Integration and innovation of China’s manufacturing and logistics industries and carbon emissions

Abstract

Integration and innovation of manufacturing and logistics industries (IIML) has become important on improving economic systems, with the concern of the carbon neutral targets. However, there is still a lack of studies on how IIML affects regional carbon emission (RCE). By applying jointly resource dependence theory and institutional theory, we construct an empirical model based on panel data of 30 provinces in China from 2014 to 2021. Our results show that: (1) the improvement of IIML level is slow and has an inverted U-shape impact on RCE; (2) the mediating effects of the industrial structure improvement (ISI) promotion is not supported, but the mediating effect of the technological innovation (TI) promotion is supported; (3) in some regions, external regulations (government regulation (GR) and social trust (ST)) make the inverted U-shaped relationship between IIML and RCE flatter; (4) the heterogeneity analysis reveals that economically developed provinces and provinces with diversified industrial structures are more effective in reducing RCE. This study points out theoretical guidance and policy insights for promoting IIML to realize the carbon-neutral plan.

Introduction

With the implementation of the dual-carbon plan, the manufacturing industries and the logistics industries, as key industries in the national economy, have an important impact on the realization of the carbon neutral targets (Khurshid et al., 2023). Concerning the manufacturing industry, China has gradually transformed into a leading manufacturing country in the world, but the overall carbon emission level is still high. As a major contributor to regional carbon reduction (RCE), the energy consumption of the manufacturing industry amounts up to 85% of the total energy consumption (Wang et al., 2023). Moreover, the logistics industry is developing rapidly, driven by a series of policies and market changes. As a link between production and consumption, it utilizes significant quantities of energy, and subsequently produces many RCE. Data shows even more serious RCE problems in 2030 with 80% more energy consumption if the logistics industry do not take any measure (Pan et al., 2020). Therefore, to realize low-carbon transformation, it becomes urgent to investigate how the carbon emission level is related to the industries.

With the emergence of technological revolution and industrial changes, integration and innovation of manufacturing and logistics industries (IIML) has become an inevitable choice for China’s development goal, which is expected to upgrade China’s economic system. As information technology advances rapidly and the servitization market emerges, China has emphasized the integration of two industries (i.e., the manufacturing industry and the logistics industry) to form a new industry through technological penetration, industrial linkage, chain extension, internal restructuring, etc., that is, IIML (Liu et al, 2024). IIML emphasizes on integration and innovation. Integration refers to the formation of strategic cooperation and common development in the whole supply chain with complementary advantages based on specialized resources (such as labor). Innovation refers to continuous improvement of the logistics industry by combining actual needs of the manufacturing industry and innovation trends of the logistics industry (gov.cn, 2020).

However, there are still inadequate empirical studies addressing the impact of IIML on RCE. For RCE, many scholars use the Environmental Kuznets Curve (EKC) to indicate the inverted U-shaped impact of an economic growth pattern on RCE (Zhao et al., 2022; Zhang et al., 2022). At early stages of economic growth, limited economic development leads to increased production and exploitation of natural resources, resulting in pollution, which is the scale effect of economic growth on the environment (Feng et al., 2024). Thus, economies of scale result in a growth in the EKC curve. Further economic growth provides opportunities for investing in capital and technology-intensive industries with a high allocation efficiency (structural effect), improving production technologies and adopting clean technologies (technological effect) (Kaika and Zervas, 2013). Consequently, the structural and technological effects overcome the scale effect, leading to a decrease in the EKC curve. (Zhang et al., 2022). Therefore, as an important feature for industry development and an important manifestation for upgrading the economic system, IIML becomes important on affecting RCE. For IIML, previous scholars have studied IIML from both micro and macro perspectives, with a greater emphasis on micro level studies (Dong et al., 2023; Wang et al., 2023). For example, Dong et al., 2023 explored manufacturing servitization for emission reduction by empirically examining green technological innovation and factor structure improvement using listed company data. From the perspective of IIML at the macro level, Kohtamäki et al., 2020 and Liu et al., 2024 have studied the correlation and economic effects of IIML. However, there is currently no clear measurement method for IIML internationally, and furthermore fewer studies have investigated the RCE effect of IIML at the macro level. It is urgent to investigate how IIML affect RCE based on the industrial integration at the macro level. It is necessary to provide theoretical guidance for the RCE of IIML in the region combining regional characteristics.

The resource dependence theory and institutional theory have provided explanations for IIML as means of RCE (Bendig et al., 2024). Both concern the inter organizational relationship and all emphasize the interaction between the organization and the external environment (Cheng et al., 2024). However, the resource dependence theory emphasizes the importance of external resources, such as complementary resources from other organizations (Shou et al., 2023). Institutional theory highlights the legitimacy of organizations derived from legal systems, cultural systems, and conceptual systems (Feng et al., 2020). That is, the relationships formed between organizations are aimed at obtaining legitimacy. Under the dual carbon goals, based on the resource dependence theory, the manufacturing and logistics industries with specialized division of labor can achieve resource complementarity and allocation through integrated innovation. It is more conducive to adapting to the development needs of the environment (Shuang et al., 2024; Bendig et al., 2024). IIML includes two aspects, namely, (1) the innovation of the manufacturing and logistics industries respectively; and (2) the cooperation between the two industries. Two industries with complementary resources may bring technological innovation (TI) resources and industrial structure improvement (ISI) resources. Although some scholars (Zhang et al., 2022; Zhao et al., 2022) have stated that TI and ISI help RCE reduction, there is no empirical indication whether IIML can achieve RCE reduction through TI and ISI. Furthermore, the organizational legitimacy requirement of institutional theory can be used to supplement the resource dependence theory. Under the requirement of organizational legitimacy, organizations will be influenced by external regulations, such as formal regulations from the government and informal regulations from the public (Shuang et al., 2024). This means that IIML in achieving carbon reduction will be subject to environmental regulations from the government (i.e., government regulation (GR)), as well as rely on public supervision and trust (i.e., social trust (ST)). Therefore, the legitimacy-based resource dependence perspective can provide a complete theoretical framework for the carbon reduction effect of the IIML.

In short, from legitimacy-based resource dependence perspective, there is an urgent need for theoretical guidance on utilizing IIML to achieve RCE reduction. To address this gap, we propose RQs:

RQ1: What is the impact of IIML on RCE?

RQ2: How do TI and ISI mediate the relationship between IIML and RCE?

RQ3: How do GR and ST moderate the relationship between IIML and RCE?

RQ4: From the perspective of provincial heterogeneity, what are the differences in the relationship between IIML and RCE in different regions?

To explore the relationship between IIML and RCE, this paper constructs an empirical model using panel data of 30 provinces in China from 2014 to 2021. Our results show: (1) by measuring IIML level, we find that the current level exhibits a slow upward trend. Further analysis reveals an inverted U-shaped impact of IIML on RCE, and currently the industry is at the stage of promoting RCE with diminishing marginal effect; (2) by analyzing mediating effects, we discover the mediating effects of IIML affecting RCE reduction. At present, the promoting effect of ISI in IIML fails to appear, and IIML mainly reduces the RCE by promoting the MTI and LTI; (3) by analyzing moderating effects, we find that in regions with high energy utilization, GT makes the inverted U-shaped relationship between IIML and RCE flatter. While in areas with high urbanization, ST makes the inverted U-shaped relationship between IIML and RCE flatter; (4) with heterogeneity analysis, we further discover differences among regions from the perspective of economically developed level and industrial structures differences. Specifically, economically developed provinces and provinces with diversified industrial structures are more effective in exerting the inhibiting effect of IIML on RCE.

There are four theoretical contributions in this study. First, from the legitimacy-based resource dependence perspective, this study establishes a theoretical framework for achieving carbon reduction through IIML and provides the empirical evidences. It effectively expands the combination and application of inter organizational relationship theory in carbon reduction research (Shuang et al., 2024; Bendig et al., 2024). Second, this study reveals an inverted U-shape illustrating the relation between IIML and RCE, which supplements studies on realizing RCE reduction by industrial integration (Zhang et al., 2022). Third, from the macro perspective, unlike previous studies that focused on the correlation and economic effects of IIML (Liu et al., 2024), this study calculates IIML and discovers the mechanism by which IIML affects RCE via TI currently. The impact mechanism effectively supplements the research concerning IIML. Fourth, this study uses provincial-level data from a macro perspective, taking regional characteristics such as economy development level and industrial structure differences into the consideration of the impact of IIML on RCE. It is different from investigating single industry and RCE based on micro data, such as manufacturing servitization and logistics services outsourcing (Wang et al., 2023).

The rest of the paper is organized as follows. Section “Literature review” is the literature review. Section “Theoretical framework and hypotheses” establishes the theoretical framework and introduces assumptions. Section “Research design” presents research design and model construction. Section “Empirical results” is the benchmark regression analysis and robustness tests. Section “Further analysis” provides a further analysis of the mediating, moderating, and heterogeneous effects. Section “Supplementary analysis using data of 276 Chinese cities from 2014 to 2021” is a supplementary analysis based on the RCE effect of IIML in Chinese cities. Finally, research conclusions and future research directions are drawn in Section “Conclusions, contributions, and limitations”.

Literature review

Carbon emissions reduction

To achieve the carbon neutral target, previous scholars have studied the carbon reduction from different perspectives, including policy and regulatory promotion (Feng et al., 2024), local economy (Wang et al., 2021), ISI (Zhao et al., 2022), TI (Zhang et al., 2022). For example, industrial structure improvement helps industries move towards the high part of a value chain, thereby reducing resource consumption and promoting carbon reduction. Both Zhang et al. (2022) and Zhao et al. (2022) confirmed that technological upgrading helps promoting carbon reduction. Although previous scholars (Zhang et al., 2022; Wang et al., 2023) have stressed the importance of TI and ISI, they have paid less consideration on promoting carbon reduction through integration among different industries. On the one hand, both manufacturing and logistics industries may have achieved innovation development in soft and hard technologies via improving their technological capabilities and enhancing organizational collaborative models (Zhang et al., 2022). It helps the two industries adapt into the new industrial ecosystem requiring integration and development. On the other hand, the manufacturing and logistics industries may promote deep integration between them through asset investment, as well as improvement of supply chain processes, and enhance the level of strategic cooperation. This helps to achieve rationalization and advancement of industrial structure (Wang et al., 2023). Therefore, we examine the RCE reduction mechanism of IIML, to supplements the research on the RCE reduction effect.

IIML

Previous research on IIML have mainly included macro and micro aspects. From a micro perspective, scholars mainly conducted research on the industrial integration at the micro level in terms of such as manufacturing servitization and logistics outsourcing. Previous scholars analyzed the key influencing factors that affect the IIML. For example, Yuan et al. (2020) used TCT theory and the ERBV theory to conclude that management and transaction uncertainty affect logistics outsourcing by using data from 250 Chinese manufacturing enterprises. Baines et al. (2020) pointed out five main influencing factors from the process of building service-oriented manufacturing such as organizational preparation and organizational commitment. Concerning benefits of manufacturing servitization, more scholars have begun empirical research on the impact mechanism of manufacturing servitization on RCE. These studies have explored the roles of factors such as technological innovation, structural improvement, and value chain participation in this process. Some scholars (Huang et al., 2022; Wang et al., 2023) pointed out the importance of TI economic development, and factor structure improvement. However, by exploring other factors, it was found that manufacturing servitization does not necessarily provide a positive effect on RCE. For example, when logistics companies may take risks in greenwashing, the high investment cost of blockchain technology promotes the RCE (Dong et al., 2023). From a macro perspective, research on IIML is still in its early stage, mostly focusing on the correlation relationship and the economic effects. For example, some scholars (Tang et al., 2018; Liu et al., 2024) calculated the correlation between manufacturing and logistics industries based on the coupling coordination model from a static perspective. However, there was no unified standard for measuring IIML level. In addition, some scholars (Kohtamäki et al., 2020) explored economic effects, which, based on 131 manufacturing enterprises, showed the non-linear U-shaped interaction in industrial integration.

