. The latter helps strengthen their HQMED. This finding clearly confirms H2 and thus emphasizes the pivotal role of MIA in shaping not just the immediate locality’s marine economic quality but also in fostering positive development in the surrounding regions.

5.4 Robustness tests

To bolster the credibility of our benchmark findings, we subject them to rigorous robustness testing using three methods. The test results are shown in Table 5.

Table 5 Results of robustness tests
Full size table

Our first method addresses potential endogeneity concerns arising from the bidirectional causality between MIA and HQMED. This intricate relationship results from the fact that although HQMED can influence MIA, provinces with higher HQMED are also more likely to foster MIA. To address this issue, we adopt a dynamic SDM regression, as recommended by Zheng et al. (2023). The results are presented in columns (1)–(2) of Table 5. Remarkably, we find that in the short and long term, the direct, indirect, and total effect coefficients of MIA on HQMED have significant positive associations.

The second method involves transforming the economic–distance matrix used in our benchmark model into an economic matrix. This modification allows us to account for potential variations in spatial relationships. The outcomes of this adjustment are presented in column (2) of Table 5.

Finally, our third approach uses principal component analysis (PCA) to compute HQMED, replacing the entropy method used in the benchmark analysis. The estimation results derived using this method are displayed in column (3) of Table 5.

Remarkably, the estimated outcomes derived from these three methods show remarkable consistency with our benchmark results. This convergence confirms the robustness of our benchmark findings.

5.5 Results of the mediating effect

Having established KS as a theoretical mechanism through which MIA influences HQMED, this section uses a spatial mediating effect model to empirically substantiate this relationship. The outcomes of this analysis are presented in Table 6.

Table 6 Estimation results of mediating effect model
Full size table

In columns (1)–(2), MIA shows a significant positive influence on HQMED and KS, characterized by coefficients of 0.061 and 0.587, respectively. These coefficients reach statistical significance at least at the 5% level.

In column (3), the coefficients of MIA and KS both retain their significant positive attributes. This observation indicates that both MIA and KS can significantly promote HQMED. Importantly, this means that KS plays a positive and partial mediating role in the relationship between MIA and HQMED. The mediating effect is quantified at 0.037 (calculated as 0.587 multiplied by 0.063), signifying approximately 60.66% of the total effect. This phenomenon agrees with the findings of Xie and Li (2021) and provides concrete empirical evidence highlighting the pivotal role of KS as a channel through which MIA influences HQMED. This result supports H3 and contributes to a comprehensive understanding of this complex dynamics. Therefore, according to the MIA’s stage of development, a better strategy is needed to have an agglomeration effect in terms of the KS.

5.6 Further analysis: threshold effect

5.6.1 Model specification

To further clarify the moderating role played by KS between MIA and HQMED, we examine their relationship from a nonlinear perspective. Referring to Hansen (2000), we construct the following threshold effect model:

$${HQMED}_{it}={{\varvec{\gamma}}}_{0}+{{\varvec{\gamma}}}_{1}{MIA}_{it}*I\left({KS}_{it}\le{\varvec{\lambda}}\right)+{{\varvec{\gamma}}}_{2}{MIA}_{it}*I\left({KS}_{it}>{\varvec{\lambda}}\right)+{{\varvec{\gamma}}}_{3}{X}_{it}+{{\varvec{\varepsilon}}}_{it}$$
(12)

where KSit indicates the threshold variable; \(\lambda\) is the threshold value; I (·) is an indicator function that returns 1 when the condition in the parentheses is satisfied and returns 0 otherwise.

Given that this model ignores the spatial spillover effect of MIA on HQMED, following Zhang et al. (2022a, 2022b) and Lu et al., (2021a, 2021b), a spatial threshold regression model is further constructed:

$${MEHQD}_{it}={{\varvec{\gamma}}}_{0}+{{\varvec{\gamma}}}_{1}W*{MIA}_{it}*I\left({KS}_{it}+W*K{S}_{it}\le{\varvec{\lambda}}\right)+{{\varvec{\gamma}}}_{2}W*{MIA}_{it}*I\left({KS}_{it}+W*{KS}_{it}>{\varvec{\lambda}}\right)+{{\varvec{\gamma}}}_{3}{X}_{it}+{{\varvec{\varepsilon}}}_{it}$$
(13)

where W*MIAit represents the spatial lag term of MIA. KSit + W * KSit are threshold variables used to reflect the total KS of the local and the surrounding provinces. The remaining variables are consistent with the benchmark regression.

5.6.2 Threshold effect test

To confirm the presence of a threshold effect, we used the Bootstrap method and iterated 400 times. The results are presented in Table 7 and illustrated in Fig. 6.

Table 7 Results of threshold effect test
Full size table
Fig. 6
figure 6

Threshold effect test results

Full size image

Our observations show that in both the threshold effect model and the spatial threshold effect model, a single threshold test consistently rejects the null hypothesis, reaching a 1% significance level. Conversely, the double threshold does not meet the significance criterion. Overall, this insight indicates that both models exhibit a solitary threshold, with estimated threshold values of 8.4822 and 13.4879, as presented in Table 8 and Fig. 6.

Table 8 Threshold estimated value
Full size table

Given this information, we conducted an empirical evaluation of the nonlinear impact of MIA on both local and surrounding HQMED using a single threshold model. This sophisticated approach enables a deeper understanding of the intricacies that govern the dynamics between MIA and HQMED.

