Analysis of technological innovation on provincial green development levels of logistics industry in China

The transition from traditional logistics to green and low-carbon logistics is crucial and inevitable due to the pressure of climate change and sustainable development in China. Meanwhile, technological innovation is perceived as an important factor affecting the development of the logistics industry. To explore the impacts of technological innovation and other factors and to propose proper policies based on the results, this study utilizes a generalized estimating equations (GEE) regression model to analyze panel data of 30 provinces during 2001–2019. Firstly, the entropy weight method is applied to calculate the green logistics development level based on an index system considering green factors. Secondly, a GEE model which considers the correlation among different observations is used to investigate the impacts of crucial factors on the green logistics development level. Moreover, regional heterogeneity is also analyzed in this paper by comparing the regression results of the Eastern region, Central region, and Western region. Based on the above analysis, several conclusions are drawn: (1) In terms of the average green logistics development levels, the Eastern region ranks 1st, the Central region ranks 2nd, and the Western region ranks 3rd. (2) GEE regression model is proved effective in our sample. (3) For the full sample, technological innovation, trade openness, and logistics infrastructure positively affect the green logistics development level; while, government regulation and energy intensity negatively influence the green logistics development level. (4) Regional heterogeneity is confirmed in our sample. Related policy recommendations are proposed based on our regional regression results. Take the Eastern region as an example, the local governments in the Eastern region should upgrade the manufacturing industry, reduce government financial investment in the transportation sector, and enhance environmental control expenditure in the transportation sector.