In short, the existing studies (Wang et al., 2023; Dong et al., 2023) have studied IIML from a micro perspective, focusing on the manufacturing industry in one or more countries. However, they have rarely from a macro perspective considered whether IIML will have similar effects on carbon reduction. Therefore, from the macro perspective of Chinese provinces, this paper evaluates IIML from the perspective of dynamic and static perspective, examines the impact of IIML on RCE, and considers regional characteristics into RCE reduction mechanism.

Application of inter organizational relationship theory in RCE

The inter organizational relationship theory refers to the relatively persistent exchange and allocation of resources between organizations. It is often used to study the interactions between a specific enterprise (or industry) and other organizations in business activities (Shuang et al. 2024; Bendig et al., 2024). Common inter organizational relationship theories include resource dependence theory, institutional theory, stakeholder theory, etc. With the increasing attention on RCE, scholars (e.g., Feng et al., 2020) have used the inter organizational relationship theory to explain social behaviors of enterprises/industries in RCE. For example, based on the institutional theory, some scholars (e.g., Feng et al., 2020) have studied the relationship between sustainable practices and market value of luxury goods companies. With the increasing complexity of inter organizational relationships, some scholars (Shou et al., 2023; Cheng et al., 2024) have combined multiple theories to enrich their research. Combining resource dependence theory and stakeholder theory, scholars (e.g., Shou et al., 2023) have studied the moderating effect of external stakeholders on the relationship between green technology innovation and market value. More scholars (e.g., Cheng et al., 2024) have combined resource dependence theory, institutional theory, and stakeholder theory to study the impact of stakeholder concerns on corporate green innovation. Obviously inter organizational relationship theories are widely applied in RCE research. However, based on the resource dependence theory, there have been only a few previous studies on achieving RCE through building resource alliances (such as industrial integration). Meanwhile, following the principle of organizational legitimacy based on institutional theory, under the dual carbon goals, resource alliance activities will be subject to environmental regulations from external sources such as government and the public (Hao et al., 2023; Shuang et al., 2024).

In summary, we should combine the resource dependence theory with the organizational legitimacy of institutional theory. It is worthwhile to study whether the manufacturing and logistics industries can contribute to RCE by forming resource alliances, and how external regulations can regulate this relationship (Cheng et al., 2024; Shuang et al., 2024; Bendig et al., 2024). This study therefore constructs a complete theoretical framework for achieving RCE through IIML, and expands the combined application of institutional theory and resource dependence theory.

Summery of literature summery

By reviewing previous literature, some gaps are identified. First, in terms of RCE, scholars (Zhao et al., 2022; Zhang et al., 2022) have indicated the importance of TI, ISI, and other factors for carbon reduction. However, they have often ignored industrial integration factors, such as IIML. Second, previous studies on the impact of IIML on RCE have mostly focused on, such as manufacturing servitization for empirical exploration in the micro level (Wang et al., 2023). However, there is less focus of IIML in macro level, especially with the concern of carbon reduction effect. Third, IIML is a long-term dynamic development process, and may play a role in RCE reduction (Kohtamäki et al., 2020; Liu et al., 2024). A suitable calculation method of IIML considering both the static and dynamic perspectives and the mechanism of how IIML affecting RCE are worth studying. Fourth, there have been only a few studies on the theoretical framework between resource alliance and carbon emissions from the legitimacy-based resource dependence perspective (Cheng et al., 2024; Shuang et al., 2024; Bendig et al., 2024). Accordingly, this study examines the RCE reduction impact mechanism of IIML, mediating effects of TI and ISI, and moderating effects of GR and ST. In doing so, this study attempts to make managerial contributions with a deep understanding of the relevant issues.

Theoretical framework and hypotheses

From the legitimacy-based resource dependence perspective, this section will explore the relationship between IIML and RCE, explain mediating effects of TI and ISI, and elaborate the moderating effects of GR and ST.

IIML and RCE

Under the dual carbon goal, following the resource dependence theory, the manufacturing industry and the logistics industry complement each other’s resources (Shuang et al.,2024; Bendig et al., 2024). By forming a resource alliance, there exists a relationship between IIML and RCE as follows. When the IIML level is low, enterprises may need to invest more resources to promote integrated innovation, and to provide various basic equipment required for IIML. For example, increasing vehicles and expanding storage facilities may lead to RCE increase (Zhang et al., 2022). Moreover, at the early stages, due to limitations in technological level and industrial structure, energy consumption and RCE in the logistics process may be relatively high (Wang et al., 2024). The continuous expansion of the manufacturing industry has led to a corresponding increase in demand for logistics, which may result in an increase in pollution emissions due to higher outputs and which may lead to an increase in RCE (Feng et al., 2024). With the increase of IIML level, the scale of manufacturing and logistics continues to expand, enterprises are capable of effectively utilizing resources. The EKC curve points out that there is a critical point between environmental pollution and economic development (Zhang et al., 2022). Before the critical point, environmental pollution intensifies with economic development. When adopting effective policies or new technologies, it may result in similar or even faster economic development with lower environmental problem. Industrial integration (e.g., IIML), can form a typical economic development model through resource allocation and complementarity. Therefore, there should be a turning point in the relationship between IIML and RCE. When IIML exceeds the critical point, the scale effect will be more obvious (Ye et al., 2020), and RCE per unit output will gradually decrease. On the one hand, the technology of manufacturing and logistics industries continues to upgrade, and the management of the two industries is also more collaborative. Enterprises are gradually applying more advanced technologies and management methods, which can help with effectively utilization of resources and reduction of energy consumption and RCE (Wang et al., 2023). For the manufacturing industry, enterprises may be more capable of adopting environmentally friendly and efficient production methods to reduce RCE. For the logistics industry, enterprises will also reduce RCE by optimizing transportation routes (Yi et al., 2020), improving warehousing efficiency, and promoting clean energy. On the other hand, the cooperation improvement of the two industries will also help reduce RCE. The two industries have more resources and motivation to promote low-carbon development, such as developing green manufacturing and circular economy (Zhao et al., 2022). For example, Midea Group, a Chinese manufacturing company, and its logistics partner Ande Logistics have formed a green supply chain alliance through asset investment and process optimization. They have achieved full chain carbon footprint accounting and optimization from manufacturing, circulation to after-sales (Chinadaily.com.cn., 2023).

According to the above discussion, at the early stage of IIML, RCE may increase, but with technological progress and industrial improvement, RCE will gradually decrease, ultimately achieving a low carbon level. Based on this, H1 is proposed:

H1: At the early stage of IIML, IIML can promote RCE. When the level of IIML exceeds a critical point, IIML reduces RCE.

The mediating effects of TI and ISI

Following the resource dependence theory, the interdependence between two industries leads to strategic management of these dependencies, resulting in TI and ISI as a response to the strategic resource management (Zhang et al., 2022; Zhao et al., 2022; Bendig et al., 2024). IIML is a typical performance of industrial integration development. It has two basic characteristics: the innovation of the two industries, and the integration between the two industries. Therefore, we mainly explain the inherent mechanism hypothesis between IIML and RCE from two aspects: TI and ISI.

The mediating effect of TI

With the widespread application of digital technologies, there is a mandatory shift underway in production systems, transitioning from high emissions to low-carbon operations (Zhang et al., 2022; Liu et al., 2024). By improving production efficiency, energy utilization can be improved, and RCE can be reduced.

On the one hand, IIML improves the TI in manufacturing and logistics industries. For example, the manufacturing industry may invest efficient production technologies and equipment, and the logistics industry may introduce more energy-efficient transportation and warehousing facilities (Wang et al., 2021). On the other hand, combining with the inverted U-shaped impact of IIML on RCE, there are two stages, in which TI plays a role in RCE along with the improvement of IIML. IIML promotes TI in manufacturing and logistics industries. However, at the early stage of IIML, because the research and application of new technologies often require a large amount of energy and resources, which leads to limited TI level in the early stages of IIML. Moreover, at this stage, as indicated by previous scholars (Yang et al., 2017), the scale and effectiveness of TI are relatively low, the absorptive capacity of TI is insufficient, and the utilization of TI is insufficient, thereby TI is incapable of effectively reducing RCE. With the development of IIML, the technological level still improves continuously. Differently, TI starts to exert its carbon reduction effect (Yi et al., 2020). In this process, production and logistics efficiency can be improved through continuous TI, thus reducing energy consumption and waste, and reducing RCE (Zhao et al., 2022). Moreover, the technology effect of IIML at this stage also helps to optimize and integrate the supply chain. It not only help both parties with better logistics planning and reduce ineffective transportation, but also promote information sharing, improving the efficiency of collaboration, and further reducing RCE (Liu et al., 2020; Ye et al., 2020).

In summary, IIML promotes TI, optimizes the operation of the enterprise itself and the collaborative operation of the supply chain, and effectively suppresses RCE. Based on this, H2a and H2b are proposed:

H2a: Technological innovation in the manufacturing industry (MTI) is an important channel for IIML to act on RCE reduction. At the early stage of IIML, IIML can promote RCE through MTI. When the level of IIML exceeds a critical point, IIML reduces RCE through MTI.

H2b: Technological innovation in the logistics industry (LTI) is an important channel for IIML to act on RCE reduction. At the early stage of IIML, IIML can promote RCE through LTI. When the level of IIML exceeds a critical point, IIML reduces RCE through LTI.

The mediating effect of ISI

It is generally believed that ISI consists of two aspects. The first aspect is the advancement of industrial structure (AIS), which is the process of industrial structure transitioning from lower to a higher level. In the process, the contribution of the primary and secondary industries to GDP has relatively decreased, while the contribution of the tertiary industry to GDP has gradually increased, becoming the main driving force for economic growth (Zhu et al., 2019). For example, many traditional manufacturing enterprises shift from labor-intensive structure with a low allocation efficiency to capital and technology-intensive structure with a high allocation efficiency (Zhu et al., 2019). The first aspect is the rationalization of industrial structure (RIS), which refers to the process of increasing the correlation between industrial structures. Scholars often used the Theil index to measure RIS (Zhao et al., 2022). Therefore, we analyze the mediating role of IIML in RCE reduction from the perspectives of AIS and RIS.

First, IIML improves AIS. Meanwhile, combining with the inverted U-shaped impact of IIML on RCE, there are two stages along with the development of IIML in which AIS plays a role in RCE. IIML helps to adjust industrial structure and bring AIS. Specifically, it helps to optimize the supply chain process, promote the transformation of manufacturing servitization, and contribute to the advancement of the logistics industry. Thereby, it increases the proportion of the tertiary industry (Kohtamäki et al., 2020; Wang et al., 2023). In the early stages of IIML, because the industrial structure is still dominated by the primary and secondary industries, the overall energy consumption was relatively high, leading to a gradual increase in RCE level. With the development of IIML, AIS still improves continuously. Compared to the secondary industry, the tertiary industry produces less energy consumption. When the proportion of the tertiary industry with low energy consumption increases significantly, the overall consumption of fossil energy in the industrial chain decreases. It will further reduce RCE (Feng et al., 2024). Therefore, IIML promotes AIS, which in turn decrease RCE.

Second, IIML improves RIS. Similarly, with the improvement of IIML, there are two stages in which RIS plays a role in RCE. IIML helps to improve the coordination and cooperation among the three major industries and promote the RIS (Tian and Pang, 2023). Specifically, IIML emphasizes the collaborative transformation of the organization. It can enhance coordination among industries and the RIS (Tian and Pang, 2023). In the early stages of IIML, due to the low RIS and uneven competition in resource factors among industries, industrial efficiency is low and RCE increases (Zhao et al., 2022). With the development of IIML, RIS still improves continuously. In this situation, resource allocation can be improved, which further reduces energy consumption per unit of GDP and thus reduces RCE (Feng et al., 2024). Therefore, IIML promotes carbon reduction by improving RIS.

