It is widely believed that opening HSR generates economic influence: 1) Regional accessibility. Hall (2009) believes that HSR causes spatial influences; Jiao and Fang (2018); Wang C. Y et al. (2020) consider different lines and demonstrate that opening HSR significantly improves accessibility. 2) Transfer of firm-level production factors. Li et al. (2020) indicate that HSR markedly accelerated regional labor circulation and capital accumulation. Zhao et al. (2021) suggest that constructing transportation infrastructure for HSR facilitates high-quality firms. Cai et al. (2021) show that high-speed rail affects polluting firms’ locations through effects such as environmental cleaning. 3) Industrial structure at city level. Wang Z. H et al. (2020) reveal that HSR facilitates industrial structure over time; Wang and Lu (2021) reveal that opening HSR tends to reduce emissions and increase economic efficiency at city level; Chen (2021) demonstrate that HSR reduces energy consumption by facilitating industrial upgrading at prefecture level; Wong et al. (2022) believe that HSR facilitates upgrading urban industrial structure and curbing pollutant emissions; Lin and Jia (2022) suggest that HSR tends to significantly enhance a city’s total-factor carbon productivity. 4). Economic growth. Magazzino and Mele (2021) use a time series approach and panel data for 28 provinces over the time 1990–2017 with a Machine Learning technique and an econometric approach (time lag model) and reveal that investment in transport is positively and significantly correlated with economic growth for the short-term test. Li and Jian (2022) construct a Spatial Lag Model (SLM) with panel data of 283 cities in China from 2007 to 2019 and indicate that carbon emissions are spatially correlated, and high-speed railways negatively impact carbon emissions. Sahu and Verma (2022) take advantage of the upcoming Science City at Challakere in the state of Karnataka, India and optimize infrastructure investment from an organizational perspective and demonstrate that maximum achievable productivity gets affected with transportation connectivity level. Sun et al. (2023) apply data on 285 prefecture-level cities of China from 2003 to 2018 and staggered DID and demonstrate that HSR has a positive effect on urban economy. Wang and Dong (2022) use data of more than 2000 counties with time period from 2005 to 2016 and DID model and reveal that HSR connection promotes economic growth measured by GDP per capita or nighttime light data.

Technological progress, financial development, environmental regulation, and industrial structure are cited as influencing factors on high-quality economy. 1) Technological progress. Zhao et al. (2020) demonstrate that the digital economy can promote high-quality development by increasing entrepreneurial activity. 2) Financial development or green finance. Sun and Tang (2022) demonstrate that digital finance tends to contribute to sustainable economy at provincial level. Yang et al. (2021) reveal that green finance comprehensively facilitates a high-quality economy at the provincial level. 3) Environmental regulation. Wang and Lu (2018) reveal that it significantly promotes China’s high-quality economy. Similarly; Song et al. (2022) utilize 28 prefecture-level cities from 2005 to 2018 with spatial Durbin model and suggest that the impact of environmental regulation on economic high quality is an inverted “U" shape. 4) Industrial structure. Sun (2021) suggest that advanced and rationalized industrial structures tend to positively influence high-quality economies at urban level. Zheng and He (2022) confirm that industrial co-agglomeration positively affects the high-quality economy of the Chengdu-Chongqing Economic Circle. Lin and Zhou (2022) apply 30 provincial data during 2000–2017 in China with entropy weight method and bidirectional fixed effects method and find a U-shaped relationship between energy efficiency and economic growth quality and the upgrading of industrial structure plays a mediating role. Zhou et al. (2022) use provincial data with fixed effect model and find that infrastructure construction promotes high-quality economic development.

On one side, at city level, Li and Wang (2021) use the DID model to reveal that HSR promotes a high-quality economy across 108 cities along the Yangtze River Economic Belt. Sun and Zhang (2021) demonstrate that implementing HSR facilitates urban economy and improves the urban environment. Kong et al. (2022) use 284 Chinese cities over 2005–2018 with DID and indicate that the introduction of HSR can promote urban economic growth at city level, but they do not conduct impact mechanism analysis. More microcosmically, at county level, Zhang (2017) reveals that operating HSR facilitates economic growth measured by GDP per capita or nighttime light data. On the other side, existing literature focuses on influencing factors on high-quality economy (e.g., Li and Jian, 2022). Very few studies have investigated HSR’s influence on high-quality economy at county level.

