Do China State-Level Economic and Technological Development Zones Have a Positive Effect on Regional Total Factor Productivity? A Perspective Based on the Moderating Effect of Transportation Infrastructure

Abstract

With the deceleration of China’s economic growth, the crude economic model will progressively diminish in its competitive edge, thereby posing challenges for state-level economic and technological development zones (ETDZs) in terms of transitioning their development model and grappling with low levels of total factor productivity (TFP). This study aims to evaluate the TFP of prominent cities in China, examine the influence of the establishment of state-level ETDZs on urban TFP, and investigate the moderating effect of transportation infrastructure on this relationship. The results show that the aggregate TFP of Chinese urban areas declined from 1999 to 2020, a trend linked to structural economic adjustments and persistent underutilization of capital in several regions. The establishment of state-level ETDZs has been found to exert a notable positive influence on regional TFP. The presence of transportation infrastructure plays a moderating role in facilitating state-level ETDZs, thereby enhancing regional TFP. Among various modes of transportation, highways and railways are particularly prominent in this regard. These conclusions provide a theoretical basis and decision-making reference for further unleashing the policy potential of development zones in China.

1. Introduction

Development zones have significantly contributed to China’s economic development during the period of reform and opening up. As a country with a robust manufacturing base, China has established numerous state-level economic and technological development zones (ETDZs), which have emerged as key hubs for industrial activity. During the period of rapid economic expansion, ETDZs significantly contributed to the country’s growth trajectory. In the current phase of economic transition, ETDZs are expected to play an active role in facilitating structural transformation and enhancing the quality of development. Total factor productivity (TFP) reflects the additional production efficiency achieved under given levels of input factors. Its growth mainly stems from technological progress and efficiency improvement, which play a crucial role in measuring the quality and efficiency of economic progress. In recent decades, the establishment of ETDZs in China has been shown to have a significant impact on regional TFP. Has this impact had a positive or negative effect? Are there additional factors beyond ETDZs that can exert an impact on the efficacy of development zone policies? The responses to these inquiries will assist both the central and local governments in effectively harnessing the potential of ETDZs in fostering high-quality growth and accelerating the establishment of a new development paradigm.

When scholars engage in the examination of state-level ETDZs in China, their attention is frequently directed to the comprehensive influence of these zones on economic growth. It is widely observed that ETDZs are consistently associated with economic activities, including the facilitation of factor migration [1] and the restructuring of economic factors [2]. The establishment of ETDZs has been found to have positive effects on the advancement of economic development [3,4,5]. This is achieved by enhancing infrastructural development, which in turn attracts potential investors [6], encourages greater foreign investment [7,8], and facilitates the promotion of international trade [9,10]. Furthermore, it has been contended by certain academics that the influence of ETDZs on the regional economy is not solely promotional. For instance, the land titling system employed in ETDZs. This process primarily relies on reduced land prices to entice foreign investors and may result in inefficient utilization of land resources and fail to yield immediate or intermediate advantages for local financial stability [7]. The correlation between the extent of land occupied by a development zone within a city and its corresponding impact on economic growth is not always directly proportionate [11]. The impact of ETDZs on the urban economy during their expansion can be characterized as the spatial reorganization effect of the city [2]. ETDZs have effectively achieved the objective of facilitating the restructuring of the city’s industrial structure and enhancing the potential for reallocating factors of production to drive economic development by establishing target industries and increasing their share within the city’s industrial sector [12]. The productivity of enterprises in ETDZs has been influenced by various factors, namely the “agglomeration effect”, “multiplier effect”, “policy effect”, and “selection effect” [13,14,15,16].

Transportation has a crucial role in facilitating the exchange of production inputs and stimulating regional economic growth [17]. As a foundational pillar of national economic development, the transportation sector maintains a dynamic and reciprocal relationship with regional economic systems. One strategy to enhance the efficiency of regional economic operation through transportation is by fostering the social division of labor. Additionally, the ongoing progress of the economy and society will inevitably influence transportation systems [18]. Actively promoting the construction of an integrated transportation network is conducive to the coordinated development of provinces and cities [19]. The presence of transportation infrastructure has been found to have a beneficial spillover effect on economic growth [20]. The enhancement of transportation infrastructure is advantageous for the advancement of urban businesses [21,22]. The transportation sector is intricately connected to several other sectors [23], and it indirectly stimulates the growth of these industries [24]. Various modes of transportation have distinct impacts on different industries [25]. For instance, air transportation notably stimulates the output of information industry services, medical equipment manufacturing, and electronic communication equipment manufacturing [26]. According to scholarly research, the influence of transportation on various industries can be categorized in a descending sequence as follows: primary industry, tertiary industry, and secondary industry [27]. The influence of transportation infrastructure on industrial development can be observed through its ability to lower transportation expenses, enhance transportation effectiveness, and stimulate the exportation of products [28]. Through the foregoing research, we believe that the future development of China’s economy should pay greater attention to the enhancement of TFP and give full play to the positive function of ETDZs. This study starts from the perspective of complex systems and regards state-level ETDZs as a key subsystem embedded in regional economic systems. It investigates how policy interventions associated with ETDZs can contribute to regional TFP growth by reshaping the flows and configurations of key production factors, including technology, capital, and labor. Although many studies have investigated the economic effects of China’s state-level ETDZs, few have systematically examined how transportation infrastructure interacts with these ETDZs to influence regional TFP. To address this research gap, this study integrates insights from transportation economics and regional development theory. Specifically, we employ a panel dataset of 282 prefecture-level cities in China from 1999 to 2020 and utilize a Difference-in-Differences (DID) approach to empirically examine the impact of ETDZs on regional TFP, as well as the moderating role of transportation infrastructure in this relationship.

The contributions and innovations of this study are as follows. First, this study systematically evaluates the impact of state-level ETDZs on regional TFP, moving beyond the traditional focus on innovation and revealing how ETDZs enhance regional TFP through mechanisms such as improved resource allocation, technological progress, and dynamic market selection. Second, this study introduces a moderating effect analysis framework for transportation infrastructure. Unlike the traditional approach of using transportation infrastructure as a control variable, this study includes various transportation infrastructure, such as highways, common railways, and high-speed railways, in empirical testing to identify their strengthening role in the relationship between ETDZs and TFP, reflecting the synergistic effect of transportation infrastructure and industrial policies. Third, to explore the heterogeneity differences of ETDZs in different types of regions and cities, this study systematically analyzes the heterogeneity effects of ETDZ policies under different development and governance backgrounds from two dimensions: geographical regions and administrative levels. These findings provide empirical support for improving classification policies and regional coordinated development strategies.