In summary, IIML promotes AIS and RIS, which helps to adjust and optimize the industrial structure, thereby effectively reducing RCE. We then propose the following hypotheses:

H3a: AIS is an important channel for IIML to act on RCE reduction. At the early stage of IIML, IIML can promote RCE through promoting AIS. When the level of IIML exceeds a critical point, IIML reduces RCE through promoting AIS.

H3b: RIS is an important channel for IIML to act on RCE reduction. At the early stage of IIML, IIML can promote RCE through promoting RIS. When the level of IIML exceeds a critical point, IIML reduces RCE through promoting RIS.

The moderating effects of GR and ST

The legitimacy-based resource dependence perspective states that in order to obtain organizational legitimacy, the resource alliance formed needs to consider its external system (Cheng et al., 2024; Shuang et al., 2024). External regulation includes formal regulation (such as laws, regulations, policy documents from the government) and informal regulation, such as social awareness and codes of conduct from the public (Tolmie et al., 2019; Chen and Wan, 2020). In exploring the carbon reduction effect of IIML, the key external regulations are GR and ST. GR refers to the regulatory behavior of government in issuing policy documents related to low-carbon, which can be expressed by the frequency of attention to RCE in local government documents (Tolmie et al., 2019; Cheng et al., 2024). ST refers to public trust in the business environment and commercial credit of a region, which can measure the quality of the commercial credit environment in the region. This section explains how GR (formal regulation) and ST (informal regulation) affect the relationship between IIML and RCE (Chen and Wan, 2020; Shuang et al., 2024).

The moderating effect of GR

Following the legitimacy-based resource dependence perspective, GR is an important moderating factor in the relationship between IIML and RCE. To achieve organizational legitimacy, GR, as an external formal regulation, regulates the social behavior of industries through the formulation of laws, policies, and standards (Li et al., 2023). Under the dual carbon goal, GR mostly refers to the government issuing policy documents related to pollutant emission restrictions and resource utilization in the production process (Berrone et al., 2013). It forces various industries to consider carbon reduction as a necessary strategy from a long-term aspect (Tolmie et al., 2019). Under GR, to achieve organizational legitimacy, the manufacturing and logistics industries can accelerate IIML, complement and allocate resources between them, and jointly promote RCE strategies.

Government regulations of low-carbon behavior make IIML more inclined to flatly realize the RCE reduction (Hao et al., 2023). Based on the inverted U-shaped relationship between IIML and RCE while under the same circumstances, the manufacturing and logistics industries tend to invest human, material, and financial resources in carbon reduction activities earlier (Shou et al., 2023). When IIML is low, GR plays a positive role by setting environmental standards and implementing tax policies, making resource investment in both industries rational and resulting in a relatively flatter increase for RCE (Cheng et al., 2024). As IIML continues to increase, compared to GR, the resource utilization ability of both industries is more important. Only when a sufficient number of enterprises achieve low-carbon development can their RCE be reduced (Shuang et al., 2024). Therefore, under the same circumstances, compared to the absence of GR, the overall RCE reduction of the industry has become more flatter with the moderation of GR. Thus, the presence of GR makes the inverted U-shaped curve flatter.

In summary, under the influence of GR, the initial increase in RCE may be relatively flatter, and as IIML increases, the decrease in RCE will also be relatively flatter. Based on this, the following hypotheses can be proposed:

H4: GR adjusts the inverted U-shaped relationship between IIML and RCE, making the inverted U-shaped curve flatter.

The moderating effect of ST

Following the legitimacy-based resource dependence perspective, another important factor affecting the relationship between IIML and RCE is ST (Chen and Wan, 2020). In order to achieve organizational legitimacy, ST is an external informal regulation that represents the intangible constraints of the public on social behavior (Dong et al., 2018). Under the dual carbon goal, ST represents the public’s supervision of low-carbon market behavior, which enforces companies to take proactive environmental actions and regard carbon reduction as an important focus (Shuang et al., 2024). In other words, subject to public supervision, the manufacturing and logistics industries will also accelerate the resource complementarity and resource allocation effects of IIML, promoting the achievement of RCE.

The public’s trust in industry social behavior has enhanced the action power of the manufacturing and logistics industries to reduce carbon emissions (Liao et al., 2018). From the perspective of the inverted U-shaped relationship between IIML and RCE, when IIML is low, ST prompts logistics and manufacturing companies to take proactive actions to reduce RCE earlier. Therefore, under the same circumstances, the existence of ST has strengthened the confidence in RCE reduction in both industries, and the degree of increase in RCE is relatively flatter (Mubarak et al., 2021), compared to the absence of ST. With the continuous enhancement of IIML, compared to ST, the resource allocation of logistics and manufacturing industries more important in the RCE reduction process. When the resource alliance formed by manufacturing and logistics industries is more capable of achieving low-carbon operations, carbon emissions will gradually decrease (Su and Tsang, 2015). This means that the presence of ST also makes the inverted U-shaped curve flatter.

In summary, under the influence of ST, the initial increase in RCE may be relatively flatter, and as IIML increases, the decrease in RCE will also be relatively flatter. Based on this, the following hypotheses can be proposed:

H5: ST adjusts the inverted U-shaped relationship between IIML and RCE, making the inverted U-shaped curve flatter.

Our research model about the mechanisms of RCE reduction effect of IIML can be shown in Fig. 1, illustrating H1-H5 mentioned above.

Fig. 1

The mechanisms of RCE reduction effect of IIML.

Research design

Sample selection

Given the principle of data availability, reliability and validity, we set the research sample as 30 provinces and cities in China (excluding Hong Kong, Macao Special Administrative Region, Taiwan and Xizang Autonomous Region as their data was missing) with balance panel data from 2014 to 2021. The data sources mainly include (1) the total carbon emission data are sourced from CEADS and China Energy Statistical Yearbook; (2) the data related to indicators for IIML, TI, ISI, and control variables is collected from the 2014–2021 China Statistical Yearbooks, China Population and Employment Statistical Yearbooks, provincial statistical yearbooks, and statistical bulletins; (3) referring to Shuang et al., (2024), GR data is obtained by using Python to segment the reports of various provincial governments and calculate the frequency of keywords related to environmental regulation; (4) ST is retrieved from the Blue Book of China Urban Commercial Credit Environment Index (CEI) (Yang and Niu, 2023). Due to limitations in data collection (e.g., carbon emission data on CEADS is only available up to 2021), this study included the data from 2014 to 2021. This treatment is however common in the existing studies (Richard and Ram, 2024; Liu et al., 2024; Yang et al., 2024). For example, Richard and Ram (2024) examined the relationship between vertically integrated stakeholder risk exposure and environmental performance in 2023, using a dataset from 2002 to 2012. Moreover, scholars have hardly touched upon the latest 2023 or 2024 data when studying the inverted U-shaped relationship of carbon emissions (Liu et al., 2024; Yang et al., 2024). Therefore, as Richard and Ram (2024) indicated, when there is no structural change in the research problem in the short term, data that differs by one or two years will have a relatively small impact on the original conclusion. This further proves the rigor of the data selection scope in this study.

Variable selection and definition

Independent variables

IIML and (IIML)²: In the past, there was no unified standard for measuring IIML level. Thus, following the approach of Liu et al. (2024), the level of coupling coordination between the manufacturing and logistics industries is used to represent IIML level. We use (IIML)² to measure the quadratic term of IIML. IIML and (IIML)² serve as explanatory variables. The process includes the following three steps:

Step 1: Selection of indicators

This paper extracts four characteristics (asset investment, process improvement, technological upgrading, and organizational collaboration) of IIML according to the policy plan “Implementation Plan for Promoting the Deep Integration and Innovation Development of Logistics Industry Manufacturing Industry”, and uses them as primary indicators (Liu et al., 2024). The indicator system with two subsystems is shown in Table 1.

Table 1 Selection of indicators.

Step 2: Evaluation method

IIML is a long-term dynamic development process. We should consider both the relative importance between indicators (i.e., static perspective) and the dynamic changes of indicators over time (i.e., dynamic perspective). However, the commonly used evaluation method entropy weight method only considers the relative importance between indicators from a static perspective. Therefore, when evaluating IIML level, a dynamic evaluation method needs to be developed to include cross-period comparison, in addition to the relative importance between indicators based on the entropy weights (Liu et al., 2024).

Therefore, we combine the vertical and horizontal levelling method in dynamic evaluation and the entropy weights in static evaluation to comprehensively evaluate IIML level. The vertical and horizontal levelling method is used for cross-period comparison of time-series three-dimensional data tables (i.e., panel data), and weighting information comes from the differences between the evaluated objects (Zhang et al., 2022). The entropy weight method uses entropy value to show the influence of an indicator on a comprehensive evaluation (Liu et al., 2024). Alternatively speaking, the vertical and horizontal levelling method considers the differences between evaluation objects, effectively calculating panel data dynamically, while the entropy weight method considers the relative importance of indicators.

The process is as follows. First, referring to Liu et al. (2024), we use the entropy weight method to weigh the indicators considering the relative importance. Second, referring to Zhang (2022), the vertical and horizontal grading method is adopted to assign weights to the indicators considering the differences between the evaluation objects. Finally, the weights are obtained by combining the weights of the two indicators based on the Eq. (1); the comprehensive development level of the two subsystems are obtained based on the Eq. (2) (Feng et al., 2018).

$$\omega =\frac{{\omega}_{1}{\omega}_{2}}{\mathop{\sum}\nolimits_{j=1}^{n}{\omega}_{1}{\omega}_{2}}$$

(1)

$${u}_{i}=\mathop{\sum }\limits_{j=1}^{m}{\omega }_{{ij}}{u}_{{ij}}\,,{\rm{i}}.{\rm{e}}.,\,\mathop{\sum }\limits_{j=1}^{m}{\omega }_{{ij}}=1,\,i=m,{l;j}=1,\ldots ,8$$

(2)

${u}_{i}(i=m,l)$ represents the comprehensive development level of the subsystem i, ${u}_{{ij}}$ $(i=m,{l;j}=1,\ldots ,8)$ is the standard value of the jth indicator in system $i$, ${\omega }_{{ij}}$ is the weight of each indicator.

Step 3: Evaluate IIML level

We adopt the coupling coordination degree to measure IIML level, which is the most widely used for evaluating the correlation and fusion degree between two systems, in which coupling is a phenomenon originating from the physical sciences (Tang et al., 2018; Liu et al., 2024). It has been extensively applied to assess the overall development status of distinct systems. The coupling level (C) and development level (T) of the two subsystems are calculated, respectively in Eq. (3) and Eq. (4), which lead to IIML level expression in Eq. (5).

$${C}_{{ml}}^{t}=2\frac{\sqrt{{u}_{m}^{t}\times {u}_{l}^{t}}}{{u}_{m}^{t}+{u}_{l}^{t}}$$

(3)

$${T}_{{ml}}^{t}=\alpha {u}_{m}^{t}+\beta {u}_{l}^{t}$$

(4)

$${{IIML}}_{{ml}}^{t}=\sqrt{{C}_{{ml}}^{t}\times {T}_{{ml}}^{t}}$$

(5)

Where ${C}_{{ml}}^{t}$ is the coupling degree value, ${T}_{{ml}}^{t}$ is the coordination degree value, $\alpha$ and $\beta$ represent the contributions of the two industries to the overall system. Referring to Liu et al. (2024), considering that the importance degrees of the two industries are same in the process of IIML, we assign value $\alpha =0.5$, $\beta =0.5$. ${{IIML}}_{{ml}}^{t}$ is the coupling and coordination of two industries in year $t(t=1,\ldots ,8)$.