In summary, some scholars probe factors influencing a high-quality economy, such as technological progress, human capital, environmental regulation, financial development level, institutional factors, and the degree of openness, but few studies have investigated the opening of HSR’s influence on high-quality economy. While these studies focus on macro provincial or city levels, no research exists on its influence at county level, where HSR is the primary vehicle for achieving China’s high quality economy and “dual carbon” target. No research has explored its mediating effect at the county level, although data availability is limited. In China’s case, probing the HSR’s influence at the county level illuminates the Chinese high-quality economy and also provides evidence for other developing countries regarding HSR’s facilitation of high-quality economic development.

The growth pole theory points out that after a growth pole is formed, factors always flow to regions with convenient transportation and a developed economy, that is, the growth pole region, which produces a greater diffusion effect and stimulates surrounding areas. HSR implementation is an important manifestation of a potential growth pole. The point axis theory also suggests that HSR belongs to the development axis, which is conducive to extending the geographical scope and development pattern of “point” cities, attracting industries and population, promoting a county’s industrial development and the formation of a wider regional economic belt. Furthermore, new economic geography believes that HSR tends to speed up factor movement. Transportation cost reduction and the emergence of economies of scale will also continuously promote regional industrial agglomeration and economic development, form a virtuous cycle, and continuously promote the formation of a “core-periphery” economic system and regional economic growth. In China, HSR adopts electrification technology, which coincides with sustainable development of energy savings and environmental protection; it has obvious environmental protection advantages, helping China reach the “double carbon” goals. Lu and Wang (2020) and Sheng and Shi (2021) reveal that HSR markedly improved overall economic quality across 270 prefecture-level cities, and this positive promotion is particularly significant in central-western regions. Considering that most counties upstream of the Yangtze River are located in central-western China, HSR’s implementation greatly shortens the distance between counties, brings about rapid factors and resources flows, upgrades the industrial structure, and accelerates county level high-quality economy. Accordingly, this study posits research hypothesis 1: Implementing HSR positively influences a high-quality economy at the county level in the upstream Yangtze River area.

At present, HSR’s advantages include facilitating accessibility between counties and the circulation of factor resources, for which consensus exists in the literature. However, whether HSR influences high-quality economic growth at the county level across different regions requires further consideration and discussion. According to Zhang (2012), He and Liu, 2020, Wang et al. (2014), and Zhang (2017), transportation facilities’ influence on economic development depends on the “diffusion effect” and “agglomeration effect,” and HSR tends to cause different economic impacts in different regions along the line, forming a differentiated regional economic structure. Most studies conclude that opening HSR has differential impacts on different geographical locations or regions with differing economic scales depending on whether the “diffusion effect” and “agglomeration effect” are large or small. There are many involved counties in the upstream Yangtze River area; natural and social conditions between them differ, and HSR policy affects high-quality economy at the county level in these regions differently. Accordingly, we present research hypothesis 2: HSR has a heterogeneous influence on a high-quality economy at the county level in the upstream Yangtze River area.

Many scholars have concluded that operating HSR influences the industrial structure, and its improvement and optimization drive economic development. Chen and Hall (2011) demonstrate that Eurostar train in the United Kingdom accelerates factor gathering in tertiary industries and promotes industrial upgrading. From the perspective of division of labor, convergence, and learning effects, Sun et al. (2022) show that HSR raises population mobility and capital inflow, significantly improves urban industrial structure index, and positively influences economic growth.

Therefore, The above-mentioned literature reveals that opening a high-speed railway influences high-quality economy mainly through the intermediary effect of industrial structure, which means that the influence of HRS on high-quality economy is realized by M. That is, M is function of HRS and high-quality economy is function of M. Mathematically, this can be expressed as follows. Y = α X + ε 1 (1) M = β X + ε 2 (2) Y = α ’ X + γ M + ε 3 (3)

In formulae 1,2 and 3, Where Y indicates high-quality economy, X represents operating high-speed railway, M stands for industrial structure. ε is the error term. α = γ × β + α ’ . If α ′ is insignificant, the mediating effect is called full mediation. Otherwise, If α ′ is significant and less than α , the mediating effect is called partial mediation. Accordingly, this study proposes research hypothesis 3: opening a high-speed railway influences industrial structure, which further facilitates a high-quality economy at the county level in the upstream Yangtze River area. The logical relationship diagram is shown in Figure 1.