2. Theoretical Analysis and Hypotheses Development

China’s state-level ETDZs serve as pioneering platforms for institutional innovation. Through a series of industrial support policies, including tax relief, preferential land use arrangements, and infrastructure subsidies, state-level ETDZs significantly reduce the fixed costs of enterprise operations, ease budget constraints, and enable firms to reallocate capital toward productive investment. This reallocation promotes improvements in regional TFP [29,30]. Additionally, by alleviating financing constraints and enhancing firms’ capacity to manage risk, state-level ETDZs facilitate enterprise expansion and production upgrading. As critical nodes within regional development strategies, state-level ETDZs often receive substantial infrastructure investment from local governments aimed at enhancing operational efficiency and fostering economies of scale. The resulting business environment not only attracts more firms but also encourages industry specialization and agglomeration. By clustering related industries, state-level ETDZs support the formation of localized labor markets, enhance the sharing of intermediate inputs, and enable firms to engage in vertical specialization and horizontal scaling, together contributing to increasing returns to scale and enhanced productivity and thus achieving the improvement of regional TFP [13,31]. From the perspective of regional development theory, the mechanisms observed in state-level ETDZs are consistent with the principles of endogenous growth theory and new economic geography. According to endogenous growth theory, sustained productivity improvements and innovation are largely driven by localized knowledge spillovers, human capital accumulation, and R&D investment, all of which are fostered within the institutional and physical frameworks of state-level ETDZs. Meanwhile, new economic geography emphasizes the role of agglomeration economies in shaping regional disparities and fostering cumulative causation. State-level ETDZs concentrate economic activity through preferential resource allocation, infrastructure investment, and spatial clustering, thereby creating self-reinforcing cycles of growth. Through the promotion of inter-firm collaboration, labor market specialization, and supply chain optimization, ETDZs play a critical role in enhancing regional TFP.

Moreover, the preferential conditions in state-level ETDZs, such as tax incentives, land allocations, and funding support, reduce the risks of entrepreneurial activities and enhance firms’ capacity for innovation. These policies not only stimulate entrepreneurial activity and the formation of new businesses but also support the expansion of existing firms, enabling them to scale operations, invest in R&D, and pursue technological innovation [32,33,34]. The dense concentration of enterprises within the zones also encourages frequent technological exchanges, reinforcing knowledge spillovers and productivity gains [35]. In addition, the intensified competition within state-level ETDZs induces a market selection process whereby less productive firms are driven out, enhancing the overall efficiency of regional resource allocation. This competitive dynamic promotes the entry of high-productivity firms and reallocates scarce policy and capital resources toward more capable enterprises, thereby enhancing regional TFP [36,37,38,39]. Based on the above analysis, the following hypothesis is proposed:

Hypothesis 1: 

The establishment of state-level ETDZs significantly improves regional TFP.

Although state-level ETDZ regulations have been found to be effective in reducing operational costs for enterprises, the spatial effectiveness of such zones is closely linked to the degree of transportation infrastructure development. Transportation costs represent a critical component of firms’ total trade costs, including those associated with acquiring capital goods, accessing factor markets, and distributing products to end consumers. When transportation costs become sufficiently high, they can offset the agglomeration economies generated within state-level ETDZs, thereby reducing the zones’ attractiveness to firms. In this context, the effectiveness of state-level ETDZs in improving regional TFP is not solely a function of internal industrial policies but is also contingent upon the accessibility and efficiency of external transportation networks. Insights from transportation economics emphasize that improvements in transportation infrastructure reduce not only direct transportation costs but also a range of indirect costs, including information costs, transaction costs, logistics risks, and supply chain disruptions [40,41,42,43]. A well-functioning transportation system enhances firms’ ability to connect with upstream suppliers and downstream markets, stabilize procurement channels, and reduce inventory and warehousing needs. These mechanisms help lower overall production costs and increase productivity by enabling firms to engage in a broader and more competitive factor market, selecting inputs of higher quality and better price [44,45].

Moreover, transportation infrastructure can play a pivotal role in expanding the effective market size of firms. By reducing the cost of factor mobility, transportation improvements facilitate the reallocation of resources across space, thereby enhancing regional labor market integration and capital mobility [46,47]. These dynamics contribute to a more efficient spatial distribution of economic activity and support interregional specialization, aligning with the foundational principles of classical transportation economics and the spatial equilibrium model. The enlargement of market access also fosters regional integration and facilitates higher levels of interregional division of labor, enhancing the dynamic efficiency of the broader economic system. Therefore, the degree of transportation infrastructure development not only affects the direct operational costs of firms but also mediates the realization of policy-induced regional TFP gains from ETDZs. Well-developed transportation systems enhance the spatial spillover effects of ETDZs, reinforcing their capacity to attract productive firms, stimulate resource reallocation, and promote technological diffusion, ultimately achieving an improvement in regional TFP. Based on previous theoretical findings and research results, the following hypothesis is proposed:

Hypothesis 2: 

Transportation infrastructure positively moderates the relationship between the establishment of state-level ETDZs and regional TFP.

3. Materials and Methods

3.1. Study Area

This study centers its attention on ETDZs that house a substantial quantity of manufacturing businesses. These state-level ETDZs also boast the most extensive geographical reach within China, as indicated by the Catalog of China’s Development Zones Audit Bulletin (2018 Edition) [48]. In contrast, it is challenging to draw a comprehensive conclusion regarding the impact of geographical characteristics on the regional economy, specifically in the case of bonded zones, pilot free trade zones, and national tourism resort zones. This study primarily focuses on the empirical analysis of ETDZs due to their more favorable industrial policies compared to provincial-level development zones. The ETDZs exhibit greater policy standardization, stronger factor agglomeration ability, a more significant impact on the economy, and possess a unique conceptual framework. These characteristics contribute to the enhanced reliability of the results obtained from econometric analyses. The distribution of cities with state-level ETDZs at the end of 2020 is shown in Figure 1.

3.2. Model Specification

In order to assess the influence of the implementation of ETDZs on regional TFP, it is important to conduct an analysis of the outcomes both prior to and after to the development of ETDZs. In line with the majority of research conducted on development zones, this study regards the creation of ETDZs as a quasi-natural experiment. In contrast with the conventional difference-in-differences modeling approach, where the policy shock is assumed to occur in a single year, this study examines the establishment of ETDZs in several cities at different points in time. Consequently, we employ the Time-varying Difference-in-Differences (DID) method, commonly known as the “progressive” Difference-in-Differences method. The focus of this study pertains to prefecture-level cities across the nation. These cities are categorized into two groups: the treatment group, which includes cities with ETDZs and is denoted by the variable “treat” set as “1”, and the control group, which comprises cities without ETDZs and is denoted by the variable “treat” set as “0”. Simultaneously, an assessment is conducted by distinguishing between two periods in the region based on the establishment of ETDZs: the period prior to their establishment (referred to as “period 0”) and the period subsequent to their establishment (referred to as “period 1”). In terms of the temporal distribution of ETDZ establishment, there is a relatively even distribution across months each year. Consequently, in order to assess the impact of ETDZs, this study adopts a current-year approach for cities that establish ETDZs in the first half of the year and a one-year lag approach for cities that establish ETDZs in the second half of the year. As an illustration, in the event that a certain area initially implemented a development zone at the state level in March 2010, the temporal designation for the city from 2010 onwards is assigned a value of 1, while the preceding temporal designation is assigned a value of 0. The establishment of the initial state-level development zone in October 2010 marked the beginning of a new era for the city. Subsequently, the period from 2011 onwards is denoted as period 1, while the preceding period is designated as period 0. The estimated model of the impact of ETDZs on regional TFP set up in this study is as follows:

$$TF{P}_{it}={\alpha }_0+{\alpha }_{1}trea{t}_i\times perio{d}_t+{\displaystyle {\sum }_j{\alpha }_{j}contro{l}_{jit}+{\mu }_i+{\gamma }_t+{\epsilon }_{it}}$$

(1)

The explanatory variable TFP represents regional TFP. The variable “treat” is a binary variable, where a value of 1 indicates that the city has established a state-level ETDZ, and a value of 0 indicates that the city has not established such a zone. The variable “period” is also a binary variable, where a value of 1 indicates that a state-level ETDZ has been established in period t, and a value of 0 indicates that no state-level ETDZs have been established in period t. The coefficient α₁ of the interaction term “treat × period” quantifies the impact of establishing state-level ETDZs. The word “control_jit” represents the control variables in the model. The variable “i” represents the region, “t” represents the year. The term “μ_i” represents the fixed effects for each area, “γ_t” represents the fixed effects for each year, and “ε_it” represents the disturbance term in the model, which is assumed to follow an independent homogeneous distribution.