Introduction
With rapid development in recent years, the logistics industry has gradually become an important component of the Chinese national economy. It is a newly developed service industry on the base of the integration of various industries. Specifically, it consists of express transportation, energy transportation, cargo transportation, and various activities generated in the transportation process. It serves as a key link that connects different industries and consumers; therefore, it not only affects our daily life but also the manufacturing processes of various products.
With the rapid development of both transportation and information networks, the Chinese logistics industry has entered a stage of prosperous development. By the end of 2020, the total value of social logistics goods increased by 3.5% compared to its counterpart in 2019. The value exceeded 301 trillion Yuan in China, indicating that the Chinese logistics market is one of the largest markets in the world. The logistics industry has become another key driving force in the Chinese economy. It plays a crucial role in many economic aspects of China, such as promoting the transition and upgrading of the Chinese industrial structure and enhancing the overall economic strength of China. However, the ecological environment has been damaged to a certain extent in China due to the rapid development of the logistics industry, posing a threat to Chinese sustainable development (Ren et al., 2023b). Taking air pollutants as an example, the CO 2 emissions produced by the transportation, storage, post, and telecommunication services (TSPTS) industry was 732.48 million tons in 2019, ranking 4th place in all sectors in China. The serious air pollution caused by the logistics industry is mainly related to the increasing number of vehicles used for transportation. Other types of pollution caused by the logistics industry include the overuse of resources, plastic pollution, and waste. Take gross energy consumption as an example, the gross energy consumption of the TSPTS industry reached 439.09 million tons coal equivalent in 2019, accounting for approximately 9% of Chinese gross energy consumption. Pollution issues caused by the logistics industry are also urgent problems faced by many governments worldwide, and the approach to reducing pollution while achieving high-quality development in the logistics industry receives great attention (Barut et al., 2023;Tsolakis et al., 2022). It becomes an important task for governments to explore a way to realize green and sustainable development in the logistics industry.
Green and low-carbon development become an inevitable trend for the Chinese logistics industry as much attention has been attached to sustainable development. The Chinese government has endeavored to achieve sustainable development since 1995 when sustainable development was perceived as a crucial development strategy. Later on, the Chinese government made many plans and promulgate several policies to facilitate sustainable development. With the guidance of relative policies and plans, the economic development mode has shifted from high-speed development to high-quality development in China since 2017. Consequently, to accomplish high-quality development and promote sustainable and green development, the Chinese government has promulgated several policies, laws, and regulations to support the transition of the traditional logistics industry, like "Opinions on Promoting High-Quality Development of Logistics to Form Strong Domestic Market," "Promotion law of Chinese Environmental Recycling Economy," "Plan for the Development of a Modern Logistics System during the 14th Five Year," and "Comprehensive Working Scheme for Conserving Energy and Reducing Emissions during the 14th Five Year." With the help of these policies, the Chinese government hopes to establish a modern and green logistics system. The attempt to transform from a traditional logistics industry to a green and sustainable logistics industry is very common in the world as many countries try to achieve sustainable development and combat climate change (Guarnieri et al., 2020;Gupta and Garg, 2020). In this context, the promotion of green development in the logistics industry is an inevitable trend in China. Therefore, one of our motivations is to evaluate the development statuses of the logistics industry in different provinces.
Technological innovation is a crucial factor for the development of the green and low-carbon logistics industry (DGLCLI) (Zhang et al., 2020c). Specifically, technological innovation can not only improve the efficiency of different processes of the logistics chain but also facilitate the transition progress of the logistics industry. The improved efficiency is beneficial for the reduction of energy consumption, while the decrease of energy consumption leads to the decline of CO 2 emissions (Ren et al., 2023a). Thus, technological innovation can contribute to the DGLCLI in China. Trade openness, government regulation, infrastructure construction, environmental protection, and energy intensity are also perceived as key factors deciding the DGLCLI in previous studies. To be specific, the booming international trade may be conducive to the DGLCLI due to the scale effects and technology transition related to international trade (Zheng et al., 2020). Government regulation may enhance the DGLCLI as efficiency can be improved due to reasonable resource allocation, but it may also reduce the DGLCLI due to the crowding-out effect . Infrastructure construction can also affect the DGLCLI as it is a crucial factor in deciding the scale of the logistics industry (lo Storto and Evangelista, 2022). Environmental protection is also important for the DGLCLI as it can partly reflect the government's attitude . Energy intensity represents the energy utilization efficiency of the logistics industry which is a determinant factor of the DGLCLI (Bai et al., 2022). Another motivation of this study is to explore the impacts of these important factors on the DGLCLI.
The DGLCLI is usually denoted by two factors in previous studies, i.e., the logistics efficiency (LE) and development levels of the logistics industry (DLLI). LE is usually denoted by green total factor productivity (GTFP) or total factor productivity (TFP), and it is usually calculated by using data envelopment analysis (DEA), slack-based measurement (SBM), and other methods (Ren et al., 2023c;Ren et al., 2022). DLLI is usually evaluated based on an indication system. As the data used to calculate DLLI is usually objective data published by the government, the entropy weight method is widely applied to calculate DLLI. Regarding the impact of technological innovation on logistics, it is also widely explored in previous literature (see examples in Long et al. (2020) and Fan et al. (2022)). However, previous studies concentrate on the impact of technological innovation on the LE. Studies which measure the green development levels of the logistics industry in China are limited, let alone the impact of technological innovation on the green development levels of the logistics industry. Although the ordinary least square (OLS) is a classical regression method, and many regression methods are developed based on OLS, there are some limitations on the application of the OLSbased methods. The difference among the individuals in the sample is neglected when using the OLS-based methods; meanwhile, this kind of heterogeneity is very common in our sample as the development levels are various for different provinces in China. Unlike the OLS-based methods, the generalized estimating equations (GEE) considers the distinction among different provinces and the changes over different period of each province.
In summary, the main aims of this study include (1) evaluation of the development of green and low-carbon logistics industry in different Chinese provinces during 2001-2019; (2) investigation of the impacts of technological innovation and other control factors on the development of green and low-carbon logistics industry applying the GEE model; and (3) policy proposal based on the regression results. Moreover, the possible impact of regional heterogeneity is also considered in this study. To accomplish this goal, we firstly use the entropy method to calculate the green development levels of the logistics industry for 30 provinces based on an index system. The index system is proposed based on previous studies, and considers the environmental levels additionally. Then, we explore the influence of technological innovation on the green logistics development level by applying the GEE model. The control factors are also chosen based on previous literature. The data is collected in Chinese authoritative database, which can yield reliable results. Moreover, the distinction between different regions is also analyzed in our study. Finally, related policy recommendations are proposed based on our regression results.
The rest of the paper is organized as follows: "Literature review "section summarizes related literature. "Evaluation of green logistics development level in China" section evaluates the green development levels of the logistics industry for 30 provinces in China. "Model and analysis" section is about the regression model and its results. "Conclusions and policy recommendations" section concludes this study.