Dependent variable

Regional Carbon Emission (RCE): it is the logarithmic value of the ratio of carbon dioxide emissions in a province to regional GDP (Zhou et al., 2019). Because in some provinces with high economic development levels, there could be higher total carbon emissions and IIML levels. Compared to total carbon emissions, the carbon emission intensity considers the GDP level of the region and is more suitable for measuring carbon emissions (Zhou et al., 2019). For example, China often uses carbon emissions per unit of GDP as an important indicator to measure low-carbon development, expecting that by the year 2030, China aims to reduce its carbon emissions per unit of GDP by over 65% (gov.cn, 2021). Therefore, this study use the carbon emission intensity to measure RCE. Total amount of carbon emission is calculated using CEADS’s carbon dioxide accounting data (Zhao et al., 2022; Guo and Pang, 2023).

Mediating variables

①TI Level (MTI and LTI): this variable includes two parts, namely Manufacturing Technology Innovation Level (MTI) and Logistics Technology Innovation Level (LTI). Patent applications, measured by the ratio of patent applications in the manufacturing and logistics industries to the number of employees, have a good reflection on the TI level associated with IIML (Wurlod and Noailly, 2018). Moreover, compared to patent authorizations, the patent applications eliminate time lag and clearly reflect the TI of the year. Therefore, patent applications in the manufacturing and logistics industries are selected to measure the TI level of the province.

② ISI (AIS and RIS): ISI consists of two aspects, namely AIS and RIS. Referring to Zhou et al. (2019), AIS is measured by the ratio of the tertiary industry to the secondary industry, as shown in formula (6). RIS is measured by the Theil index, the smaller the Theil index, the more reasonable the industrial structure (Zhao et al., 2022), as shown in formula (7).

$${AIS}=\frac{{Y}_{3}}{{Y}_{2}}$$

(6)

$${RIS}=\mathop{\sum }\limits_{i=1}^{3}\left(\frac{{Y}_{i}}{Y}\right){\mathrm{ln}}\left(\frac{{Y}_{i}}{{L}_{i}}/\frac{Y}{L}\right)=\mathop{\sum}\limits_{i=1}^{3}\left(\frac{{Y}_{i}}{Y}\right){\mathrm{ln}}\left(\frac{{Y}_{i}}{Y}/\frac{{L}_{i}}{L}\right),{i}={1,2,3}$$

(7)

Where, $Y$ represents the output value. $L$ represent the employment of the three major industries. $i$ is the industry index, representing three major industries, i.e., i = 1 corresponds to the first industry (i.e., agriculture industry), i = 2 corresponds to the secondary industry (i.e., industrial industry), and i = 3 corresponds to the tertiary industry (i.e., service industry).

Equations (6) and (7) show the calculation of AIS and RIS. It is worth noting that Eq. (7) of RIS is the result of using the Theil index, and a large Theil index indicates a low RIS. Therefore, the larger AIS is, the higher value of the industrial structure is; the larger RIS is, the more unreasonable the industrial structure is.

4.2.4 Moderating variables

① Government regulation (GR): it is measured by summing the frequency proportion of environmental regulation related words with respect to the total text frequency. As in Shuang et al. (2024), GR is measured by using Python to segment the reports of various provincial governments. The keywords include environmental protection, environmental protection, pollution, energy consumption, emission reduction, pollution discharge, ecology, green, low-carbon, air, chemical oxygen demand, sulfur dioxide, carbon dioxide, PM10, PM2.5, etc. The total frequency of keywords related to environmental regulation is calculated, and the proportion of these keywords in the total frequency of the government report is calculated.

② Social trust (ST): it is measured by the logarithmic value of the average commercial credit environment index of each city in the province. Referring to Yang et al. (2024), the China Information Center Business Institute releases the China Municipal Commercial Credit Environment Index. By measuring the credit environment of local enterprises in commercial activities, the index reflects the ST for local enterprises’ business operations. We summarize the commercial credit environment index of each city in the province, and calculate the mean as the social trust level of the province.

4.2.5 Control variables

There are various influential factors on RCE. Based on a summary of previous studies, the following control variables have been selected: ①Energy intensity (ENE): energy consumption per unit of GDP. Referring to Zhang (2022), a high energy intensity often means more fossil fuels such as coal, leading to an increase in RCE. ②Urbanization level (UR): the proportion of the urban population to the total population. When introducing UR as the control variable, previous scholars (Zheng et al., 2021) often introduced its quadratic term, because the impact of IIML on RCE often conforms to the EKC curve, which means there is an inverted U-shaped impact. Thus, introducing a quadratic term can better control the impact of urban population on its results. Similarly, we introduce the quadratic term ((UR)²) of urbanization level to our control variables. ③Government Support (GOV): the ratio of local general public expenditures to GDP. Zhang et al. (2022) pointed out that government fiscal expenditure would affect RCE. ④Foreign Investment Ratio (FDI): the logarithm of actual amount of foreign investment is utilized. Economic globalization is constantly advancing, and research suggests that increasing economic connections between different countries may affect RCE (Zhang et al., 2022). ⑤Population Density (PS): this is the logarithmic value of year-end population. Referring to Zhang et al. (2022), the continuous growth of population can lead to excessive exploitation of resources and excessive emissions of pollutants, which increase RCE.

4.3 Model construction

This study uses RCE as the dependent variable. According to H1 above, there is an inverted U-shaped impact of IIML on RCE. Therefore, IIML level and its quadratic term are used as explanatory variables. Model 1 is set as follows:

$${{RCE}}_{{mn}}={\beta}_{0}+{\beta}_{1}{{IIML}}_{{mn}}+{\beta}_{2}{({IIML}_{mn})}^{2}+{\beta }_{3}{X}_{{mn}}+{\delta }_{m}+{\gamma }_{m}+{\varepsilon }_{{mn}}$$

(8)

Jiang (2022) indicated that the three-stage stepwise regression method may increase the standard error of coefficient estimation due to the correlation between independent variables and mediating variables. When testing for mediation effects, the effectiveness of statistical testing is relatively low. When the relationship between the selected mediating variable and the dependent variable is obvious, the two-stage regression method is preferred for testing the mediation effect. Therefore, we examine the impact mechanism of IIML on RCE by the two-step regression method. Meanwhile, we refer to Lin and Li (2022) and use non-parametric percentile Bootstrap method to verify the mediating effect. It ensures a rigor mediation effect test and further indicates the existence and significance level of indirect effects. In summary, based on the hypotheses in section “The mediating effects of TI and ISI”, we analyze whether there exists a mediating mechanism for carbon reduction through TI and ISI in the process of IIML. Model 2 is set as follows:

$${{MTI}}_{{mn}}={\beta }_{2}+{\beta }_{1}{{IIML}}_{{mn}}+{\beta }_{2}{X}_{{mn}}+{\delta }_{m}+{\gamma }_{m}+{\varepsilon }_{{mn}}$$

(9)

$${{LTI}}_{{mn}}={\beta }_{2}+{\beta }_{1}{{IIML}}_{{mn}}+{\beta }_{2}{X}_{{mn}}+{\delta }_{m}+{\gamma }_{m}+{\varepsilon }_{{mn}}$$

(10)

$${{AIS}}_{{mn}}={\beta }_{2}+{\beta }_{1}{{IIML}}_{{mn}}+{\beta }_{2}{X}_{{mn}}+{\delta }_{m}+{\gamma }_{m}+{\varepsilon }_{{mn}}$$

(11)

$${{RIS}}_{{mn}}={\beta }_{2}+{\beta }_{1}{{IIML}}_{{mn}}+{\beta }_{2}{X}_{{mn}}+{\delta }_{m}+{\gamma }_{m}+{\varepsilon }_{{mn}}$$

(12)

Moreover, based on the hypotheses in section “The moderating effects of GR and ST”, we analyze whether there exists a moderating mechanism for carbon reduction through GR and ST in the process of IIML. Referring to Shuang et al. (2024), Model 3 is set as follows:

$$\begin{array}{ll}{{RCE}}_{{mn}}={\beta }_{0}+{\beta }_{1}{{IIML}}_{{mn}}+{\beta }_{2}({{IIML}_{{mn}}})^{2}+{\beta }_{3}{{IIML}}_{{mn}}\times {{GR}}_{{mn}}\\\qquad\qquad\qquad+{\beta }_{4}{({{IIML}}_{{mn}})}^{2}\times {{GR}}_{{mn}}+{\beta }_{5}{X}_{{mn}}+{\delta }_{m}+{\gamma }_{m}+{\varepsilon }_{{mn}}\end{array}$$

(13)

$$\begin{array}{ll}{{RCE}}_{{mn}}={\beta }_{0}+{\beta }_{1}{{IIML}}_{{mn}}+{\beta }_{2}{({{IIML}}_{{mn}})}^{2}+{\beta }_{3}{{IIML}}_{{mn}}\times {{ST}}_{{mn}}\\\qquad\qquad\qquad+{\beta }_{4}{({{IIML}}_{{mn}})}^{2}\times {{ST}}_{{mn}}+{\beta }_{5}{X}_{{mn}}+{\delta }_{m}+{\gamma }_{m}+{\varepsilon }_{{mn}}\end{array}$$

(14)

In Eqs. (8)–(14), subscript $m$ represents province, subscript $n$ represents the year, ${{RCE}}_{{mn}}$ represents RCE of province (city) $m$ in year $n$, ${{IIML}}_{{mn}}$ represents the coupling coordination level of two industries in province (city) $m$ in year $n$. ${X}_{{mn}}$ are control variables in the model, ${\delta }_{m}$ is the fixed effect of the province, and ${\gamma }_{m}$ is a fixed year effect, ${\varepsilon }_{{mn}}$ is a random error term. ${{MTI}}_{{mn}},{{LTI}}_{{mn}},{{AIS}}_{{mn}}$ and ${{RIS}}_{{mn}}$ represents mediating variables in province (city) $m$ in year $n$. ${{GR}}_{{mn}}$ and ${{ST}}_{{mn}}$ represents moderating variables in province (city) $m$ in year $n$. In addition, to eliminate the influence of extreme outliers, this article performed tail reduction on continuous variables. We also centralized the interaction term to avoid multicollinearity caused by the interaction term between the moderating variable (GR and ST) and the explanatory variable.

Empirical results

Benchmark analysis

Overall description of IIML

According to Fig. 2, IIML level has increased from 0.437 in 2014 to 0.480 in 2021. Although the overall IIML level is not high, the interaction between the two industries is in a relatively slow but with a stable upward trend.

Fig. 2

IIML level.

Variable descriptive analysis

In this section, we focus on statistical analysis of variables and correlation analysis between variables.

First, statistical analysis was conducted on the variables, including mean, standard deviation, minimum and maximum values. The results are shown in Table 2 below.

Table 2 Variable descriptive analysis results.

The results show an average level of IIML in each province 0.4254, with a standard deviation of 0.1397, and the minimum and maximum values 0.1868 and 0.8166, respectively. The average RCE of each province is 0.0196, with a standard deviation of 0.0213, and the minimum and maximum values are 0.0017 and 0.1216, respectively. Therefore, there are certain differences in IIML and RCE among provinces. It is worth noting that AIS can reach 5.2968, with its 1.3709, and the lowest value only 0.6661, indicating significant differences of AIS among many provinces, with the overall AIS relatively low.

Second, the correlation analysis was conducted on the variables selected in 4.2, and Table 3 shows the results.

Table 3 Correlation analysis of coefficients.