In comparison, the use of the DID research method requires the treatment and control groups to follow strictly parallel time trends; otherwise, it will make the results of the study highly biased, and the use of the combination of Propensity Score Matching (PSM) and DID has the potential to reduce this bias to some extent. Based on the propensity value causal inference analysis method of Imbens [49], among the many macro variables including regional gross domestic product (GDP), the correlation variables with a more desirable degree of matching are finally screened out for the trend test, including regional GDP, the share of regional tertiary industry in GDP, the share of regional primary industry in GDP, the ratio of fiscal expenditure to GDP, the logarithm of the regional population number, the share of foreign investment in GDP, and many other variables. In terms of matching methods, this method of kernel matching is superior compared with these methods of nearest neighbor matching and radius matching, so this study uses the Kernel Propensity Score Difference-in-Differences method for estimation. The estimation steps of this method are as follows: use the kernel density equation to estimate the weights of the selected matching variables and then estimate the DID based on the weights estimated by the said propensity score. The specific estimation equations are as follows.

$$TF{P}_{it}\times weight{s}_{it}={\beta }_0+{\beta }_{1}trea{t}_i\times perio{d}_t+{\displaystyle {\sum }_j{\beta }_{j}contro{l}_{jit}+{\mu }_i+{\gamma }_t+{\epsilon }_{it}}$$

(2)

The weights used in this study are derived from an estimation equation based on density. The coefficient β₀ corresponds to this equation. The term μ_i represents the fixed effect for each region, while γ_t represents the fixed effect for each year. The perturbation term ε_it is assumed to follow an independent homogeneous distribution in the model. Finally, β₁ is the coefficient of interest in the DID regression analysis conducted in this study.

The modeling of the moderating effect is similar to the above, also using the DID approach, with the difference being the addition of an interaction term to validate the moderating effect treat_i × period_t × M_it, where M is the moderating variable. The moderating effect mechanism test model is shown below:

$$TF{P}_{it}={\rho }_0+{\rho }_{1}trea{t}_i\times perio{d}_t+{\rho }_{2}trea{t}_i\times perio{d}_t\times M_{it}+{\displaystyle {\sum }_j{\rho }_{j}contro{l}_{jit}+{\mu }_i+{\gamma }_t+{\epsilon }_{it}}$$

(3)

Application of PSM: The PSM-DID method can more accurately estimate policy effects and reduce endogeneity issues caused by sample self-selection bias by comparing and analyzing matched samples [50]. The validity of the PSM-DID findings is contingent upon the presence of dissimilarities in the matching variables between the treatment and control groups subsequent to the matching process. If a notable disparity exists, it suggests that the selection of matching variables and the matching method was inadequate, rendering the estimation results of the DID approach invalid. Conversely, if no significant difference is observed, it indicates that the treatment and control groups have a similar probability of being chosen as policy pilots, thereby validating the matching variables and matching method employed. In order to ascertain the accuracy and reliability of the estimation outcomes for this PSM-DID analysis, a smoothness test was performed. The PSM-DID approach can be employed to choose city j from the set of non-pilot cities. This is achieved by matching the weights of the estimator in order to minimize the dissimilarity between the observable variables of city j and the pilot city i. Specifically, it is imperative to ensure that the pilot regions and their corresponding non-pilot regions exhibit similar trends in their observable variables to the greatest extent possible.

3.3. Evaluation of Regional TFP

This study uses the DEA-Malmquist index approach established by Färe et al. to measure the change in TFP, which has the advantage of being able to decompose the change in TFP, making it easier for us to examine the source of the change in TFP [51]. For the treatment of missing values, some cities lack samples for only one or two years, and we obtained data from the statistics bulletin provided by the city in that year to fill in the gap. The data that are still missing are treated in two groups; for data in 2019 and earlier, this research employs linear regression point estimates to fill in the missing values. For data in 2020, we use the average growth rate of the last five years to generalize backwards.

Input variables. The classical Cobb–Dauglas production function suggests that the most basic input factors in an economic production system consist mainly of two factors of production, capital and labor. The input labor force is expressed in terms of the number of people employed at the end of the period in society as a whole (10,000 people). The capital factor is expressed by the stock of fixed assets. Referring to the method of Zhang et al., this study uses the perpetual inventory method for calculations [52]. Considering the influence of inflationary factors, this study selects the fixed asset price index for deflating the total fixed asset investment to obtain the comparable total fixed asset formation for the period, but the fixed asset price index is missing more at the prefecture-level city level, which is replaced by the Consumer Price Index, and the individual residual values are replaced by the Consumer Price Index of the province where the city is located. The study uses 1998 data as a benchmark for estimation.

Output variables. Output is chosen to be expressed by GDP, and since the statistics of this indicator use monetary units, in order to eliminate the influence of inflation, referring to the method of Wu & Yu [53] and Li et al. [54], this study uses 1998 as the base period and uses the constant price index of consumption in each province to deflate it. Considering the lack of city-level GDP deflators, this study utilizes the GDP deflator of the province where the city is located to deflate the city’s GDP at 1998 constant prices. The relevant input–output indicators are shown in

Table 1. Input–output indicators for urban TFP.

Classification	Variables	Specific Indicators	Unit
Input indicators	Labor	Number of employees	10,000 persons
	Capital	Fixed assets	100 million CNY
Output indicators	Capital	GDP	100 million CNY

.

Based on the study period of 1999–2020, this study excludes the cities that have changed during the study period and those with serious levels of missing statistics, leaving 282 cities for the remaining study.

3.4. Data

Data sources: The data in this study span from 1999 to 2020, and are mainly composed of three parts: (1) regional macro data, including regional GDP, year-end resident population, regional fiscal revenue, and regional proportion of three industries, etc., which are obtained from the National Bureau of Statistics, provincial and municipal bureaus of statistics, and the corresponding year’s China Urban Statistical Yearbook; (2) data on state-level ETDZs, which are obtained according to the “2018 edition of the Catalog of China’s Development Zones Audit and Announcement” [48] and the official website of the China Association of Development Zones; and (3) regional TFP data, based on the DEA-Malmquist index method to measure the change in regional TFP.

This study requires further elucidation regarding the endogeneity of state-level development zone establishment and its impact on economic development during the data processing phase. Indeed, there exists an inherent correlation between the establishment of state-level ETDZs and the economic advancement of the respective regions. However, subsequent to the introduction of the western development strategy by the state in 1999, the scope of state-level ETDZs has expanded beyond cities with already elevated levels of development. Consequently, these zones have evolved into instruments for fostering regional economic and social progress. Consequently, this study deems it appropriate to disregard the aforementioned endogenous issue.