Literature review
Scholars usually use two factors to represent the DGLCLI, i.e., the LE and DLLI; thus, there are two main branches in previous studies related to the factors affecting the development of the logistics industry. Several recent papers are summarized in Tables 1 and 2, respectively. Apart from these studies, other scholars pay attention to the issues related to CO 2 emissions in the logistics industry, like the decoupling issues (Quan et al., 2020;Zhang et al., 2021) and influencing factors of CO 2 emissions (Guo and Wang, 2022).
With regard to LE, many scholars utilized GTFP or TFP as the indicator of LE, see examples in Zheng et al. (2020) and Li and Wang (2021). Few scholars applied other methods to calculate LE, like Yao et al. (2020) used a non-directional distance function method to estimate LE. Regarding the research contents, many scholars explored the impacts of different variables on LE; these factors included economic development, foreign direct investment, government regulation, technological innovation, human resources, and industry agglomeration level (details can be found in Table 1). The methods applied in these studies include spatial regression, Tobit regression, fixed-effect OLS, and GMM. Other scholars divided LE into technical progress and technical efficiency and studied the evolution of LE in different regions, see Liu and Xu (2020) and Mao et al. (2022). Moreover, Zheng et al. (2021) investigated the influencing factors of the link efficiency between the manufacturing industry and the logistics industry. Lei et al. (2020) gauged the impact of LE on the employment structure of 49 listed logistics companies in China. These studies are very valuable as they explore the impacts of different factors on LE, providing a foundation for policy proposals aiming to improve LE. However, LE is calculated based on serval common factors related to inputs and outputs; it cannot reflect the DGLCLI comprehensively.
For the strand of DLLI, scholars usually calculated DLLI based on an index system firstly. Generally, the index system considers the infrastructure and scale of    Tian and Zhang (2019) and Zhang et al. (2020b)). As attention has been paid to sustainable and low-carbon development in recent years, indicators related to the environmental and green development of the logistics industry are also included in some recent works (Fan et al., 2022;Zhang et al., 2020a;Sun et al., 2023). The common method used to calculate DLLI is the entropy method; other methods included fuzzy-AHP, combined weights ,and ISM. After obtaining the DLLI, scholars may (1) help company make decisions by appraising the DLLI for different objectives (Ni et al., 2019;Zhang et al., 2020a); (2) explore the coupling relationship between logistics development, economic development, and ecological development (Lan and Tseng, 2017;Zhang et al., 2020b); and (3) investigate the impacts of other factors on logistics development (Fan et al., 2022;Sun, 2017;Tian and Zhang, 2019;Wang et al., 2023). These factors include economic development, urbanization, and human resource. Moreover, Xu and Wang (2017) studied the impacts of DLLI on economic development. These studies are also very valuable as they provide many policy recommendations based on their results. However, the impact of technological innovation is overlooked in previous studies.
To sum up, although the studies related to the development of the logistics industry are rich, plenty of them concentrate on the analysis of the influencing factors of logistics efficiency. Studies which aim to investigate the impact of technological innovation on provincial green development levels in China are limited. Tian and Zhang (2019) utilized a spatial regression model to investigate the impacts of logistics infrastructure, economic development, information level, and human resources on provincial logistics development levels in China. However, environmental and green indicators are neglected when they appraise the provincial logistics development level; moreover, they failed to explore the impact of technological innovation in their regression analysis. Regarding the methods, OLS-based methods are widely applied in previous studies, but these methods cannot solve the possible individual heterogeneity issues. Therefore, the GEE model is utilized in this paper. In summary, to bridge the literature gap, we firstly propose an index system, which contains indicators related to logistics infrastructure, logistics scale, economic development, information level, and environmental level, and utilize an entropy method to evaluate the green logistics development level for 30 provinces between 2001 and 2019. Then, we conduct a regression analysis to investigate the impact of technological innovation on the provincial green logistics development level (GLDL). Their impacts on the Eastern region, Central region, and Western region are also compared and analyzed. Related policies are proposed based on the results.

Evaluation index system of green logistics development level
Based on previous studies, we conclude that the logistics development level is usually evaluated from four aspects, i.e., the logistics infrastructure, the logistics scale, the economic development, and the information level (see examples in Tian and Zhang (2019) and Fan et al. (2022)). Specifically, logistics infrastructure is the basis for the development of the whole industry; the scale of logistics is another crucial dimension; the economic development is the driving force for the development of the industry; informatization level is the driving force for the transformation of the traditional logistics industry to the green logistics industry. However, the environmental level is additionally considered in this paper to reflect the GLDL comprehensively. It reflects the impact degree of the logistics industry on the environment. The proposed evaluation index system of GLDL is shown in Table 3. The chosen secondary indexes are widely applied in previous studies. According to previous literature, 16 criteria are selected from the above five aspects.
The original data of 30 provinces (Hongkong, Macao, Tibet, and Taiwan are excluded due to the data accessibility) are mainly obtained from China Statistical Yearbook, China Energy Statistical Yearbook, China Science and Technology Statistical Yearbook, and provincial statistical yearbooks from 2001 to 2019. These data are collected by the authoritative institutions in China, and they are real statistical data reflecting the changes in different provinces. Most of the data can be obtained directly from the yearbook, and a small part of the data needs to be calculated and converted. The collected data has been confirmed and checked many times to ensure the accuracy, reliability, and objectivity of the data.