5.6.3 Results of threshold effect analysis

The outcomes of the threshold effect estimation are provided in Table 9, with columns (1) and (2) presenting the results for the threshold and spatial threshold effect models, respectively.

Table 9 Regression results of threshold effect model
Full size table

Column (1) demonstrates that the coefficients of MIA are consistently and significantly positive across both threshold intervals. These values are 0.085 and 0.252, respectively, signifying that MIA consistently contributes to the advancement of local HQMED, irrespective of whether KS is below or above the threshold value. Intriguingly, the promoting effect of MIA on local HQMED increases as the scale of KS expands.

Column (2) shows that using KS as the threshold variable reveals a significant single threshold effect of MIA on surrounding HQMED. Specifically, within the low-level KS interval (KS + W * KS < 13.4879) and high-level KS interval (KS + W * KS ≥ 13.4879), the influence of local MIA on improving the surrounding HQMED is both positive and statistically significant at the 1% level. When the KS level falls below the threshold value, each 1% increase in local MIA yields an average increment of 0.178% in surrounding HQMED. Even when the KS level exceeds the threshold, the positive impact of local MIA on surrounding HQMED persists, with the impact coefficient experiencing a 0.314 increase. This alignment with the findings of Zheng and He (2022) emphasizes that a higher KS level increases the potency of local MIA in improving surrounding HQMED.

Considering the KS threshold interval identified within the spatial threshold effect model, we divide China’s 11 coastal provinces into two groups: the low-level (KS + W * KS < 13.4879) and high-level KS groups (KS + W * KS ≥ 13.4879), as presented in Table 10.

Table 10 Threshold values and their distribution by region
Full size table

Our analysis shows that only Shanghai and Guangdong provinces have KS levels exceeding the determined threshold value. These provinces stand out because of their increased ability to foster surrounding HQMED. An explanation for this phenomenon lies in the relatively concentrated presence of marine science and technology talents within Shanghai and Guangdong. This concentration facilitates the seamless dissemination and diffusion of novel technologies, products, and models across provincial boundaries, collectively contributing to the overarching promotion of HQMED.

6 Conclusions and policy implications

6.1 Conclusions

Using the panel data covering China’s 11 coastal provinces from 2008 to 2020, we adopted the entropy method to quantify the HQMED. We then empirically examined the direct, mediating, and nonlinear impacts of MIA on HQMED. Our key findings can be summarized as follows: (1) From 2008 to 2020, the overall HQMED in China’s coastal provinces showed a consistent upward trend. The disparity in HQMED across these provinces showed a gradual reduction. (2) The influence of MIA extends beyond merely enhancing local HQMED. It shows a positive spatial spillover effect that echoes across the neighboring provinces, thereby improving their HQMED as well. (3) Our mediating effect analysis shows that KS functions as a constructive and partially mediating intermediary between MIA and HQMED. This mediating role is highlighted by its effects, contributing 60.66% to the total observed effects. (4) By analyzing the threshold effects, we show that when KS is chosen as the threshold variable, there is a significant and single threshold impact of MIA on local and surrounding HQMED.

6.2 Policy implications

Based on these conclusions, we make the following policy recommendations:

(1) Given the significant role of MIA in promoting HQMED, governments should adopt specific measures to strengthen MIA while considering the carrying capacity of marine resources and environment within a reasonable range. First, outdated production capacity should be eliminated in a timely manner by phasing out firms with outdated technologies and high resource consumption, guiding the industry to shift from quantity to quality agglomeration. Second, marine industrial resource allocation should be optimized by rationalizing industrial layout and avoiding redundant construction and resource waste, thereby addressing potential congestion and excessive competition effects. In addition, cross-regional marine industrial cooperation can be strengthened through regional linkage effects by promoting resource sharing, technology exchange, and collaboration among regions to achieve complementary advantages and jointly promote HQMED. This can serve as an important means to achieve HQMED and fully use the positive spatial spillover effect of MIA on HQMED.

(2) Given that KS is an important channel for MIA to influence HQMED, the following measures can help further deepen the KS effect of MIA and build a favorable marine innovation environment. First, through policy and resource support, encouraging the mobility and cooperation of marine science and technology talents among different regions, promoting knowledge exchange and technological cooperation among different regions can promote KS in MIA areas. Second, establishing green technology exchange platforms among MIA areas can promote technological cooperation and sharing among different regions and the dissemination and application of green technologies in environmental protection, sustainable development, and other fields among MIA areas, thereby achieving KS. Third, the KS effect can be achieved by incentives such as setting up research project funds and protecting intellectual property rights, guiding firms and research institutions to conduct technological innovation in MIA areas, and promoting the dissemination and promotion of innovation achievements within and outside the agglomeration areas.

HQMED is influenced by various factors. Therefore, to promote HQMED, governments should adopt specific measures to strengthen infrastructure construction, promote urbanization, and facilitate foreign trade exchanges. First, strengthening infrastructure construction is crucial to improving the level of HQMED, including the construction and improvement of marine infrastructure such as ports, terminals, and waterways to enhance the production, transportation, and circulation efficiency of marine industries and promote IA and KS. Second, reasonable promotion of urbanization plays a positive role in HQMED. Urbanization can promote population and resource agglomeration, form talent, technology, and market advantages for marine economic development and drive industrial upgrading and innovation-driven development. In addition, facilitating foreign trade exchanges is also of great significance to HQMED. International trade and foreign trade cooperation can introduce advanced technologies, management experiences, and market resources and promote internationalization and HQMED.