Selection of variables: Explained variable: Regional TFP. Regional TFP in this study refers to the growth of output due to technological progress and capability realization, etc., excluding factor inputs such as capital and labor, within the scope of prefecture-level and above cities, using the results of regional TFP measurements in Table 1.

Core explanatory variable: establishment of state-level ETDZs (treat). The policy variables of DID are generally represented by dummy variables, which take the value of 1 if city i has a state-level ETDZ in year t; otherwise, they take the value of 0. The data of state-level ETDZs studied in this study are based on the Catalog of China’s Development Zones Audit and Bulletin (2018 Edition). An additional note on data organization is needed: due to the long time span in this study, some prefecture-level cities changed their administrative divisions during this period due to the act of splitting or merging (e.g., Laiwu City, Shandong Province, was merged with Jinan City in 2019, and Chaohu City, Anhui Province, was assigned to Hefei, Wuhu, and Ma’anshan, respectively, in 2011). In order to avoid biased results caused by administrative division changes, this study will exclude prefecture-level cities that involved administrative division changes during this period. China’s state-level ETDZs were first set up in 1984, but due to the large area of missing statistics of prefectural-level cities before 1998, and the obvious advantages of state-level ETDZs set up before 1998 in terms of administrative level or geographic location, the stochastic nature of the policy intervention is significantly weaker than that of state-level ETDZs set up after 1998. Therefore, the starting year of the panel data in this study is set as 1999. Since the macroeconomic data for 2021 have not been published as of the completion of this study, the panel data in this study are thus as of 2020. Due to the limitation of the measurement method, the samples of cities that first set up state-level ETDZs in 2019 and later cannot be objectively evaluated, so the samples of state-level ETDZs actually studied in this study are included in the Catalog of China’s Development Zones Audit Bulletin (2018 Edition). Furthermore, it should be noted that the Jiuquan Economic and Technological Development Zone and Shizuishan Economic and Technological Development Zone were slated for removal from the list of state-level ETDZs in 2020 and 2021, respectively. However, their exclusion from the sequence does not impede the empirical investigation conducted in this study. As they are still considered state-level ETDZs within the scope of this study, their removal after 2020 does not affect their classification.

Controlled variables: Drawing upon extant research [55,56,57,58], six control variables were selected to eliminate the influence of other factors on regional TFP growth. (1) The level of the city’s economy (lngdp) is measured using the logarithm of the city’s real GDP per capita. The level of the city’s economy directly affects technological investment capacity, industrial upgrading, and resource allocation efficiency, all of which influence TFP. (2) Foreign direct investment (fdi) is measured as the ratio of actual utilization of foreign capital to regional GDP for the year. Foreign direct investment introduces advanced technologies, management experience, and promotes knowledge spillovers, thereby affecting regional TFP. (3) The level of government financial input (fin) is measured by the ratio of government financial expenditure to regional GDP. The level of government financial input supports infrastructure, education, and R&D investment, which can enhance production efficiency and TFP. (4) Population density (lnhum) is expressed as the logarithm of the ratio of resident population to administrative area at the end of the year in the region. Higher population density fosters labor market pooling, knowledge spillovers, and agglomeration economies, all contributing to TFP growth. (5) The industrial structure (pi, ti) is expressed in terms of the primary sector’s share of GDP and the tertiary sector’s share of GDP, respectively. Different industrial structures lead to varying productivity levels and resource allocation patterns, thus affecting TFP. (6) The level of scientific and technological development (edu) is measured by the number of students enrolled in general higher education per 10,000 population. A higher level of human capital and innovation capability improves firms’ production efficiency and drives TFP improvement. These control variables comprehensively capture key economic, demographic, structural, and innovation-related factors that may influence regional TFP, ensuring the robustness of the empirical analysis.

Moderating variable: The construction of transportation infrastructure is the moderating variable (M) in this study, specifically divided by the city where the state-level ETDZs is located whether there are highways (hw), common railways (rw), high-speed railways (hsr), inland river ports (rp), coastal ports (cp), and civilian airports (ap). The descriptive analysis of the data is shown in

Table 2. Descriptive analysis of data.

Variable	Mean	Std. Dev	Min	Max
TFP	0.8846	0.9703	0.0085	16.1287
treat	0.4787	0.4996	0.0000	1.0000
lngdp	15.8228	1.2152	12.0368	19.7740
fdi	0.0205	0.0310	0.0000	1.1937
fin	0.1557	0.1220	0.0028	0.9988
lnhum	5.8519	0.6966	2.6776	8.1362
pi	0.0241	0.0092	−0.0351	0.0660
ti	0.3894	0.9820	0.8500	0.8534
edu	0.0142	0.0205	0.0000	0.1311

.

4. Results

4.1. Analysis of Regional TFP

Regional TFP based on the Malmquist index is a dynamic measurement result, and the decomposition of the index can analyze the changes in regional TFP, as well as explore the impact of technology level and innovation on the change in TFP.

The Malmquist index is a metric used to assess the alteration in TFP between two consecutive time periods. A Malmquist index exceeding 1 signifies an increase in TFP during the corresponding period, while a value below 1 indicates a decrease. Conversely, an index of 1 denotes that TFP remains unaltered.

Table 3. ML index results (1999–2020).

Year	effch	techch	pech	sech	tfpch
1999	1.155	0.971	1.011	1.143	1.122
2000	1.084	0.962	1.1	0.985	1.042
2001	0.965	1.058	0.951	1.016	1.022
2002	1.04	0.952	1.071	0.971	0.99
2003	1.034	0.93	1	1.034	0.961
2004	1.224	0.806	1.147	1.067	0.986
2005	0.692	1.348	0.846	0.818	0.933
2006	1.433	0.632	1.201	1.194	0.907
2007	1.001	0.912	0.962	1.041	0.913
2008	0.957	0.983	1	0.957	0.941
2009	0.16	5.198	0.445	0.359	0.831
2010	6.109	0.163	2.17	2.815	0.993
2011	0.913	1.021	0.935	0.977	0.933
2012	0.987	0.892	0.996	0.991	0.881
2013	1.015	0.827	1.088	0.933	0.84
2014	0.987	0.909	0.952	1.036	0.897
2015	1.17	0.769	1.116	1.048	0.9
2016	1.05	0.888	1.019	1.03	0.932
2017	0.528	0.812	0.665	0.793	0.429
2018	1.81	1.256	1.451	1.247	2.273
2019	1.028	0.934	0.993	1.035	0.96
2020	0.977	1.013	0.991	0.986	0.99
Mean	1.241773	1.101636	1.050455	1.067091	0.985273

provides a decomposition of regional TFP in various regions. This analysis is conducted using the DEA-ML model measurements for 283 cities in China from 1999 to 2020. The decomposition includes factors such as change in technical efficiency (effch), change in technological progress (techch), change in pure technical efficiency (pech), change in scale efficiency (sech), and change in TFP index (tfpch). The data presented in Table 3 represent the mean values of national cities from 1999 onwards. These values serve as indicators for assessing the overall changes in technical efficiency, technological progress, pure technical efficiency, scale efficiency, and domestic regional TFP level over the specified time period. From 1999 to 2020, China’s regional TFP exhibited a general decline. The average value of the ML index measurement result was 0.985. Technical efficiency, technological progress, pure technical efficiency, and scale efficiency exhibit a modest overall rising trend, with a mean value exceeding 1.