Evaluation method
In this section, we use the entropy weight method to calculate the weights of different indexes, and then the GLDL of different provinces can be obtained. As the data in our study is objective data, the entropy weight method is widely applied to calculate the weights based on objective data (see other examples in Fan et al. (2022) and Zhang et al. (2020b)). Another widely applied method to calculate the weights of different indexes is the analytic hierarchy process (AHP). The weights obtained by the AHP method are influenced by experts' subjective opinions. Unlike the AHP method, the entropy weight method takes advantage of the information of the objective data, and provides objective weights.
As the units of the original data are inconsistent, it is necessary to standardize the original data to overcome the influences of different dimensions and orders of magnitude, ensuring that the calculated results are authentic and reliable. Let n denotes the number of sample, T represents the time span of the observation period, and m denotes the number of indicators. In our case, n equals to 30, T is 19 years, and m equals to 16. Let x t ij represents the value of the j-th index of sample i in year t (i =1, 2, …,n; t =1, 2, …, T; j=1, 2, …, m). There are 16 indexes in the proposed evaluation index system of the green development level of the logistics industry, including 14 positive indexes and two negative indexes. The standardized treatment of these indexes is as follows: For positive indexes, Eq. (1) is used for standardization: For negative indexes, Eq.
(2) is used in the standardization process: where max x t ij represents the maximum value of the j-th index in the sample in year t. min x t ij denotes the minimum value of the j-th index in the whole sample in year t. x t′ ij stands for the standardized value of the j-th index of sample i in year t, and x t� ij ∈ [0, 1]. Based on the standardized data, this paper uses the entropy weight method to determine the weight of each index, and calculates the values of the green development level of the logistics industry of each province in the sample years. The calculation process is shown as follows: (1) Calculate the proportion of the standardized value of each index: where y t ij represents the standardized proportion of the j-th index of sample i in year t.
(2) Calculate the information entropy of each index: where e t j denotes the entropy of the j-th index in year t. (3) Calculate the difference coefficient of each index: where d t j stands for the difference coefficients of the j-th index in year t.
(4) Calculate the weight of each index: where w t j represents the weight of the j-th index in year t. (5) Calculate the green logistics development level: where GLDL t i represents the green logistics development level of i-th province in t-th year.

Spatial-temporal analysis of green logistics development level
The average GLDL for each province during 2001-2019 is listed in Table 4. As shown in Table 4, eight of the top 10 provinces are located in the Eastern region, two of them are in the Central region, and only one of them situate in the Western region. Moreover, the rank of the average GLDL of the Eastern region, Central region, and Western region also indicates that the development of green logistics in the Eastern region is faster than that of the other two regions. To be specific, a solid economic foundation in the Eastern region contributes to the expansion of consumption market, which is beneficial for the expansion of the logistics industry as logistics serve as a bridge that connects producers and consumers. A good economic foundation also promotes the application of information technology. Moreover, the flat terrain in the Eastern region also is conducive to the construction of logistics infrastructure. Therefore, the green logistics development in the Eastern region is fast. Similarly, the Central region not only connects the Eastern region and the Western region, but also links the North and South of China. Several provinces in the Central region are the centers of the Chinese traffic network, like Henan, Anhui, and Hunan provinces. The location advantage of the Central region is beneficial for the development of its green logistics. On the contrary, many of the provinces in the Western region have a vast territory with a sparse population, and the economic scale and information level in these provinces are also relatively low. Consequently, the development level of green logistics in the Western region is low. However, owing to the "Sustainable development" strategy and "One Belt and One Road" Initiative, the development level of green logistics in the Western region rises quickly in recent years. The spatial distribution of GLDL of different provinces in 2001,2005,2008,2012,2015, and 2019 is shown in Fig. 1. It can be observed from Fig. 1 that the GLDL has a significant decreasing trend from the Eastern region to the Western region. Moreover, the GLDL of the coastal provinces is also obviously higher than that of the inland provinces. In other words, the spatial distribution of the GLDL confirms the ranks of the average GLDL of different provinces during 2001-2019. Guangdong ranks 1st place in terms of GLDL as it has a good economic foundation and logistics infrastructure. Moreover, Guangdong is also an important manufacturing base and a crucial trade base in China. Its environment also improves recently due to the continuous effects of its government and enterprises. Therefore, the GLDL is high in Guangdong. On the other hand, Qinghai is an inland province with weak economic performance and a fragile ecological environment. The logistics infrastructure in Qinghai is weak due to the fragile environment. The scale of the logistics industry is also small due to the weak logistics infrastructure and economy. As a result, the development level of green logistics in Qinghai is low.

Model and analysis
In this section, we firstly propose the econometric model we use in our paper; then we collect data from Chinese authoritative institutions; Thirdly, two traditional tests-the LM test, CD test, and unit root test are conducted to confirm the characteristics of the data. After that, the results of a full sample and subsamples considering regional heterogeneity are presented. Finally, we conduct a robustness test to confirm the reliability of our results.