The results of Table 3 are explained as follows. First, except for the correlation coefficient between the variable and its quadratic term, the correlation coefficients between other variables do not exceed 0.8. This indicates insignificant correlation and no multi-collinearity between these variables. Second, by analyzing the correlation coefficients of some core variables, H1 can be preliminarily verified. The correlation coefficient between RCE and (IIML)² shows a significant result, with a value of −0.28. This to some extent indicates a possible inverted U-shaped impact of IIML on RCE. H1 will be further supported through regression models. Similarly, the correlation coefficients between MTI, LTI, and (IIML)² are 0.44 and 0.68, respectively, and both are significant. To some extent, IIML may have a critical point when generating carbon reduction effects through TI, namely H2a and H2b are supported. Again, the hypothesis will be further supported through regression models. Furthermore, the correlation coefficients between AIS, RIS, and (IIML)² indicate a significant negative correlation with (IIML)² with a value of −0.43. This is inconsistent with H3a and H3b, and further empirical analysis is needed to explain the reasons. Analyzing the correlation coefficients between GR, ST, (IIML)² and RCE, we found that their correlations are relatively small and suitable as moderating variables. It is worth mentioning that although ST is slightly correlated with some control variables, the VIF value was tested and found to be 3.56 < 5, indicating no multi-collinearity.

Benchmark regression analysis

According to model 1 in section “Model construction”, regression was performed on the entire sample data. This section examines the impact of IIML on RCE.

First, referring to previous study (Xiang, 2023), the Hausman test was used to determine the choice of a fixed-effects model or a random-effects model. The results in Table 4 show a p-value of 0.0018, therefore suggesting a fixed effects model. Therefore, we adopted bidirectional fixed effects to improve the robustness of regression results.

Table 4 Hausman test.

Second, by regressing the first and second terms separately, we can obtain the results. Columns (1) and (2) in Table 5 respectively represent the estimated results of the impact of the first and second terms of IIML on RCE. Specifically, in column (1), the coefficient of IIML is 0.0224(1% confidence level), indicating that the current IIML level has not effectively curbed RCE. In Column (2), the coefficient of IIML is 0.0640(1% confidence level), and the second-order coefficient is 0.0464(the 5% confidence level). The results indicate that although the current IIML level has not effectively curbed RCE, there is an inverted U-shaped impact of IIML on RCE. In other words, there is a threshold for the effect of IIML on RCE. When IIML in a province is lower than this threshold, the enhancement of IIML promotes RCE. When IIML in a province exceeds this threshold, the inhibitory effect of IIML on RCE dominates.

Table 5 Benchmark Regression on the impact of IIML on RCE.

In the inverted U-shape curve, the extremum turning point of IIML level was calculated as 0.69. To statistically confirm the existence of turning point, we further conducted a Wald test (Greene, 2019), which is a test to determine whether both the first-order and second-order coefficients satisfy specific conditions. We assumed that the null hypothesis (H0: coefficient of the first term is less than or equal to zero, and coefficient of the second term is greater than or equal to zero) holds. The Wald test explained whether the coefficients of IIML and IIML² are significant and illustrated the inverted U-shaped relationship by combining the p-value significance of one-sided test. The results show that the Wald test statistic of IIML is 11.13, p = 0.0010 < 0.05, and the p-value of its unilateral test is 0.0005 < 0.05. The Wald test statistic for IIML² is 5.37, p = 0.0215 < 0.05, and the p-value of its unilateral test is 0.0108 < 0.05. Hence, it can be concluded that the results reject the null hypothesis, which states that the coefficient of the first-order term is positive, and the coefficient of the second-order term is negative; and that there is an inverted U-shaped relationship between IIML and RCE. Furthermore, referring to Hanns et al. (2016), we calculated that in the inverted U-shaped curves of IIML and RCE, the extremum turning point of 0.69 is within the data range of the independent variable. The lower bound is 0.1953 and the slope of the lower bound is 0.0459. The upper bound is 0.79 and the slope of the Upper bound is −0.0094. Moreover, we find that the turning extremum point (i.e. 0.69) is located at the 95% Fieller interval ([0.5528, 1.1158]). In summary, we believe that there exists an inverted U-shaped relationship between IIML and RCE. Thus, H1 is supported. To clearly demonstrate the existence of an inverted U-shaped relationship, we fit the relationship between RCE estimates and IIML using a regression model, as shown in Fig. 3.

Fig. 3

The relationship between RCE estimates and IIML using a regression model.

Therefore, on the one hand, we can conclude that IIML is still at its early stages, as the lower bound (i.e. 0.1953) < 0.69 < the upper bound (i.e. 0.79) at 95% Fieller interval ([0.5528, 1.1158]). Based on Section “Overall description of IIML”, we can also see that the average IIML has not yet crossed the extreme point in 2021. This again indicates that the IIML in various provinces has not completely crossed the extreme point, and IIML is still mainly in the early stage. Manufacturing industry is still expanding its scale, and the demand for logistics is high. Frequent energy consumption leads to RCE increase, which promotes RCE. On the other hand, the inverted U-shaped relationship indicates that the marginal promoting effect of IIML on the total RCE is continuously decreasing. Thus, as IIML increases, IIML in each province shows an inverted U-shaped effect on RCE.

Robustness test

Measurement method for replacing the dependent variable

This section performs robustness test by changing the dependent variable. In the benchmark model, we referred to the RCE data from the CEADS database to calculate the RCE, and to measure the low-carbon level. However, there are currently several methods for measuring RCE, without a unified accounting standard for carbon dioxide emission indicators. Previously, some scholars (Tol, 2007; Zhang et al., 2022; Xiang et al., 2023) used the apparent emission accounting method to calculate RCE level and further calculate RCE using IPCC RCE data. To support the robustness of the results, this section replaces the existing dependent variable with IPCC RCE data for a robustness test, as shown in column (1) of Table 6. The result shows that there is still an inverted U-shaped impact of IIML on replaced RCE. The robustness of the results in this article is supported.

Table 6 Robustness test results.

Excluding sample data from major energy provinces (autonomous regions)

RCE is not only influenced by the industrial structure, but also by the energy structure in the regions, which may affect the relationship between RCE and IIML (Xiang et al., 2023). In order to further confirm the robustness of study results, we adapted the approach in Guo and Zhang (2023), excluded the sample data of six major provinces/autonomous regions with high energy intensity (Shanxi, Shaanxi, Inner Mongolia, Gansu, Ningxia, and Xinjiang), and reconducted the regression analysis. From the column (2) of Table 6, we can find that the results support the robustness of the results presented in Section “Benchmark analysis”. It shows that for the sample data excluding six major provinces/autonomous regions with high energy intensity, there is still the phenomenon that at the early stage of IIML, IIML can promote RCE. When the level of IIML exceeds a critical point, IIML reduces RCE. The robustness of the results in this article is supported.

Endogeneity test

Scholars (e.g., Diebel et al., 2024) stated that using fixed effects can to some extent eliminates the interference of disturbance terms. Along with the use of the Hausman test and fixed effects in the benchmark analysis, endogeneity has been partially eliminated in this study. In addition, endogeneity issues such as omitted variables, measurement errors, and reverse causality may also lead to biased estimation of coefficients. Therefore, we conducted endogeneity tests in this section, including independent variable lag for endogeneity test and instrumental variable method for endogeneity test.

Lagged one-period test of independent variable

There may be a reverse causality problem between IIML and RCE. When IIML promotes RCE, changes in RCE may in turn affect IIML. Referring to Diebel et al. (2024), we lagged the independent variable IIML and (IIML)² by one period, denoted as IIML_L and (IIML) ²_L, respectively, to address the issue of reverse causality. The endogeneity results are shown in Table 7.

Table 7 The independent variable lag for endogeneity test.

The results show that even if the independent variable lags one period, an inverted U-shaped relationship still exists between IIML and RCE. The endogeneity issue has been resolved.

Instrumental variable method for endogeneity test

Referring to Bendig et al. (2024), we used the 2SLS instrumental variable method to test bidirectional causality problems. Another effective way to address endogeneity issues is to find an exogenous explanatory variable that does not directly affect the dependent variable. Referring to Xu et al. (2024), we planned to select a policy shock that is highly correlated with the explanatory variable IIML. Although the most relevant policy to IIML is the two-industry integration policy issued in 2020, considering that the time span of this study is from 2014 to 2021, the two-industry integration policy as a policy shock is difficult to form an experimental group and a control group with good comparative effects. Therefore, we have selected a policy with similar functions as the two-industry integration policy, namely the “Pilot Plan for Innovative Development of Trade in Services” (hereinafter referred to as the “Trade in Services Innovation Policy”) issued by the State Council of China in February 2016. This policy agreed to carry out pilot projects for innovative development of trade in services in 13 provinces (cities) including Tianjin, Heilongjiang, Shanghai, Jiangsu, Zhejiang, Shandong, Hubei, Guangdong, Hainan, Chongqing, Sichuan, Guizhou, and Shaanxi. The policy content pointed out in multiple places about the innovative development of the service industry and the integrated development of manufacturing and service industries. As a typical industry of traditional service trade, the logistics industry is closely related to IIML, but it has good “exogeneity” compared to RCE, meeting the requirements of instrumental variables.

Based on this, this section uses the Service Trade Innovation Policy issued in February 2016 to identify the impact of IIML on RCE. Referring to Xu et al. (2024), the impact of service trade innovation policies on IIML was first studied through a difference-in-differences model. On the provincial dimension, we set the dummy variable DPR to 1 for 13 provinces (cities) including Tianjin, Heilongjiang, Shanghai, Jiangsu, Zhejiang, Shandong, Hubei, Guangdong, Hainan, Chongqing, Sichuan, Guizhou, and Shaanxi. Otherwise, set it to 0. On the year dimension, we set the dummy variable DYE, with 2016 and beyond set to 1, and before 2016 set to 0. The interaction term coefficient obtained from this (denoted as DPY = DPR × DYE) can measure the effectiveness of the service trade innovation policy issued in 2016. As shown in column (1) of Table 8, this policy has promoted the level of integration between the two industries, with a coefficient of 0.0110, and being significant at 10%. At the same time, in order to prove that this effect is indeed caused by the service trade innovation policy, we conducted counterfactual analysis before and after the event. We assume that the policy occurred in 2015 and 2018, respectively. As shown in columns (2) and (3) of Table 8, the results are not significant. This reduces the likelihood of the estimation results violating the same trend assumption.

Table 8 Results of the relationship between DPY and IIML.

Second, after determining that innovation policies in service trade can promote IIML, we used the instrumental variable method to test endogeneity. Considering that this study has two explanatory variables, IIML and IIML2, we select two instrumental variables. The first one is the product of the policy shock dummy variable (denoted as ${DPY}={DPR}\times {DYE}$) and the IIML of each province in the previous year of policy implementation (denoted as IIML2015). The second one is the product of the policy shock dummy variable (DPY) and the IIML2 of each province in the previous year of policy implementation (denoted as IIML22015). They were denoted as ${{DPY}2015}_{{it}}={DPY}\times {IIML}2015$ and ${{DPY}22015}_{{it}}={DPY}\times {IIML}22015$, respectively. The logic of using ${{DPY}2015}_{{it}}$ and ${{DPY}22015}_{{it}}$ as instrumental variables is as follows. In terms of correlation, there are differences in the impact of service trade innovation policies on different provinces. For provinces with higher initial IIML, the effectiveness of service trade innovation policies is better. Therefore, ${{DPY}2015}_{{it}}$ and ${{DPY}22015}_{{it}}$ satisfy the correlation condition of instrumental variables. In terms of exogeneity, this service trade innovation policy is a decision of the State Council of China and is not related to local provinces (cities). It is exogenous relative to each province (city). In addition to using the interaction terms ${{DPY}2015}_{{it}}$ and ${{DPY}22015}_{{it}}$, we also set provincial fixed effect to control for potential endogeneity in IIML2015. To ensure the reliability of the interaction term, we can also consider the product of the policy shock dummy variable (DPY) and the average IIML and IIML² of each province in the years before the policy was issued (denoted as ${\rm{IIMLV}}=\frac{{IIML}2014+{IIML}2015}{2}$, ${\rm{IIMLV}}2=\frac{{IIML}22014+{IIML}22015}{2}$, denoted as ${{DPYV}}_{{it}}={DPY}\times {IIMLV}$, and ${{DPYV}2}_{{it}}={DPY}\times {IIMLV}2$, respectively. Similar to ${{DPY}2015}_{{it}}$ and ${{DPY}22015}_{{it}}$, ${{DPYV}}_{{it}}$ and ${{DPYV}2}_{{it}}$ also satisfy correlation and exogeneity.