Since the sought tfpch reflects the change in TFP, in order to analyze the influencing factors to facilitate the establishment of the regression relationship between the influencing factors and TFP, referring to the studies by Cheng & Lu [59] and Li et al. [60], in this study, we applied such a process, i.e., assuming that the TFP of the base period is 1, and if the year 1999 is the base period, the TFP of 2000 is equal to the Malmquist index of 2000 multiplied by the TFP of 1999, the other years, and so on to perform the cumulative multiplication. The process of cumulative multiplication is applied to subsequent years, as well as to technological progress (techch) and the change in technical efficiency (effch) [59,60]. The newly obtained variables are labeled as technological progress (tech) and technical efficiency (eff). The findings are depicted in Figure 2.

TFP encompasses three fundamental components, namely capital, labor, and technological progress. During the initial three decades following the implementation of reform and development policies, China benefitted from an abundant labor supply, which significantly propelled its economy forward. The regional TFP in China has experienced a consistent fall over time, primarily attributed to the influence of economic structural adjustment. During the time of structural adjustment, there was a notable increase in household spending among residents in the sectors of education and medical care, whereas expenditures on apparel, food, and household appliances experienced somewhat slower growth. The economy has reached a stage where capital accumulation has been achieved, leading to a structure predominantly characterized by the service sector. This sector operates in a manner that does not necessitate extensive inputs or an excess of capital. In 2017, China’s service sector accounted for more than 50% of the economy. As China transitions toward a post-industrial society, the growth in TFP within the service sector has been sluggish, consequently impacting the overall TFP of the entire economy. The manufacturing sector has witnessed the accumulation of numerous capitals in its early stages, as well as in the past. However, it is not feasible to continue utilizing all of these accumulated capitals in the present day. Despite attempts to incorporate this factor into regional TFP accounting through the utilization of traditional methods, the impact of this factor has not been successfully mitigated. The observed decrease in technical efficiency and TFP does not necessarily imply a decrease in the efficiency of capital use. Rather, it suggests that a significant portion of fixed capital remains unutilized and is nevertheless included in the calculation of fixed capital stock. The reason for the sudden increase and decrease in technical efficiency and technical progress of TFP in our region in 2009 is mainly due to the fact that after the financial crisis in 2008, China responded to the crisis by means of short-term expansion of government investment (especially in the area of infrastructure construction), which led to a sharp increase in the amount of factor inputs in 2009, whereas the effectiveness of infrastructure construction usually takes a few years to take effect and cannot be quickly reflected in 2009 output. To mitigate the potential influence of outliers in 2009 on the upcoming research conducted in this study, and considering the absence of any newly constituted state-level ETDZs during that year, the regression analysis presented below removes the data from 2009.

4.2. Analysis of PSM-DID Benchmark Regression Results

The results of the impact of state-level ETDZs on regional TFP from 1999 to 2020 are shown in

Table 4. Empirical results of the influence of ETDZs on regional TFP.

Model	(1)	(2)	(3)	(4)
DID (treat × period)	0.265 **	0.112 ***	0.545 ***	0.023 ***
	(0.098)	(0.044)	(0.050)	(0.080)
lngdp				0.097 ***
				(0.020)
fdi				0.058 **
				(0.033)
lnhum				−0.034 *
				(0.026)
fin				0.085 ***
				(0.040)
edu				0.064 *
				(0.055)
pi				−0.120 ***
				(0.047)
ti				0.248 ***
				(0.058)
Time fixed effect	No	No	Yes	Yes
Individual fixed effect	No	Yes	Yes	Yes
Observations	5911	5911	5911	5911
Adjusted-R2	0.419	0.528	0.562	0.644

Note: The data are calculated using Stata 15.0. ***, **, and * refer to 1%, 5%, and 10% significance levels, respectively. Numbers in parentheses represent t values of bilateral tests.

, where DID is the double-difference result before and after the establishment of state-level ETDZs. From Table 4, it can be seen that with the establishment of state-level ETDZs, regional TFP has a positive growth, and the results of the model before and after the addition of time, individual fixed effects and control variables are all significant, which indicates that the establishment of state-level ETDZs has a certain positive impact on the enhancement of regional TFP. Hypothesis 1 is validated.

The parallel trend test is employed as a crucial assumption in the DID model. To mitigate the potential bias in measurement results arising from unsatisfactory parallel trends, this study employs a dynamic effect test to perform an assessment. The first phase before pilot implementation is set as the base period. The findings are presented in Figure 3. Prior to the implementation of the state-level ETDZs, the treatment coefficient does not exhibit a statistically significant departure from zero, suggesting a lack of discernible distinction between the treatment and control groups. However, subsequent to the establishment of the state-level ETDZs, the treatment effect coefficient becomes positive and significantly non-zero in all years, with the exception of the third year following the implementation of the state-level ETDZs. In addition, it is worth noting that the treatment effect exhibits a progressive upward trend over time, suggesting that the impact of state-level ETDZs on regional TFP is gradually realized. The contribution of state-level ETDZs to regional TFP increased each year after their inception but began a declining trend in the sixth year after their establishment.

Sensitivity test: However, the latest studies argue that parallel trend testing before processing is not effective for DID methods and may have serious issues. To address this issue, some scholars have proposed methods that allow for non-parallel pre-treatment trends, assessing the sensitivity of the estimated processing effect to violations of the parallel trend assumption by imposing bounds on the degree of relative deviation [61,62]. Following the approach of Rambachan & Roth [62], this study sets the maximum deviation degree (Mbar) equal to 1 multiplied by the standard error. Figure 4 shows the results of the sensitivity test. It can be seen that under the constraint of relative deviation degree, the estimated coefficients do not include 0 within the 90% confidence interval, indicating that the promotion effect of ETDZ policies on regional TFP is very robust. This result also indicates that although there is a certain degree of deviation from the parallel trend assumption, it can still be concluded that ETDZ policies have a significant treatment effect on regional TFP.

Stabilization test: In the previous study, there are still controversies about the calculation and classification criteria of some indicators, mainly in the following aspects: whether the treatment of the time of establishment of the state-level ETDZs is reasonable; whether the method of measuring the regional TFP of the explanatory variables is reasonable and whether the results are reliable; and whether the establishment of the subsequent ETDZs of some cities with multiple ETDZs will affect the policy effect of the first ETDZ that is the focus of this study. These controversies may also affect the final results, so in this section, we conduct a robustness check on these controversies.

In the robustness test of the time of establishment of state-level ETDZs, this study screens out the cities that established their first state-level ETDZs in 2010, and the sample size decreases from the original 282 prefectural-level cities to 94 prefectural-level cities, and the analysis is conducted using the traditional DID method on the grounds that 2010 is the year with the highest number of state-level ETDZs in the observation period, and the results of the test are shown in Model (5) in

Table 5. Stabilization test.