Econometric model
Technological innovation is a crucial factor affecting the development level of green logistics. Specifically, research and development (R&D) can yield technologies or products which can improve energy efficiency or operation efficiency in the logistics industry, such as electronic motor cars with long endurance mileage, goods sorting systems, and the Internet of things. The promotion of such technology and high-tech products is beneficial for the development of green logistics. Moreover, technological innovation can also change the future development direction of the industry. With the development of the green development of logistics industry in China, companies in the logistics industry tend to utilize green technology. In summary, technological innovation and R&D can effectively promote the development of the industry and enhance competitiveness. Referring to Li and Wang (2022) and Shen et al. (2019), this paper uses the R&D expenditure of the high-tech industry (TI) to express innovation ability. Trade openness (TR) is an important influencing factor for GLDL. To be specific, the promotion of international trade can contribute to economic development, and most of the countries which actively involved in international trade have a vigorous economy. The promotion of international trade can drive the circulation of goods and services and facilitate the development of the logistics industry and its related industries. With reference to Khan et al. (2019) and Zheng et al. (2020), this paper uses the total values of import and export to denote TR.
Government regulation is also an important factor for GLDL. In China, the government mainly regulates the market order of the logistics industry through macro-control such as financial investment in the transportation section. The application of government regulation may facilitate the development of logistics infrastructure, optimize resource allocation of different links in the logistics industry, reduce related costs and increase efficiency, and improve the ecological environment. However, it may also impede the development of green logistics as too much government financial investment may hinder the investment from companies, which may reduce the market efficiency. Moreover, if the financial investment is used for infrastructure construction with high pollution, the development of green logistics would also be damaged. With reference to Long et al. (2020) and Xu and Xu (2022), this paper utilizes government financial input in the transportation industry (GFI) to represent government regulation.
Infrastructure construction is also an important factor affecting GLDL. Logistics infrastructure not only involves warehousing, computers, and transportation equipment, but also involves road, railway, and other transportation infrastructure. In summary, infrastructure construction lays the foundation for the development of green logistics. Unlike Tan et al. (2019) and lo Storto and Evangelista (2022) who used the total mileage of different transportation manners to represent logistics infrastructure (LF), we use the area of logistics storage land to represent the LF.
Environmental protection (EP) is an important factor which has an impact on the GLDL. The attitude of the logistics industry towards energy protection can be partly reflected by the governance capability and the construction of a governance system related to the environment in the logistics industry. To be specific, the utilization efficiency of resources can be improved by enhancing the EP of the logistics industry, which are conducive to the long-term and sustainable development of green logistics. In order to explore the impact of EP on GLDL, we use the proportion of investment in environmental pollution control to GDP to denote the EP (see other examples in Deng et al. (2020) and Wang et al. (2022)).
Energy intensity (EI) is also a crucial factor affecting the GLDL. As the market demand for the logistics industry is expanding, its energy consumption is increasing accordingly, especially the energy consumption of the transportation process. Although the government has implemented plenty of policies and measures to control energy consumption and pollution emissions, these policies still cannot resist the continuous growth of energy consumption, which jeopardizes the green development of the logistics industry. With reference to Bai et al. (2022) and Long et al. (2020), this paper uses the proportion of energy consumption to the added value of the logistics industry to express EI.
Although the main goal of this paper is to explore the impacts of TI on GLDL, other control variables are also considered based on the analysis above, such control factors include TR, GFI, LF, EP, and EI. STIRPAT model is a classical model in which the environment is assumed to be influenced by population, affluence, and technology. STIRPAT model is widely applied to explore the impacts of different factors on the environment. Thus, we also use the STIRPAT model to investigate the impacts of TI and other control factors on GLDL. Their relationship can be described by Eq. (8): where i and t denotes i-th province and t-th year respectively. GLDL represents green logistics development level, TI denotes technological innovation, TR stands for trade openness, GFI represents government financial investment, LF denotes logistics infrastructure, EP stands for environmental protection and EI denotes energy intensity.
In order to avoid the possible impact of data fluctuation, various units of different variables and heteroscedasticity, we use the logarithm form of the original data in our analysis. Thus, the regression model in our paper is shown in Eq. (9) where i is the provinces (i =1, 2, 3…, 30), and t is the year (t =2001, 2002…, 2019). β 1 -β 6 is the regression coefficient of these variables. ε it is a random error term.

Data
In order to explore the impact of TI on GLDL, this paper collects data for 30 provinces (Tibet, Hong Kong, Macao, and Taiwan are excluded due to the lack of data) during the period 2001 to 2019. The data used in this study is collected and sorted from several statistical reports of authoritative institutions, like the China Statistical Yearbook, China Financial Yearbook, and China Science and Technology Yearbook. These data are real data collected by the Chinese National Statistics Bureau, People's Bank of China, and so on. For some missing data in a particular year, the average interpolation method is adopted to calculate the missing data. The definition of each variable and its descriptive statistics are listed in Table 5.