As shown in Table 9, the results of the first stage are reported in columns (1) and (2) when using ${{DPY}2015}_{{it}}$ and ${{DPY}22015}_{{it}}$ as instrumental variables, and in columns (3) and (4) when using ${{DPYV}}_{{it}}$ and ${{DPYV}2}_{{it}}$ as instrumental variables. The results show that these two sets of instrumental variables are significantly correlated with IIML and IIML², respectively.

Table 9 Results of the relationship between DPY and IIML.

As shown in Table 10, the results of the second stage are reported in column (1), when ${{DPY}2015}_{{it}}$ and ${{DPY}22015}_{{it}}$ were used as instrumental variables, and in column (2), when ${{DPYV}}_{{it}}$ and ${{DPYV}2}_{{it}}$ were used as instrumental variables. It can be observed that the results in columns (1) and (2) are significant, and there is still a significant inverted U-shaped relationship between IIML and RCE. Moreover, both sets of instrumental variables meet the applicable conditions. For ${{DPY}2015}_{{it}}$ and ${{DPY}22015}_{{it}}$, their unidentifiable test statistic (KP-LM statistic) is 3.763, Chi-sq p = 0.0524 < 0.1, rejecting the null hypothesis of insufficient identification of instrumental variables, i.e., passing the unidentifiable test. The weak identification test statistic (KP-F statistic) is 23.574 > 7.03 (10% Stock Logo threshold), indicating that it has passed the weak instrumental variable test. The unidentifiable test statistic (KP-LM statistic) for ${{DPYV}}_{{it}}$ and ${{DPYV}2}_{{it}}$ is 3.769, Chi-sq p = 0.0522 < 0.1, rejecting the null hypothesis of insufficient identification of instrumental variables, i.e., passing the unidentifiable test. The weak identification test statistic (KP-F statistic) is 22.306 > 7.03 (10% Stock Logo threshold), indicating that it has passed the weak instrumental variable test (Bendig et al., 2024).

Table 10 Instrumental variable method for endogeneity test.

Further analysis

Mediating effects

According to the analysis in Section “Empirical results”, we find that IIML has an inverted U-shaped effect on regional RCE. To further study its impact mechanism, based on the discussion in section “The mediating effects of TI and ISI”, further analysis in this section considers two aspects: TI and ISI.

The mediating effect of TI (MTI and LTI)

TI includes MTI and LTI, which are used as the dependent variables for further exploring the impact of IIML on TI, and thereby further explaining the mechanism of IIML on RCE. We can see it in Table 11 and Table 12.

Table 11 The mediating impact of IIML on MTI and LTI.

Table 12 Bootstrap test of mediating variables MTI and LTI.

First, column (1) estimates the impact of IIML on MTI, whereas column (2) shows the impact of IIML on IIML. Specifically, For MTI, the coefficient of IIML is 303.0645 (5% confidence level), indicating that IIML contributes to MTI. For LTI, the coefficient of IIML is 1818.42(1% confidence level). The IIML level also contributes to LTI. In summary, the hypotheses H2a and H2b are supported.

Second, to further explore whether the mediating effect is complete mediation or partial mediation, we conduct another test. Considering that this study applies a nonlinear model, this paper is more suitable to use the non-parametric percentile Bootstrap method for testing instead of the stepwise method and Sobel test (Lin and Li, 2022). We used a total of 500 Bootstrap samples. Regarding the mediating effect of MTI, according to the results in Table 12, the confidence interval for the indirect effect (_cs_1) does not include 0, indicating the existence of a mediating effect. The confidence interval for the direct effect (_fs_2) does not include 0, indicating the partial mediation effect of MTI. Similarly, regarding the mediating effect of LTI, according to the results in Table 12, LTI also partially mediates the impact of IIML on carbon emissions reduction. In summary, H2a and H2b are again supported. At present, IIML can improve manufacturing production efficiency and logistics circulation efficiency through technological innovation, achieve resource conservation, and thus reduce RCE.

In summary, IIML can achieve TI, improve manufacturing production efficiency and logistics circulation efficiency, enhance resource conservation, and thus reduce RCE.

The mediating effect of ISI (AIS and RIS)

ISI includes AIS and RIS, which are used as the dependent variables for exploring the impact of IIML on ISI, and further for illustrating the impact mechanism. The results are given Table 13.

Table 13 The mediating effect of IIML on AIS, RIS.

Column (1) estimates the impact of IIML on AIS, whereas column (2) shows the impact of IIML on RIS. Only the coefficient of IIML in the column (1) is −0.7961 (10% confidence level). The coefficients of IIML in the column (2) are not significant. The current IIML level is not conducive to AIS and is not related to RIS. In summary, H3a and H3b are not supported.

This could be due to the current lack of IIML, and AIS and RIS are still at an initial stage. On the one hand, the new industrial and business models nurtured by IIML may still be under-exploration and under-development (Tian and Pang, 2023). Currently, emerging industries have not completely replaced traditional industries, which often have a high RCE. On the other hand, although IIML promotes cooperation between the manufacturing industry and the logistic industry, in the case of a low IIML level, cooperation between the two industries is still dominated by traditional forms of cooperation, which is inclined to be a buyer-seller relationship. The secondary industry dominates the relationship. While the tertiary industry serves the secondary industry. This results in an immature service-oriented transformation and a low degree of industrial upgrading. Therefore, ISI (AIS and RIS) may temporarily lead to an increase in RCE. However, as IIML further develops, the industrial structure may gradually be optimized and advanced.

To further explain why ISI cannot serve as a mediating variable between IIML and RCE, we conducted the following data analysis. First, we set a dummy variable high_IIML, where high_IIML = 1 indicates that the IIML value is greater than the IIML mean, otherwise it is set to high_IIML = 0. We considered AIS and RIS as mediating variables in ISI in these two scenarios. The results in Table 14 show that when high_IIML = 0, AIS is significantly negative as a mediating variable, while RIS is positive but not significant as a mediating variable. In other words, considering the inverted U-shaped relationship between IIML and RCE, when IIML is low, IIML promotes an increase in carbon emissions, but IIML can to some extent suppress the increase in carbon emissions by inhibiting the upgrading of industrial structure. When high_IIML = 1, AIS is positive but not significant as a mediating variable, while RIS is significantly negative as a mediating variable. In other words, when IIML is high, it can promote the rationalization of industrial structure, thereby reducing carbon emissions.

Table 14 The role of ISI under different IIML level.

On this basis, we analyzed the indicator composition of AIS and RIS, constructed advanced indicators for the industry itself, and explored the role of advanced manufacturing and logistics industries in the process of integrated development. Following the previous studies (Cheng et al., 2018; Zhang et al., 2022), we have constructed two indicators of industry upgrading, namely Logistics Industry AIS (LIAIS) and Manufacturing Industry AIS (MIAIS), based on existing AIS. In doing so, we attempted to explain the performance of ISI indicators as mediating variables from the perspective of the industry itself.

The results shown in Table 15 indicate that for LIAIS, no matter high_IIML = 0 or high_IIML = 1, LIAIS is significantly positive as a mediating variable. This may be because, regardless of the level of IIML, the logistics industry is a necessary condition for the operation of other industries such as manufacturing. When IIML is low, IIML can promote the specialization of the logistics industry, such as forming specialized third-party logistics enterprises. This helps manufacturing enterprises improve logistics operation efficiency, thereby helping to accelerate the RCE to reach its extreme turning point and achieve carbon emissions reduction. When IIML is high, the industrial upgrading of the logistics industry is conducive to the deep integration of the logistics and manufacturing industries, i.e., participating in the entire supply chain management of the manufacturing industry and jointly promoting the low-carbon development of the supply chain. The above results are also consistent with empirical evidence. China’s JD Logistics and home appliance manufacturing enterprises continue to integrate and innovate, empowering JD Logistics construction and promoting the continuous improvement of LIAIS. When IIML is low, JD Logistics optimizes the operational efficiency of manufacturing enterprises and reduces energy consumption during transportation through specialized warehousing management and distribution algorithms. When IIML is high, JD Logistics delves into the operation process of manufacturing enterprises. For example, JD Logistics has deep cooperation with Midea Group (a Chinese home appliance manufacturing enterprise) to jointly create an intelligent supply chain system (Tech.cnr.cn., 2024). As IIML becomes more higher, JD Logistics promotes the use of green packaging materials and circular packaging technology to achieve a higher level of IIML (JDL.com., 2024). This not only helps JD Logistics to upgrade itself, but also further achieves carbon reduction.

Table 15 The performance of industrial upgrading at different levels of IIML.

Furthermore, for the MIAIS, when high_IIML = 0, MIAIS is positive as a mediating variable, but not significant. When high_IIML = 1, MIAIS is significantly negative as a mediating variable. This may be because, with the increasing integration of manufacturing and logistics industries, resources are often invested in areas that promote the coordinated development of the two industries, such as intelligent logistics systems. This reduces the investment in advanced manufacturing enterprises, making it easier for them to face resource shortages in their advanced construction, thereby hindering MIAIS. In other words, a negative correlation between IIML and MIAIS can exist and lead to an increase in total carbon emissions, which is not conducive to promoting carbon reduction. These results can also find the support from practice. We take Gree Electric Appliances, a manufacturing enterprise in China, as an example. As the degree of integration and innovation between Gree Electric Appliances and its logistics cooperation enterprises continue to improve, they are more inclined to invest in areas of coordinated development, such as the construction of intelligent logistics systems and supply chain information sharing platform (Gree.cn., 2023). In this way, making diversified investment in different areas can reduce the investment of Gree Electric in e.g., industrial products, namely squeezing the resources used for its own advanced construction (Thepaper.cn., 2023). However, this does not mean that there is no need to promote the upgrading of the manufacturing industry. When IIML is high, promoting MIAIS can also help the business cooperation between the logistics and manufacturing industries, thereby achieving carbon reduction through promoting RIS.

In addition, to analyze whether there are potential measurement issues, we selected two alternative indicators of ISI for robustness testing: (1) the natural logarithm of AIS and RIS for processing; (2) the share of high-tech industries in the province (the ratio of high-tech industry output value to the province’s GDP). As shown in Table 16, we find that compared with the case of selecting AIS and RIS as mediating variables, the results are still not significant, when using the two alternative indicators of ISI. This rules out the possibility that ISI could not be used as a mediator variable in the relationship between IIML and RCE, due to potential measurement issues. In addition, previous scholars (Zhang et al., 2022; Zhao et al., 2022) have also pointed out that optimizing industrial structure does not always play a positive role in achieving carbon reduction. For example, Zhao et al. (2022) pointed out that optimizing industrial structure may actually exacerbate carbon emissions by hindering energy efficiency and other means.

Table 16 Conduct robustness testing using alternative indicators of ISI.