Model	(5)	(6)	(7)
DID (treat × period)	0.104 *	0.028 ***	0.035 ***
	(0.047)	(0.022)	(0.028)
lngdp	0.124 ***	0.110 ***	0.098 ***
	(0.019)	(0.021)	(0.021)
fdi	0.076 **	0.068 **	0.060 **
	(0.030)	(0.031)	(0.032)
lnhum	−0.058 *	−0.046 *	−0.038 *
	(0.022)	(0.023)	(0.024)
fin	0.110 ***	0.096 ***	0.085 ***
	(0.036)	(0.037)	(0.038)
edu	0.089 *	0.078 *	0.065 *
	(0.048)	(0.049)	(0.050)
pi	−0.176 ***	−0.151 ***	−0.135 ***
	(0.042)	(0.043)	(0.045)
ti	0.290 ***	0.265 ***	0.240 ***
	(0.053)	(0.054)	(0.055)
Time fixed effect	Yes	Yes	Yes
Individual fixed effect	Yes	Yes	Yes
Observations	1974	5911	3801
Adjusted-R2	0.743	0.675	0.697

Note: The data are calculated using Stata 15.0. ***, **, and * refer to 1%, 5%, and 10% significance levels, respectively. Numbers in parentheses represent t values of bilateral tests.

. In the robustness test of replacing the explanatory variables, this study uses technological progress (tech) as a new explanatory variable, the reason being that the promotion of technological progress in ETDZs is one of the most important ways to enhance TFP; the test results are shown in model (6) in Table 5. In order to exclude the interference of the sample of cities with multiple state-level ETDZs on the results of the study, the sample of such cities is excluded in the robustness test, and the sample size decreases from 282 to 181 prefecture-level cities, and the test results are shown in model (7) in Table 5. The results of these three types of robustness tests show that, except for model (5) whose significance has decreased compared to the previous section but is still significant at the 10% level, both the sign and significance of the regression coefficients in the remaining robustness tests are consistent with the results presented in Table 4; thus, the above findings have relatively high reliability.

Placebo test: In order to further rule out the influence of other unknown factors on regional TFP and to ensure that the findings obtained in the previous section were indeed caused by the establishment of the state-level ETDZs, a placebo test was therefore required. Since the premise of using the DID method is that the experimental and control groups are comparable, i.e., without the state-level ETDZs, there would be no significant difference between the regional TFP of the experimental and control groups over time. To verify this premise, a placebo test was conducted using the counterfactual analysis method. 2015 and 2018, which have the lowest number of state-level ETDZs set up, are taken as the hypothetical time of state-level ETDZs set up to conduct a consistent test with the baseline regression to derive the mean treatment effect coefficients. The results are shown in

Table 6. Placebo test.

Model	(8)	(9)
Hypothetical processing point	2005	2018
DID (treat × period)	0.066	0.064
	(0.053)	(0.071)
lngdp	0.137 ***	0.120 ***
	(0.021)	(0.023)
fdi	0.076 **	0.071 **
	(0.031)	(0.033)
lnhum	−0.060 *	−0.052 *
	(0.024)	(0.026)
fin	0.115 ***	0.098 ***
	(0.037)	(0.040)
edu	0.089 *	0.078 *
	(0.048)	(0.050)
pi	−0.182 ***	−0.160 ***
	(0.043)	(0.045)
ti	0.295 ***	0.268 ***
	(0.055)	(0.057)
Time fixed effect	Yes	Yes
Individual fixed effect	Yes	Yes
Observations	5911	5911
Adjusted-R2	0.743	0.697

Note: The data are calculated using Stata 15.0. ***, **, and * refer to 1%, 5%, and 10% significance levels, respectively. Numbers in parentheses represent t values of bilateral tests.

, which shows that when setting the time of state-level ETDZs set up as 2005 or 2018, the coefficients of the interaction term DID are not significant, indicating that the actual year of establishment of state-level ETDZs can indeed significantly enhance regional TFP.

4.3. Analysis of the Results of the Moderating Effect

Moderating effect results: Since different modes of transportation infrastructure may have different impacts on the role of the ETDZs, in order to further analyze the differences in the moderating effects of transportation infrastructure, whether or not the city where ETDZs is located opens highways (hw), common railways (rw), high-speed railways (hsr), coastal ports (cp), inland river ports (rp), and civilian airports (ap) (all of the above are variables are set from 0 to 1, with the open ones set to 1 and the unopened ones set to 0, and they are used as moderating variables). The empirical results are shown in

Table 7. Moderating effects of transportation infrastructure.

Model	(10)	(11)	(12)	(13)	(14)	(15)
Type of moderator	hw	rw	hsr	cp	rp	ap
DID (treat × period × M)	0.420 ***	0.366 ***	0.167 **	−0.196	0.039	−0.258
	(0.096)	(0.077)	(0.052)	(0.180)	(0.079)	(0.070)
M	−2.9 ***	−2.4 ***	−3.3 ***	−2.8 ***	−2.9 ***	−2.6 ***
	(0.715)	(0.702)	(0.771)	(0.716)	(0.713)	(0.700)
lngdp	0.132 ***	0.121 ***	0.105 ***	0.093 ***	0.087 ***	0.098 ***
	(0.021)	((0.020)	(0.019)	(0.021)	(0.022)	(0.020)
fdi	0.082 **	0.076 **	0.069 **	0.058 **	0.061 **	0.060 **
	(0.031)	(0.032)	(0.033)	(0.031)	(0.030)	(0.031)
lnhum	−0.055 *	−0.048 *	−0.042 *	−0.037 *	−0.041 *	−0.045 *
	(0.023)	(0.024)	(0.025)	(0.024)	(0.023)	(0.024)
fin	0.118 ***	0.104 ***	0.091 ***	0.085 ***	0.091 ***	0.089 ***
	(0.037)	(0.039)	(0.040)	(0.039)	(0.038)	(0.038)
edu	0.084 *	0.079 *	0.068 *	0.062 *	0.065 *	0.060 *
	(0.050)	(0.051)	(0.052)	(0.053)	(0.051)	(0.050)
pi	−0.182 ***	−0.159 ***	−0.141 ***	−0.122 ***	−0.135 ***	−0.138 ***
	(0.044)	(0.045)	(0.046)	(0.047)	(0.046)	(0.045)
ti	0.301 ***	0.274 ***	0.249 ***	0.220 ***	0.234 ***	0.246 ***
	(0.053)	(0.054)	(0.055)	(0.056)	(0.055)	(0.054)
Time fixed effect	Yes	Yes	Yes	Yes	Yes	Yes
Individual fixed effect	Yes	Yes	Yes	Yes	Yes	Yes
Observations	5911	5911	5911	5911	5911	5911
Adjusted-R2	0.669	0.669	0.668	0.668	0.667	0.668

Note: The data are calculated using Stata 15.0. ***, **, and * refer to 1%, 5%, and 10% significance levels, respectively. Numbers in parentheses represent t values of bilateral tests.

from model (10) to model (15). It can be seen that the sign and significance of the coefficients of the interaction terms between the establishment of ETDZs and the transportation infrastructure are unstable, indicating that different modes of transportation infrastructure have different impacts on the role of the ETDZs. Among them, highways, common railways, and high-speed railways show significantly positive moderating effects in Models (10), (11), and (12), implying that land-based transportation infrastructure is more closely aligned with the operational needs and logistical functions of ETDZs. Highways and common railways provide essential freight capacity and regional connectivity, which directly support industrial output and supply chain operations in ETDZs. In model (12), the significance of the coefficient of the interaction term between high-speed railways and ETDZs is slightly weaker, probably because China’s high-speed railways network primarily serves passengers rather than freight, limiting its relevance for freight-intensive industrial activities typically concentrated in ETDZs.