LM test and CD test
As cross-sectional dependence may exist in panel data, it could result in biased regression results. Therefore, we conduct the LM test and CD test to check whether there exists a cross-sectional dependence in our data. Results of the two classical cross-sectional dependence tests are listed in Table 6, and it implies that cross-sectional dependence occurs in our sample as the hull hypothesis (there is (9) lnGLDL it = c + 1 TI it + 2 TR it + 3 GFI it + 4 LF it + 5 EP it + 6 EI it + it no cross-sectional dependence in the sample) is rejected. In order to solve the cross-sectional dependence issues, we apply a GEE model in our benchmark regression.

Unit root test
Before conducting a panel data regression, we should ensure that the data is stable. If the data is unstable, it is easy to yield bad regression results which affect the accuracy and consistency of regression results. Therefore, we conduct the unit root test firstly. As the data is cross-dependent, traditional first-generation unit root tests are not suitable. Therefore, we utilized two second-generation unit root tests in our paper, i.e., Pesaran CADF test and Pesaran CIPS test. The results are listed in Table 7. As shown in Table 7, these variables are stable after first difference.

National regression analysis results
As the conditions in different provinces vary greatly in China, we apply a GEE model which is proposed by Zeger et al. (1988). Compared with the traditional ordinary least squares (OLS) regression model, the distinction among different provinces and the changes over different periods of each province are considered in the GEE model. In other words, the correlation between observations is considered in the GEE model.
Regression results of the fixed-effect (FE) OLS, feasible generalized least squares (FGLS), and GEE model are shown in Table 8. The result of the GEE model is basically  consistent with that of FGLS, and is different from that of FE-OLS, which confirms that the result of OLS is biased. According to the regression result of the GEE model in Table 8, we can conclude that TI has a positive impact on GLDL. Specifically, a 1% increase in R&D can yield a 0.056% growth in GLDL. With regards to the control variables, TR and LF have a positive impact on GLDL, while GFI and EI have a negative impact on GLDL.
Several key results can be obtained from Table 8: (1) Technological innovation is conducive to the improvement of GLDL as it can enhance logistics efficiency. As shown in Table 8, TI has a significant positive impact on GLDL. The result implies that the green development of the logistics industry depends on the development of science and technology. R&D investment has continuously kept increasing in recent years, and the ever-increasing R&D severs as an important source for high-quality development and promotes the evolution of the logistics industry. Abundant R&D funds and a good research environment have yielded many research outputs which can be applied in many aspects related to the logistics industry. Such research outputs include the improvement and innovation of transportation equipment and the invention of eco-friendly materials for packaging. In summary, an increase in R&D investment have a positive effect on the green development of the logistics industry in China. It has also been confirmed that TI can reduce CO 2 and enhance green and low-carbon development in previous studies (Cheng et al., 2021).
(2) Trade openness can improve GLDL due to its positive impact on economic development. As displayed in Table 8, TR has a positive effect on GLDL. China is an important production base in the world, and international trade is a crucial component of China's economy. An increase in the total value of imports and exports can significantly stimulate economic development in Chia. As an important bridge linking both domestic and international markets, the green development of logistics will also accelerate with a continuous increase in the economy. Therefore, TR positively affects GLDL. The positive impact of TR on logistic performance has also been proved by Khan et al. (2019).
(3) Government regulation reduces GLDL due to the crowding-out effects of governmental investment. As shown in Table 8, GFI has a significant negative impact on GLDL. When government commits large financial investment in the transportation section, investment chance for investors from the private sector is occupied. According to classical economic theory, this may cause a decrease in market efficiency. Consequently, the development of green logistics may be hindered. Moreover, the government's financial investment is mainly used to build highways, railways, or airports, which may cause pollution as much steel and energy  are needed in the construction process. Therefore, an increase in GFI negatively influences the GLDL. Xu and Xu (2022) also confirmed that mandatory government regulation reduced the efficiency of the logistics industry. (4) Construction of logistics infrastructure can enhance GLDL as it is a physical foundation for GLDL. As shown in Table 8, LF has a significant positive impact on GLDL. Logistic infrastructure, which is denoted by land area for logistical storage in this paper, lays a solid foundation for the development of the logistics industry. An increase in LF implies the foundation of the logistics industry becomes more solid; thus, the GLDL increases accordingly. Similarly, Tan et al. (2019) found that LF positively influences logistics efficiency.
(5) The effect of environmental protection on GLDL is not significant.
As shown in Table 8, EP has an insignificant negative impact on GLDL. Although the government's investment in environmental control has increased year by year, the growth rate of environmental protection cannot keep up with the growth rate of GDP; this will lead to an imbalance between economic development and environmental control. Therefore, EP negatively affects GLDL, but its impact is not significant. Wang et al. (2022) also confirmed that EP positively affected air pollutants in their time-individual fixed effect, indicating that EP may jeopardize green and sustainable development.
(6) Energy intensity reduces GLDL as more pollution is emitted.
As shown in Table 8, EI has a significant positive impact on GLDL. Energy intensity of the logistics industry represents the utilization efficiency of the logistics industry. Currently, traditional fossil fuels which result in environmental pollution are still widely applied as long-distance vehicles using clean energy are still in the design and test process. An increase in EI indicates that more energy is needed to yield the same GDP; this will cause damage to the current environment and impede the green development of the logistics industry. Similarly, Bai et al. (2022) proved that EI negatively affects logistics efficiency.