In the process of proposed hypotheses, this study poses relevant research questions as follows: Does HSR facilitate high-quality economy in county of China? A candidate method is to use PSM-DID model to identify if there is a causal relationship between HSR and high-quality economy. Previous studies with either FE or OLS model cannot answer the causal relationship because of endogeneity. Further, why would HSR influence high-quality economy? This causal relationship could be explored steadily after solving endogeneity problem brought by the presence of mutual causality or omitted variables or sample selection bias with PSM-DID, GMM and IV models, which have been largely ignored in the literature. The third question is: how could HSR influence high-quality economy? A proposed method is to perform mechanism analysis of improving industrial structure derived in this paper. Unfortunately, previous studies largely ignore the mechanism analysis.

Theoretically, while much attention has been paid to the economic effects of operating HSR, less attention has been paid to HSR’s influencing on high-quality economy. We attempt to investigate influencing factors on high-quality economy from novel perspective of HRS. As the core of modern economy, infrastructure and new economic geographic theory is quite rich. With developing HSR, it comes into being, but theoretical research is still very limited. Following the framework of Acemoglu et al. (2012), this study describes a simple theoretical framework for HSR to facilitate high-quality economy. Basic logic states that technological progress relies on relatively expected values of low-quality and high-quality capitals. HSR represents high-quality technology. Combining the market clearing conditions, we reveal that on endogenous growth path, there exists a dynamic equilibrium: marginal return of low-quality or unsustainable technology equals marginal return of high-quality or sustainable technology. This implies that economic growth quality and improving sustainability seem to be realized simultaneously, which enriches relevant literature.

Following literature (e.g., Zhang et al., 2016), a multiperiod DID model was built. A basic regression model was established as follows: y it = α 0 + β 0 H S R i ∗ d t + γ X i t + λ i + υ t + ε i t (4)

In this model, Y i t is explained variable, representing the high-quality economy in county i in year t ; H S R i t is a dummy variable and stands for core explanatory variable, representing building a HSR in county i ; d _t is a dummy variable and denotes operating HSR in year t ; X i t represents several control variables that tend to influence the high-quality development of the county. λ i stands for regional fixed effect. υ t denotes time fixed effect. ε i t represents a random disturbance.

(1) A high-quality economy. At county level of China, this study draws on previous research (Gong et al., 2021; Lin and Zhou, 2022) by combining with county data availability, and constructs an index for estimation by including GDP, number of students in ordinary middle schools, number of medical beds, total power of agricultural machinery, CO₂ emissions, and PM2.5 concentration. In order to avoid the influence of subjectivity on the indicators’ weights, we used the entropy weighting method (EWM) to estimate the high-quality economy with 396 counties in upstream Yangtze River area from 2006 to 2018 ( Y i t ) objectively and accurately, which is widely used in economics and statistics. The basic logic of EWM is to determine the objective weight according to the size of index variability. In general, the smaller the information entropy of an indicator, the greater the variation of the indicator value, the more information it provides, the greater the role it can play in the comprehensive evaluation, and the greater its weight and vice versa. The detailed calculation procedure can be described as follows:

First, standardization of indicators. Since the indicators within the evaluation index system for high-quality economic development in counties are different in nature, the units and directions are not exactly the same. Thus, it is necessary to standardize (also called dimensionless) the raw data of all indicators. Z a i j = x a i j − x min x max − x min Z a i j = x max − x a i j x max − x min (5)

The two Eq. 5 represent the normalization method for positive and negative indicators, respectively. Where j is the number of indicator terms, and x a i j represents the value of the jth indicator before standardization, and Z a i j represents the value of the jth indicator after standardized treatment processing, and x max denotes the maximum value in the jth indicator, and x min denotes the minimum value of the jth indicator.