In contrast, the moderating effects of ports and airports are not significant in models (13), (14), and (15). This may be due to the fact that these types of transportation infrastructure often serve broader regional and even national functions, rather than providing immediate productivity gains at the city level. Furthermore, ETDZs located in cities without direct access to a port or airport may still benefit from interconnected land-based transportation infrastructures that link them efficiently to surrounding cities with such infrastructure. Thus, proximity, freight relevance, and integration with industrial value chains emerge as key mechanisms that explain the differential moderating roles of various transportation infrastructures. Hypothesis 2 is validated.

Heterogeneity analysis:

Table 8. Subgroup regression results.

Model	(16)	(17)	(18)	(19)	(20)	(21)	(22)	(23)	(24)
Subgroup	With hw	Without hw	With rw	Without rw	With cp	Without cp	With rp	Without rp	Adjusted
DID (treat × period)	−0.185 **	0.340 ***	−0.098 *	0.291 ***	0.393 ***	0.814 ***	0.391 ***	0.493 ***	0.371 ***
	(0.066)	(0.047)	(0.031)	(0.035)	(0.049)	(0.218)	(0.062)	(0.063)	(0.046)
lngdp	0.118 ***	0.107 ***	0.089 ***	0.091 ***	0.094 ***	0.118 ***	0.087 ***	0.093 ***	0.094 ***
	(0.020)	(0.021)	(0.021)	(0.022)	(0.020)	(0.020)	(0.021)	(0.021)	(0.020)
fdi	0.069 **	0.062 **	0.060 **	0.062 **	0.067 **	0.069 **	0.062 **	0.060 **	0.066 **
	(0.031)	(0.032)	(0.032)	(0.033)	(0.031)	(0.031)	(0.032)	(0.032)	(0.032)
lnhum	−0.063 *	−0.051 *	−0.045 *	−0.048 *	−0.053 *	−0.065 *	−0.050 *	−0.056 *	−0.052 *
	(0.026)	(0.027)	(0.028)	(0.029)	(0.026)	(0.026)	(0.027)	(0.028)	(0.026)
fin	0.103 ***	0.091 ***	0.084 ***	0.087 ***	0.092 ***	0.112 ***	0.089 ***	0.098 ***	0.094 ***
	(0.035)	(0.036)	(0.037)	(0.037)	(0.035)	(0.035)	(0.036)	(0.037)	(0.035)
edu	0.076 *	0.065 *	0.061 *	0.069 *	0.074 *	0.080 *	0.068 *	0.072 *	0.071 *
	(0.048)	(0.050)	(0.051)	(0.052)	(0.049)	(0.048)	(0.050)	(0.051)	(0.049)
pi	−0.158 ***	−0.142 ***	−0.128 ***	−0.133 ***	−0.136 ***	−0.159 ***	−0.141 ***	−0.145 ***	−0.139 ***
	(0.044)	(0.045)	(0.046)	(0.047)	(0.043)	(0.044)	(0.045)	(0.046)	(0.043)
ti	0.290 ***	0.278 ***	0.263 ***	0.245 ***	0.255 ***	0.295 ***	0.268 ***	0.272 ***	0.262 ***
	(0.052)	(0.053)	(0.054)	(0.055)	(0.053)	(0.052)	(0.053)	(0.054)	(0.052)
Observations	1050	4861	2361	3550	4926	985	4365	1546	5911
Adjusted-R2	0.691	0.638	0.684	0.663	0.670	0.674	0.661	0.692	0.658

Note: The data are calculated by Stata 15.0. ***, **, and * refer to 1%, 5%, and 10% significance levels, respectively. Numbers in parentheses represent t values of bilateral tests.

present subgroup regressions on whether a city has appropriate transportation infrastructure when it first establishes an ETDZ, and the results of models (16) and (18) indicate that the establishment of a development zone has a significant negative impact on regional TFP for cities that do not have highways and common railways when they first establish an ETDZ; in contrast, the results of models (17) and (19) show that the establishment of an ETDZ has a positive impact on regional TFP if the city already has a highway and a common railway when it first establishes an ETDZ. This also means that the opening of highways and common railways in the city is a necessary condition for the state-level ETDZs to have a positive effect on regional TFP. The results of models (20), (21), (22), and (23) indicate that the presence of a port is not a necessary condition for the state-level ETDZs to enhance regional TFP, but there is icing on the cake, i.e., if the city has a port at the time of the development zone’s establishment, the development zone’s effect on regional TFP is stronger. In terms of the magnitude of the improvement in the regression coefficient, the effect of coastal ports is more pronounced than that of inland river ports.

Considering the objective fact that the transportation infrastructure is getting better and better across China during the observation period, it is further explored whether the promotion effect of the ETDZs on regional TFP will gradually emerge after the construction of the regional key transportation infrastructure is completed. The results are shown in Table 8. Model (24), for the sample of cities that did not have highways at the time of the first establishment of ETDZs, adjusts the time point of its policy treatment from the year of the first establishment of ETDZs to the year of the opening of highways and conducts a regression analysis based on the DID model of Equation (2). The results show that the treatment coefficient of DID is significantly positive at the 1% level, i.e., for cities that did not have a highway at the time of the establishment of the ETDZs, the point in time when the ETDZs really came into play may be after the highway was opened.

Considering the fundamental reality of significant economic disparities between regions in China, the actual effects of establishing state-level ETDZs in different regions may vary. Based on the classification criteria of the National Bureau of Statistics, this study divides 282 prefecture level cities into four regions, eastern, central, western, and northeastern, and analyzes the heterogeneity of the impact of state-level ETDZs on regional TFP. The results are shown in

Table 9. Heterogeneity analysis of different regions.

Model	(25)	(26)	(27)	(28)
Type of regions	Eastern	Central	Western	Northeastern
DID (treat × period)	0.274 ***	−0.459 ***	−1.40 ***	−0.235
	(0.064)	(0.083)	(0.372)	(0.127)
lngdp	0.105 ***	0.092 **	0.084 *	0.078 **
	(0.021)	(0.036)	(0.045)	(0.035)
fdi	0.050 **	0.043 *	0.038	0.042 *
	(0.023)	(0.025)	(0.031)	(0.024)
lnhum	−0.034 *	−0.048 *	−0.055 *	−0.028
	(0.019)	(0.027)	(0.026)	(0.020)
fin	0.087 ***	0.078 ***	0.069 **	0.061 *
	(0.031)	(0.029)	(0.035)	(0.033)
edu	0.041 *	0.037*	0.029	0.023
	(0.025)	(0.022)	(0.024)	(0.021)
pi	−0.112 ***	−0.129 ***	−0.145 ***	−0.138 ***
	(0.040)	(0.041)	(0.043)	(0.042)
ti	0.155 ***	0.144 ***	0.137 ***	0.123 ***
	(0.052)	(0.053)	(0.055)	(0.054)
Time fixed effect	Yes	Yes	Yes	Yes
Individual fixed effect	Yes	Yes	Yes	Yes
Observations	1787	1641	1715	626
Adjusted-R2	0.721	0.582	0.672	0.628

Note: The data are calculated by Stata 15.0. ***, **, and * refer to 1%, 5%, an 10% significance levels, respectively. Numbers in parentheses represent t values of bilateral tests.