Regional regression analysis results
As the conditions in different regions vary greatly in China, we conduct a heterogeneity analysis in which 30 provinces are divided into the Eastern region, Central region, and Western region in this study. The heterogeneity analysis can help us gauge the impacts of each variable on the GLDL in the three different regions; then, we can propose different policies based on the regression results.
The results are listed in Table 9. For the Eastern region, TR and LF have a significantly positive impact on GLDL, while GFI and EP exert a negative influence on GLDL. With regards to the Central region, TI and LF have a significantly positive impact on GLDL, while EI exerts a negative influence on GLDL. Regarding the Western region, TI, TR, and LF have a significantly positive impact on GLDL, while GFI exerts a negative influence on GLDL.
(1) An inverted U-shaped relationship between TI and GLDL causes different regional results. For TI, the impact of TI on GLDL is not significant for the Eastern region. This may be caused by the scale effect of R&D, i.e., the impact of R&D investment on GLDL will gradually increase when the technological level is low; however, the influence of R&D investment will become small when the technological level exceeds a turning point. To be specific, when the technological level is low, it is easy to realize technological innovation as it is relatively easy to invent an early-stage technology. Meanwhile, with the continuous development of early-stage technology, it will become more and more difficult to modify sophisticated technology as more resources are needed. This is partly confirmed by the values and significance of coefficients of TI in the three regions: the coefficient of TI in the Central region is higher than that of TI in the Western region, meanwhile, the coefficient of TI in the Eastern region is insignificant.
(2) Location advantage and industry structure are the reasons for the difference in regional results. With regard to TR, the impact of TR on GLDL is insignificant in the Central region. International trade is prosperous due to the "One Belt and One Road" initiative in the Eastern region and Western regions due to their good location advantage. The volume of international trade in the Central region is small. Therefore, the impact of TR on GLDL is small in the Central region. Moreover, many heavy industries with high pollution are situated in the Central region, and an increase in TR in the Central region will result in severe environmental damage, which will hinder the green development of logistics in the Central region. Consequently, the sign of TR in the Central region is negative.
(3) The low government financial investment is responsible for the difference in regional results.
Regarding GFI, the impact of GFI on GLDL is not significant in the Central region. GFI in the Central region is less than that in the Eastern and Western regions. Thus, the crowding-out effect of government investment in the Central region is the smallest of the three regions, indicating that the impact of GFI on GLDL is small in the Central region.
(4) U-shaped relationship between LF and GLDL is the reason for different regional results. For LF, the impact of LF on GLDL is significant in these three regions, but the coefficient of LF in the Central region is smaller than the counterparts in the Eastern and Western regions. In other words, a U-shaped relationship exists between LF and GLDL, i.e., the impact of the land area of logistical storage on the GLDL will decrease firstly, then start to increase after exceeding a turning point. (5) EP in different regions causes different regional results.
With regards to EP, EP has a significant negative impact on GLDL in the Eastern region. As mentioned in "National regression analysis results" subsection, the gap between the growth rate of GDP and environmental control input causes a negative relationship between EP and GLDL. Moreover, as the economic development in the Eastern region is better than that of the Central region and the Western region, this kind of gap is more significant in the Eastern region; therefore, EP exerts a significantly negative impact on GLDL in the Eastern region. (6) Energy structure is responsible for the difference in regional results.
Regarding EI, the coefficient of EI is only significant in the Central region. Several provinces in the Central region are an important energy base and heavy industry base for China, such as Shanxi, Heilongjiang, and Jilin. In other words, many fossil energy production centers or high-pollution industries are located in the Central provinces. In contrast, many renewable energy projects or high-tech industries are situated in the Eastern or Western region. Therefore, EI has a significant negative impact on GLDL in the Central region.

Robustness test
Two kinds of robustness tests are conducted to confirm the regression results are reliable in our study. Firstly, we compare the regression results of the different regression models. The comparison result is listed in Table 8. As displayed in Table 8, the results of FE-OLS, FGLS, and GEE are basically consistent. Secondly, we use the number of patents (NP) to denote technological innovation instead of R&D in the GEE estimator. The results are listed in Table 10. As the two regression results are consistent, we can conclude that our result is robust.