Moderating effects

The moderating effect of GR

Referring to Bendig et al. (2024), we further explored how GR affects the relationship between IIML and RCE, as shown in Table 17. Among them, column (1) indicates the regression results under the full sample condition. The coefficients of the two interaction terms IIML × GR and (IIML)² × GR are not significant, indicating that H4 is not supported. This may be because IIML is relatively low, and GR is temporarily unable to significantly affect the relationship between IIML and RCE. Furthermore, in column (2), when only energy rich provinces (Shanxi, Shaanxi, Inner Mongolia, Gansu, Ningxia, and Xinjiang) are present, the coefficients of the two interaction terms, IIML × GR and (IIML)² × GR, are significant. However, in column (3), for non-energy rich provinces, the coefficients of the two interaction terms, IIML × GR and (IIML)² × GR, are not significant. This suggests GR in energy intensive provinces a greater impact on the relationship between IIML and RCE. Moreover, the interaction term coefficient of (IIML)² × GR is significantly positive, indicating that GR makes the inverted U-shaped curve flatter (Hanns et al., 2016, Bendig et al., 2024). This is because in regions with high energy utilization, GR can better guide the manufacturing and logistics industries to engage in carbon reduction activities. It makes the relationship between IIML and RCE more apparent, and results in a more gradual increase in carbon emissions when IIML is low (Cheng et al., 2024).

Table 17 The moderating effects of GR.

In summary, although H4 is not supported, GR significantly moderates the relationship between IIML and RCE in provinces that prioritize energy, making the inverted U-shaped relationship flatter.

The moderating effect of ST

Referring to Bendig et al. (2024), we further explored how ST affects the relationship between IIML and RCE, as shown in Table 18. Among them, column (1) indicates the regression results under the full sample condition, and the two interaction terms coefficients of IIML × ST and (IIML)² × ST are not significant, indicating that H5 is supported. This can be explained since the current IIML is relatively low, and ST is temporarily unable to significantly affect the relationship between IIML and RCE. Furthermore, considering that ST may be related to the overall economic development of the province, we calculated the mean UR in the panel data. We set a dummy variable HUR, with samples higher than the UR mean setting to 1 and samples not higher than the UR mean setting to 0, in order to conduct a grouping test. In column (2) of Table 18, it is indicated that the coefficients of the interaction terms IIML × ST and (IIML)² × ST are significant in samples above the UR mean. However, in samples below the UR mean in column (3), the coefficients of the interaction terms IIML × ST and (IIML)² × ST are not significant. This indicates that ST can have a greater impact on the relationship between IIML and RCE only when the urbanization level of the province is high. The coefficient of the interaction term of (IIML)² × GR is significantly positive, opposite to the coefficient of (IIML)², so ST makes the inverted U-shaped curve flatter (Hanns et al., 2016, Bendig et al., 2024). This is because in areas with higher levels of urbanization, the overall quality of the public is higher, and the moderating effect of ST in the relationship between IIML and RCE is more evident. When IIML is low, it is more conducive to making the growth of RCE flatter (Shuang et al., 2024).

Table 18 The moderating effects of ST.

In summary, H5 is not supported, but when the urbanization level of the province is high, ST significantly moderates the relationship between IIML and RCE, making the inverted U-shaped relationship flatter.

Heterogeneity analysis

Considering factors such as regional development and differences in industrial structure, there may be heterogeneity in the impact of IIML on RCE in different regions. Thus, this section conducts heterogeneity analysis from two perspectives: economic development level and industrial structure differences.

Heterogeneity analysis of economic development level

There are several important economic regions in China, including the Beijing Tianjin Hebei Economic Circle (Beijing, Tianjin, and Hebei Province), the Yangtze River Delta Economic Belt (Shanghai, Jiangsu Province, Zhejiang Province, Anhui Province), and the Pearl River Delta Economic Belt (Guangdong Province). These provinces are economically developed, while the remaining provinces are considered economically underdeveloped.

The heterogeneity regression results were obtained with different groups in columns (1) and (2) of Table 19. It shows the estimated results for economically developed provinces and economically underdeveloped provinces, respectively. Specifically, for the economically developed provinces, the coefficient of IIML is −0.0250 (10% confidence level), and the coefficient of (IIML)² is 0.0187 (not significant). For the economically underdeveloped provinces, the coefficient of IIML is 0.0771 (1% confidence level), and the coefficient of (IIML)² is −0.0466 (5% confidence level) and presents an inverted U-shaped shape. Furthermore, we used the Chow test to illustrate inter group differences and analyzed whether the relationship between IIML and RCE will vary among different provincial economies. As shown in Table 19, the results of the chow test are significant, with a statistical value of 3.5310. Thus, it can be concluded that there is a significant intergroup difference between economically developed and economically undeveloped provisions.

Table 19 Heterogeneity Test of IIML on RCE.

In sum, depending on the regional development level in different regions, IIML has heterogeneity in terms of its impact on RCE. Economically developed regions have advantages in IIML in terms of geographical location, economic development foundation, human resources, etc., which is more conducive to RCE through IIML (Wang et al., 2023).

Heterogeneity analysis of industrial structure differences

Provinces with a single industrial structure refer to areas where natural resource extraction and development are the main focus. These areas rely on a single traditional industrial model, mainly resource-based provinces and old industrial bases, including nine provinces (regions) such as Heilongjiang, Jilin, Liaoning, Shanxi, Shaanxi, Inner Mongolia, Gansu, Ningxia, and Xinjiang (People. cn., 2016). The economic characteristics of these regions are highly dependent on natural resource extraction and utilization, dominated by heavy chemical industry, and a relatively single industrial structure. Therefore, IIML may have different impacts on RCE in these regions, compared with the provinces consisting of more diverse industrial structures. In this section, these nine provinces are collectively referred to as provinces with a single industrial structure, while the other provinces have a diversified industrial structure.

The heterogeneity regression results were obtained through group regression in columns (3) and (4) of Table 19. Specifically, for the single province with a single industrial structure, the coefficients of IIML and (IIML)² are 0.0272 and 0.0010 (not significant), respectively. For the non-industrial structure single province, the coefficient of IIML is 0.0475 (1% confidence level). The coefficient of (IIML)² is −0.0324 (5% confidence level) and presents an inverted U-shaped shape. Similarly, we used the Chow test to illustrate inter group differences and further demonstrated whether the relationship between IIML and RCE varies depending on different types of industrial structures. As shown in Table 19, the results of the chow test are significant, with a statistical value of 8.6266. Thus, it can be concluded that there are significant inter group differences between provisions with a single industrial structure and provisions with a diversified industrial structure.

In sum, these indicate that, influenced by the type of industrial structure, IIML has regional heterogeneity in terms of its impact on RCE. The provinces with more diversified industrial structures and more flexible economic development models, often become easier to promote TI and ISI, resulting in carbon reduction effects.

Supplementary analysis using data of 276 Chinese cities from 2014 to 2021

This section expands the dataset to Chinese cities for supplementary analysis to enhance the statistical power of the model and improve the accuracy of regression results. In addition, it further supports the inverted U-shaped relationship between IIML and RCE. On the one hand, there are many missing data at the city level, and even with interpolation, there is still a significant margin of error compared to provincial (city) data. In the real world, manufacturing and logistics industries are mostly concentrated in developed areas, such as the southeast coast and the Yangtze River Delta. While provinces such as Xinjiang and Yunnan, have relatively few manufacturing and logistics industries. If the analysis is set at the city level, the data processed through interpolation and other methods is still difficult to reflect the reality. On the other hand, the administrative activities have a focus at the province level. The industrial performance of different Chinese cities in the same province is relatively similar, while there are significant differences in industrial performance among different provinces. For example, in Shaanxi Province, most cities have a coal based industrial structure, while in Guangdong Province, most cities have an electronic information manufacturing based industrial structure. The differences in industrial performance among cities within a province are not significant, while there are significant differences among provinces, for instance between Shaanxi Province and Guangdong Province. In summary, we have strengthened the inverted U-shaped relationship between IIML and RCE by expanding the dataset to the Chinese cities level, but still primarily using provincial (city) data as the main dataset.

Research design

Sample selection

In view of the principle of data availability, reliability and validity, this paper set the research sample as the balanced panel data of 276 Chinese cities from 2014 to 2021 (the data of Chinese cities with carbon emissions from 2022 to 2023 are not counted, due to much missing data). The data sources include: (1) IIML, TI, ISI, and control variables come from the “China Urban Statistical Yearbook” from 2014 to 2021, the China Research Data Service Platform (CNRDS), and other sources; (2) the carbon emission data of 276 Chinese cities comes from Emissions Database for Global Atmospheric Research (EDGAR), published by Joint Research Centre (JRC) of the European Union.

Model construction

Similar to the model construction based on 30 provinces, we constructed an empirical model based on data from 276 Chinese cities from 2014 to 2021. Model 3 is set as follows:

$${\mathrm{ln}}{{RCE}}_{{jt}}={\beta }_{0}+{\beta }_{1}{{IIML}}_{{jt}}+{\beta }_{2}{{{IIML}}_{{jt}}}^{2}+{\beta }_{3}{X}_{{jt}}+{\delta }_{j}+{\gamma }_{j}+{\varepsilon }_{{jt}}$$

(15)

Among them, subscript $j$ represents the city, subscript $t$ represents the year, ${{RCE}}_{{jt}}$ represents the carbon emission level of city $j$ in year $t$, ${{IIML}}_{{jt}}$ represents the coupling coordination between two industries of city $j$ in year $t$. ${X}_{{jt}}$ is the control variable set of the model, ${\delta }_{j}$ is the fixed effect of the city, ${\gamma }_{j}$ is the fixed effect of the year, and ${\varepsilon }_{{jt}}$ is the random error term. In addition, to eliminate the influence of extreme outliers, this study performed tail reduction on continuous variables.

Result analysis

Variable descriptive analysis

In this section, we focus on statistical analysis of variables and correlation analysis between variables. First, statistical analysis includes mean, standard deviation, minimum and maximum values, with results shown in Table 20. The average IIML in each city is 0.2987, with a standard deviation of 0.0827, and the minimum and maximum values are 0.1596 and 0.7215, respectively. The average RCE of each province is 0.0255, with a standard deviation of 0.0318. The minimum and maximum values are 0.0005 and 0.6426, respectively. Therefore, similar to descriptive statistical results of provincial (city) data, there are significant differences in IIML and RCE in Chinese cities.

Table 20 Variable descriptive analysis results.

Second, the test results of correlation analysis are presented in Table 21. It is shown in the table that, except for the correlation coefficient between the variable and its quadratic term, the correlation coefficient between other significant variable relationships does not exceed 0.8. It indicates no significant correlation between the research variables, further strengthening the effectiveness of the related variables in subsequent studies. Similar to the results of provincial (city) data, by analyzing the correlation coefficient between RCE and (IIML)², it can be found that the correlation coefficient between the two is significant, and the correlation coefficient is −0.1528. This indicates that there may still be an inverted U-shaped relationship between IIML and RCE based on data from Chinese cities, which will be further supported through regression models in the future.

Table 21 Correlation analysis of coefficients.

Benchmark analysis

Based on the econometric model 3 in Section “Model construction”, regression analysis was conducted on the full sample data of 276 Chinese cities using a panel model from 2014 to 2021. This section examines the impact of IIML on RCE.

First, similar to provincial (city) data, we used the Hausman test to determine whether to use a fixed effects model or a random effects model (Xiang, 2023), as shown in Table 22. The results show that for Chinese city data, bidirectional fixed effects should still be used.

Table 22 Hausman test.