. The state-level ETDZs have a promoting effect on TFP in the eastern region but are not conducive to the improvement of TFP in the central, western, and northeastern regions, especially in the western region. This may be due to the relatively early establishment of state-level ETDZs in the eastern region, which is also the region with the highest degree of openness to the outside world in China. Therefore, it can not only promote the progress of production technology through foreign investment but also absorb a large amount of cheap labor from the central and western regions. With the gradient transfer of China’s industry from the east to the central and western regions in recent years, although some mid to low end industries have been transferred to the central and western regions, their ability to absorb foreign investment and labor is not as good as that of the eastern regions. Moreover, due to their deep inland location and distance from shipping ports, there is a potential risk of increased transportation costs and reduced market accessibility. In this context, the establishment of state-level ETDZs in the central, western, and northeastern regions has increased government fiscal expenditure in the short term but may not necessarily achieve industrial agglomeration smoothly, resulting in a negative impact of state-level ETDZs on regional TFP.

In China’s political and economic system, there are differences in administrative levels between cities, and cities with different administrative levels often have significant differences in factor possession, market size, autonomy, transportation, and other characteristics. These characteristics are key factors affecting the effective implementation of development zone policies. Therefore, sub-provincial or provincial capital cities with higher administrative levels will have a strengthening effect on the policy effects of state-level ETDZs. Based on the above analysis, this study sets sub-provincial or provincial capital cities as the treatment group and general cities as the control group and conducts group regression according to the benchmark model. The heterogeneity analysis results are shown in

Table 10. Heterogeneity analysis of cities with different administrative levels.

Model	(29)	(30)
Type of administrative levels	Sub-provincial cities or provincial capital cities	General cities
DID (treat × period)	0.909 *	0.017 ***
	(0.443)	(0.043)
lngdp	0.112 ***	0.103 ***
	(0.035)	(0.021)
fdi	0.062 **	0.054 *
	(0.027)	(0.031)
lnhum	−0.075 **	−0.034 *
	(0.030)	(0.019)
fin	0.091 ***	0.084 **
	(0.033)	(0.029)
edu	0.051 *	0.061 **
	(0.028)	(0.027)
pi	−0.148 ***	−0.132 ***
	(0.051)	(0.045)
ti	0.179 ***	0.163 ***
	(0.057)	(0.048)
Time fixed effect	Yes	Yes
Individual fixed effect	Yes	Yes
Observations	500	5180
Adjusted-R2	0.809	0.694

Note: The data are calculated using Stata 15.0. ***, **, and * refer to 1%, 5%, and 10% significance levels, respectively. Numbers in parentheses represent t values of bilateral tests.

, indicating that state-level ETDZs generally perform better in cities with higher administrative levels.

5. Discussion

China’s economic trajectory has shifted from prioritizing the expansion of its scale to emphasizing the enhancement of its development quality. ETDZs play a crucial role in facilitating economic operations. Given the overarching context of a progressive decrease in regional TFP, it becomes imperative to assess the influence of China’s ETDZs on regional TFP. It is found that ETDZs, in general, exert a beneficial influence on regional TFP. However, it is worth noting that transportation infrastructure holds significant importance in the realm of regulation. In general, the establishment of ETDZs in regions characterized by robust transportation infrastructure yields a more pronounced positive effect on regional TFP compared to the establishment of such zones in regions with limited transportation infrastructure. Certain ETDZs were initially formed in regions characterized by insufficient transportation infrastructure, but the optimal functioning of these zones was only achieved subsequent to the completion of significant transportation infrastructure initiatives inside such regions.

This study exhibits some deficiencies that are attributed to the researchers’ limited level of research. It is anticipated that these shortcomings will be addressed and rectified in subsequent research endeavors. Firstly, this research primarily examines the presence of ETDZs within a city but does not extensively investigate the specific effects of the quantity and scale of such zones on regional TFP. Cities with higher administrative levels and superior economic development often possess multiple ETDZs. However, the present study does not delve into the intricate ramifications of the policy superposition effect resulting from the presence of multiple ETDZs, and its potential impact on regional TFP. It is crucial to assess the economic implications of newly established ETDZs in highly developed cities. This evaluation holds significance in determining the future trajectory of development zone policies in advanced regions. Consequently, it becomes an imperative subject for future research on ETDZs.

Furthermore, this study fails to take into account the influence of geographical factors. The significance of transportation is to achieve inter-regional connectivity, and this study only evaluates a city’s own transportation development level without fully considering the spillover effect of transportation infrastructure construction, ignoring the interaction of inter-regional transportation development, which may not truly reflect a city’s true level of transportation development. Spillover effects are present not only within the realm of transportation but also in the context of development zone policies. Since ETDZs can have good impacts on the local economy, they will certainly bring siphoning and spillover effects to surrounding cities, and how to detect these spatial effects will be the future direction of development zone research.

6. Conclusions and Policy Recommendations

This study investigates the impact of state-level ETDZs on regional TFP in Chinese cities from 1999 to 2020 and further explores how transportation infrastructure moderates this relationship. The main research findings are as follows: The state-level ETDZs enhance regional TFP by reducing firm costs, improving resource allocation efficiency, and fostering industrial agglomeration that enables economies of scale and knowledge spillovers. In addition, ETDZs also enhance regional TFP by promoting technological innovation and dynamic market selection, redistributing resources toward more productive enterprises through competitive mechanisms. In addition, transportation infrastructure plays a significant moderating role in the relationship between ETDZs and regional TFP. Among various modes of transportation infrastructure, highways, common railways and high-speed railways are especially effective due to their alignment with the logistical and operational needs of ETDZs. In light of these conclusions, this study proposes the following policy recommendations.

While state-level ETDZs have generally promoted regional TFP, their effectiveness varies across regions. Based on the findings of the study, it is evident that the decrease in regional TFP does not necessarily imply the ineffectiveness of the development zone policy. Rather, it can be attributed to the inadequate state of local transportation, which hinders the proper functioning of the state-level ETDZs. To enhance policy outcomes, local governments, especially those in the central and western regions, should coordinate the implementation of development zone policies and transportation infrastructure to achieve synergistic effects beyond the sum of individual contributions. In areas where ETDZs were established early but infrastructure remains weak, resources should be reallocated to improve connectivity rather than expanding zone coverage. Overall, aligning industrial policy with infrastructure planning is essential to maximize productivity gains and address regional disparities.

Transportation infrastructure plays a vital role in supporting economic activities, but for most prefecture-level cities, it is neither necessary nor efficient to pursue comprehensive development across all transportation modes. Instead, cities should prioritize improving road and rail networks, which have greater economic impact and better align with the logistical needs of development zones. Rather than investing heavily in costly infrastructure like ports and airports, which often serve broader regional functions, local governments should focus on enhancing connectivity with nearby major transport hubs. The government should establish a coordinated infrastructure plan, especially in the areas of highways and railways, which can improve industrial integration and better support the development of local development zones.

The relationship between regions and cities with varying levels of TFP development is intricately linked to their respective stages of economic growth and endowment conditions. Urban development should be based on the utilization of local distinctive resources and engagement of relevant market participants. It should also prioritize industrial collaboration and implement specialized division of labor within industries, with the aim of enhancing economic outcomes to their fullest potential. Less developed regions can benefit by learning from high-performing areas through experience sharing and industrial coordination. At the same time, in order to effectively learn from the experiences of more developed regions, it is imperative to enhance efforts in promoting domestic infrastructure development, scientific and technological innovation, financial support, and the establishment of an economic development model that aligns with the country’s specific stage of economic development and resource endowment.