Main conclusions
To explore the impact of technological innovation on the provincial green logistics development level in China, this study utilizes a GEE regression model to analyze panel data of 30 provinces during 2001-2019. Firstly, the entropy method is applied in this paper to calculate the green logistics development level based on an index system considering green factors. Secondly, a GEE model which considers the correlation among different observations is used to investigate the impact of technological innovation, trade openness, government financial investment, logistics infrastructure, energy protection, and energy intensity on the green logistics development level. Moreover, region heterogeneity is also analyzed in this paper by comparing the regression results of the Eastern region, Central region, and Western region. Based on the above analysis, several conclusions are proposed below: (1) The average rank of the green logistics development level for three regions in China is Eastern region > Central region > Western region.
(2) The GEE model can provide more reliable results as the values of green logistics development levels, technological innovation, and other control variables are heterogenous for different provinces in China.
(3) Technological innovation, trade openness, and logistics infrastructure positively affect the green logistics development level from the perspective of the full sample; meanwhile, government financial investment in the transportation industry and energy intensity of the logistics industry negatively influence the green logistics development level. (4) The impacts of these factors on the green logistics development levels are different for the three regions in China, confirming the existence of regional heterogeneity.

Policy implications
One important aim of this study is to provide policy-makers with some suggestions; thus, the managerial insights of this study are from the governmental perspective. As the impacts of each factor vary for the three regions in China, we propose related policy recommendations according to the regional regression results to make sure these policies are reliable.
(1) For the Eastern region, trade openness and logistics infrastructure positively affect the green logistics devel-opment level; while, government financial investment in the transportation industry and environmental protection input has a negative impact on the green logistics development level. Therefore, in order to keep a high green logistic development level, provinces in the Eastern regions should (i) upgrade the manufacturing industry. China, especially the Eastern region of China, is an important production base in the world. By upgrading the manufacturing industry, products with high technology are produced and exported, energy intensity will also reduce. These are beneficial for the green development of the logistics industry; (ii) reduce government financial investment in the transportation sector. The government in the Eastern regions can utilize other measures to encourage private investors to participate in the transportation investment, like a public-private partnership (PPP); and (iii) enhance environmental control expenditure in the transportation sector. The increase rate of environmental protection expenditure lags behind the growth rate of economic development. Therefore, economic development is achieved at the cost of environmental damage. In order to modify this situation, more environmental protection investment should be committed.
(2) For the Central region, technological innovation and logistics infrastructure positively affect the green logistics development level; while the energy intensity of the logistics industry negatively influences the green logistics development level. Therefore, to keep a high green logistic development level, provinces in the Central regions should: (i) maintain R&D investment as technological innovation are conducive to the green development of the logistics industry; (ii) modify energy supply and consumption structure. Many provinces in the Central region rely on energy production and heavy industry and these could hinder the green development of logistics; therefore, they could try to transform their energy structure to a green one; and (iii) actively participate in international trade as international trade is proven to be beneficial for the green development of the logistics industry.
(3) For the Western region, technological innovation, trade openness, and logistics infrastructure positively affect the green logistics development level; while government financial investment in the transportation industry negatively affects the green logistics development level. Therefore, to keep a high green logistic development level, provinces in the Western regions should: (i) increase R&D investment as technological innovation contributes to the green development of the logistics industry, (ii) encourage international trade by taking the opportunity of the "One Belt and One Road" initiative as international trade is beneficial for the green development of the logistics industry, and (iii) reduce government financial investment in the transportation sector. The government in the Western regions can also encourage private investors to take part in the construction of logistics infrastructure by providing PPP projects, and build-own-transfer projects.

Limitation and future research direction
Although we have explored the impact of technological innovation on the green logistics development level in this paper, there are still some limitations. Firstly, a linear relationship between technological innovation and green logistics development level is assumed in our model. However, according to the regional heterogeneity analysis, the relationship between them may not be linear. Therefore, we will further study their relationship in our future research. Secondly, the mechanism of how technological innovation affects green logistics development level is not fully investigated in our study; we will explore the impact mechanism in our future study. Lastly, although the impacts and related policies may be suitable for other countries as the conditions in these countries are similar to that in China, they are not applicable in any countries around the world. Therefore, the relationship between technological innovation and green logistics development levels in other countries can be explored in the future.
Author contribution Cheng Cheng: conceptualization, formal analysis, methodology, writing-review and editing, and visualization. Yanan Han: investigation, formal analysis, validation and data curation, and software. Xiaohang Ren: formal analysis, investigation, resources, project administration, and visualization.

Declarations
Ethics approval Not applicable.

Consent to participate Not applicable.
Consent for publication Not applicable.

Competing interests
The authors declare no competing interests.