Second, by regressing the first-order and second-order terms separately, we obtain the results in Table 23, in which columns (1) and (2) represent the estimated results of the relationship between IIML, (IIML)² and RCE, respectively. Specifically, similar to provincial (city) data, the coefficient of IIML in column (1) is 0.0280 is significantly positive at the 1% level. It indicates that the current level of (IIML)² has not effectively curbed carbon emission intensity. The first-order coefficient of IIML in column (2) is 0.0952, which is significantly positive at the 1% level, and the second-order coefficient −0.1135 is significantly negative at the 5% level. It indicates that although IIML has not effectively curbed RCE, there is an inverted U-shaped curve relationship between IIML and RCE. In other words, there is a threshold for the effect of IIML on RCE. When IIML in a city is below this threshold, the enhancement of IIML promotes RCE, while when IIML exceeds the threshold, the inhibitory effect of IIML on RCE dominates. Through calculation, the turning point of IIML is 0.4193. Similarly, referring to Greene (2018), we conducted a Wald test to statistically confirm the existence of turning points, testing whether both the first-order and second-order coefficients satisfy specific conditions. In this study, the first-order coefficient is positive and the second-order coefficient is negative. The results show that the Wald test statistic for IML is 11.13, p = 0.0010 < 0.05, and the p-value of its unilateral test is 0.0005 < 0.05; and the Wald test statistic for IIML² is 5.37, p = 0.0215 < 0.05, and the p-value of its unilateral test is 0.0108 < 0.05. The null hypothesis is accordingly rejected. Therefore, at the urban level, there is an inverted U-shaped relationship between IIML and RCE.

Table 23 Benchmark Regression on the impact of IIML on RCE.

In summary, based on data from Chinese cities, we obtain similar results as in the case of using provincial (city) data, namely the current IIML at the national level is still in its early stages. Meanwhile, the inverted U-shaped relationship indicates that the marginal promoting effect of IIML on the total RCE is continuously decreasing.

Robustness test

To verify the robustness of the relationship between IIML and RCE at the city level, this section conducts robustness tests from two aspects.

• (1) Exclude special regions. Similar to previous provincial investigation, at the city level, we also exclude energy rich cities to verify the robustness, with test results in Table 24. The results show that even after excluding major energy provinces, there is still an inverted U-shaped relationship between IIML and RCE, and the robustness issue is resolved.

  Table 24 The independent variable lag for endogeneity test.

• (2) Endogenous test: lagged tests of independent variable. Similar to previous provincial analysis, at the city level, we used lagged tests of independent variables to address endogeneity issues. The endogeneity results are shown in Table 24, showing that even after two periods of treatment lagged, there was still an inverted U-shaped relationship between IIML and RCE, and the endogeneity problem is resolved.

Heterogeneity analysis

Again, the heterogeneity of factors such as regional development level and industrial structure differences are also considered at the Chinese city level.

• (1) Heterogeneity analysis of economic development level. Unlike at the provincial level, the economic development level of Chinese cities often depends on regional GDP. We constructed a dummy variable iGDP to designate cities that meet the quasi trillion city standard (with a GDP greater than 800 billion yuan in the current year) as economically developed cities, i.e., iGDP = 1. Cities that do not meet the standard of a trillion-dollar city were considered economically underdeveloped, with iGDP = 0. Accordingly, group regression was performed to obtain the heterogeneity regression results shown in columns (1) and (2) of the Table 25. Apparently, both economically developed and underdeveloped cities exhibit an inverted U-shaped relationship between IIML and RCE. However, it can be found that the inflection point for economically developed cities is 0.09, and that for economically underdeveloped cities is 0.25. It indicates that due to the influence of urban economic foundation, IIML is more likely to achieve carbon emission reduction effects in more economically developed areas. Moreover, by comparing the IIML of cities with iGDP = 1, it can be found that all cities have crossed the inflection point and reached the stage of suppressing RCE through IIML. However, by comparing the IIML of cities with iGDP = 0, it can be found that not all cities have crossed the inflection point, that is, not all economically underdeveloped cities have reached the stage of suppressing RCE through IIML. Thus, a conclusion similar to that at the provincial level can be drawn, that is, the impact of IIML on RCE is regionally heterogeneous, influenced by the level of economic development.

  Table 25 Heterogeneity Test of IIML on RCE in city level.

• (2) Heterogeneity analysis of industrial structure differences. Similar to the provincial level, we classified all cities in provinces with relatively single industrial structures as cities with relatively single industrial structures, while the remaining cities as diversified industrial structures. The heterogeneity regression results were obtained through grouped regression, as shown in columns (3) and (4) of Table 25. It can be observed that both cities with a single industrial structure and cities with a diversified industrial structure exhibit an inverted U-shaped relationship between IIML and RCE. However, by calculating the inflection point, the inflection point for cities with a single industrial structure is at 0.39, and the inflection point for cities with a diversified industrial structure is at 0.17. These results indicate that due to the influence of differences in industrial structure, IIML is more likely to achieve carbon emission reduction effects in areas with richer industrial structures. Therefore, similar conclusions can be drawn as at the provincial level and city level, i.e., the impact of IIML on RCE exhibits regional heterogeneity.

Conclusions, contributions, and limitations

Research conclusions

Based on the empirical analysis, the following conclusions are drawn:

• (1) IIML is showing a slow upward trend, and it has an inverted U-shaped impact on RCE. However, it is still at the stage of promoting RCE with diminishing marginal promote effect. We use a coupling coordination model to calculate IIML, and further analysis on the impact of IIML on RCE. It is found that China is still on the left side of this inverted U-shaped curve, showing that the current IIML level has not reached the stage of suppressing RCE.

• (2) At present, IIML promotes MTI and LTI, while the effect of ISI has not been reflected. Through the analysis of mediating effects, it can be found that for the mediating effect of TI, the IIML is conducive to MTI and LTI. At the early stage of IIML, IIML can promote RCE through MTI and LTI. When the level of IIML exceeds a critical point, IIML reduces RCE through MTI and LTI. However, for the transmission path of ISI, the current IIML level is not conducive to AIS and is not related to RIS.

• (3) GR in areas with high energy utilization can make the inverted U-shaped relationship between IIML and RCE flatter, while ST in areas with high urbanization can make the inverted U-shaped relationship between IIML and RCE flatter. Through the analysis of moderating effects, it can be found that GR and ST have not effectively regulated the relationship between IIML and RCE in the entire sample. However, in regions with high energy utilization, GR moderates the relationship between IIML and RCE. In regions with high urbanization, ST moderates the relationship between IIML and RCE.

• (4) Considering regional heterogeneity, the inhibitory effect of IIML on RCE is more pronounced in economically developed provinces and provinces with a diversified industrial structure. From the heterogeneity analysis of economic development, IIML has a significant inhibitory effect on RCE for economically developed provinces, while for non-economically developed provinces, IIML on RCE still shows an inverted U-shaped pattern. From the heterogeneity analysis of industrial structure differences, there is no impact of IIML on RCE for provinces with a single industrial structure. However, for provinces without a diversified industrial structure, due to a more diverse industrial structure and a more flexible economic development model, the impact of IIML on RCE shows an inverted U-shape.

Research contributions

Theoretical contributions

From a theoretical perspective, this study has four contributions. First, although the inter organizational relationships theory has been used in previous research on carbon reduction, there has been no study on the relationship between resource alliances and carbon reduction (Shuang et al., 2024). Based on the legitimacy-based resource dependence perspective, this study constructs a complete theoretical framework for RCE effect of IIML. By exploring the direct, mediating, and moderating effects between IIML and RCE, it expands the combined application of inter organizational relationship theory in RCE.

Second, although previous studies on RCE (Wang et al., 2023) have investigated the importance of the industrial structure, integration between industries has not been considered. This study expands the perspective of RCE reduction and includes the mediating effect of IIML on RCE. It explores the impact mechanism of IIML on RCE through TI and ISI. This supplements the Kuznets inverted U-shaped curve hypothesis (EKC), thereby providing practical guidelines for achieving the dual-carbon plan.

Third, unlike previous studies focusing on the correlation, influencing factors, and economic effects of IIML (Baines et al., 2020; Yuan et al., 2020), we calculate IIML, focuses on analyzing the inverted U-shaped impact of IIML on RCE, and explores the impact mechanism of RCE through TI and ISI. It effectively supplements the literature related to IIML.

Fourth, differing from previous studies on the RCE reduction effects of manufacturing servitization and logistics service outsourcing phenomena in IIML from a micro perspective (Wang et al., 2023), this study adopts provincial-level data from a macro perspective. We conduct heterogeneity analysis based on economic development and industrial structure differences. It is found that economically developed provinces and provinces with a diversified structure have better carbon reduction effects in IIML. This can be combined with different regional characteristics to exert the carbon reduction effects of IIML.

Practical implications

From a practical perspective, this paper has four implications. First, from a long-term perspective, to accelerate the RCE reduction, the government should attach importance to IIML. This study finds that IIML has an inverted U-shaped effect on RCE, that is, at the early stage of IIML, RCE may increase due to technological and resource adjustments, but as IIML intensifies, RCE will gradually decrease. For the manufacturing industry, accelerating IIML not only improves its efficiency and competitiveness, but also helps achieve the dual carbon goals. Therefore, policy makers need to actively promote IIML, create favorable conditions, and enhance the enthusiasm of manufacturing industries in various provinces to participate in IIML, in order to fully release the inhibitory effect of IIML on RCE (Ye et al., 2020). For the logistics industry, IIML also has a profound impact. Through IIML, the logistics industry can optimize resource allocation, improve transportation efficiency, and thereby reduce RCE (Yi et al., 2020). Policy makers should pay attention to the role of the logistics industry in IIML and promote technological and business innovation in the logistics industry with the aid of policy guidance and market mechanisms, which ultimately facilitate the achievement of efficiency in RCE.

Second, the government should guide the manufacturing and logistics industries for RCE reduction from two aspects: TI and ISI. In terms of TI, to enhance IIML, the government should promote innovation activities in the two industries (Wang et al., 2021). Specifically, the government should help reduce innovation costs by investing infrastructures and establishing dedicated funds, so that enterprises are willing to invest resources in technology research and application. In terms of ISI, the iterative upgrading of manufacturing and logistics industries should be encouraged actively, and the improvement of industrial structure should be promoted (Zhao et al., 2022). The government should provide opportunities of communication and cooperation for promoting IIML, which jointly develop efficient and environmentally friendly technological means and management models.

Third, GR and ST can make the relationship between IIML and RCE flatter. On the one hand, GR can serve as an external formal regulation, making the role of IIML in carbon reduction flatter, especially in regions with higher energy utilization. Therefore, the government should pay special attention to implementing GR in regions with high energy utilization, such as Shanxi and Inner Mongolia, by increasing environmental policies and providing environmental subsidies. On the other hand, ST can serve as an external informal regulation and make the role of IIML in RCE flatter, which is more evident in regions with higher urbanization. Therefore, the government should actively increase the urbanization rate of each province (city), improve the quality of citizens and ST, in order to further promote the carbon reduction effect of IIML.

Fourth, in the process of IIML, the government needs to notice the imbalance and insufficient coordination between regions and promote IIML in different provinces according to local conditions. The imbalance and inadequacy of IIML level between regions have constrained IIML, thereby hindering the implementation of the dual carbon strategy. Therefore, for provinces with developed economy and provinces with a diversified industrial structure, the regional advantages should be leveraged to promote TI and ISI of the two industries. For other provinces with less developed economy and provinces with a single industrial structure, policy assistance should be increased. For example, policymakers should coordinate the relationship between industries, enterprises, and markets, encouraging those provinces to utilize regional resources reasonably on the basis of balancing economic development and environmental protection and transforming the resource advantages of provinces into regional development advantages (Tian and Pang, 2023).

Limitations and directions

This study has some limitations. First, this study mainly characterized IIML from a macro perspective and investigated the relationship between IIML and RCE. In the future, further exploration should aim to characterize IIML from a micro perspective and draw interesting conclusions through comparison. Second, due to limitations in data collection (carbon emission data on CEADS is only available up to 2021), this study included the data from 2014 to 2021. Although we have argued for the rigor of the data selection scope, a study using more recent data is still valuable in the future, especially after the carbon emission data is updated.