Next, indicator normalization process. Normalization of the index Z a i j obtained after normalization. P a i j = z a i j ∑ a = 1 n ∑ i = 1 m z a i j (6) Next, calculate the information entropy of each index. Using n denotes the year, and m denotes the number of sample counties in this paper, and P a i j is the coefficient of variation obtained after normalization of indicators, and m 1 is the reciprocal of ln n * m . Eq. 7 indicates that ln P a i j is obtained by taking the logarithm of P a i j and then multiplying it with P a i j and sum up the values of all counties for all years obtained after multiplication, and then obtain the information entropy of each index. E j denotes the entropy value of the jth indicator, which takes values between 0 and 1. E j = − m 1 ∑ a = 1 n ∑ i = 1 m P a i j ⁡ ln ⁡ P a i j (7) Next, calculate the redundancy of each index. In Eq. 8, subtract E j from one to obtain the redundancy of each indicator as R j : R j = 1 − E j (8)

Furthermore, calculate the weights of each indicator. Eq. 9 represents the calculation of the weights of each indicator based on the redundancy of the indicators W j : W j = R j ∑ j = 1 k R j (9)

Lastly, calculate the comprehensive evaluation index. After calculating the weights accounted for by each index, multiply Eq. 9 with Eq. 6, and sum up all indicators for the ith county to obtain the economic quality development index I a i for the ith county in year a. I a i = ∑ j = 1 k P a i j × W j (10)

The constructed evaluation indicator system is demonstrated in Table 1.

(2) Opening high-speed railway ( H S R i t ). This variable is our core explanatory variable, which is defined as operating a railway station along the high-speed railway and providing service for people taking high-speed trains located in a certain county in a certain year. The counties opening HSR are taken as the treatment group; otherwise, counties without HSR are set as the control group, which is represented by the group dummy variable D i t . If county i operates a HSR in year t , D i t = 1; otherwise, D i t = 0. Simultaneously, the time virtual variable T i t can be constructed. Prior to HSR, T i t = 0; after implementation, T i t = 1. It should be noted that in this study, counties whose HSR opening time is before June 30 of the current year are classified as opening and T i t = 1 in this year; counties whose opening time is after June 30 can be defined as opening in the following year, that is, T i t = 0 in this year and T i t = 1 in the following year. This study referenced the treatment method of Li and Zhu (2022). If multiple HSR are opened in a county, the moment when the earliest-built HSR station is put into operation is considered the implementation time. This study focuses on the interaction term D i t × T i t of the group 0–1 variable and time virtual variable and replaces it with the variable H S R i t to indicate whether county i is influenced by opening the HSR in year t. That is, if county i in year t operates a HSR, H S R i t = 1 in the following period; otherwise, the variable H S R i t = 0.

In 2021, there were 512 county administrative units in the Yangtze River upstream area, including 183 in Sichuan, 38 in Chongqing, 44 in Qinghai, 30 in Hubei, 129 in Yunnan, and 88 in Guizhou provinces. Some counties in Qinghai and Sichuan provinces close to Tibet do not have two consecutive year data for fixed effect use. Because of low population density by 2.2 persons/km² and location on frigid “Roof of the World” with some no man’s land unsuitable for human habitation, we cannot obtain economic data or values that are small and can be neglected. These missing counties are completely at random and independent of the value of any other variables or counties. On the other side, no HSR operates in these counties with missing whole county data in Figure 2. The limited of the information would not impact on our research theme. Thus, there are a total of 396 counties adopted for our empirical analysis where 65 counties operate HSR. We do summarize a timeline of HSR operation in the study area as Table 2.

With counties, for example, data on indicators or control variables are sometimes missing for certain years. The key issue with the unbalanced panel is determining why the panel is unbalanced. Provided the reason we have missing data for some i is not correlated with the idiosyncratic errors ε i , the unbalanced panel causes no problem (Wooldridge, 2001, p448). In our case, we do not intentionally choose particular indicators or control variables with missing years. In addition, we perform many tests such as correlation and stationary tests. Furthermore, we use three methods (PSM-DID, IV, GMM) to obtain robust conclusions by solving endogeneity induced by omitted variable bias, reverse causality, sample selection deviation.

Due to strict policy on geographic and social data disclosure in China, we explain the subdivision units are county bounaries in the map of the investigated area in data sources section. In addition, we added HSR operation county as a shaded area in red color. A map of the investigated area in shallow green color is provided in Figure 2.

The evaluation index system for high-quality economic development consists of five indicators including GDP per capita, number of general secondary school students, number of beds, and agricultural machinery from the 2019 China County Statistical Yearbook (CCSY, 2019), Data on CO₂ sourced from the National Genome Data Center (NGDC, 2020), and PM2.5 from the Atmospheric Composition Analysis Group (ACAG, 2022) at Dalhousie University. The data on the opening of high-speed rail were obtained from the China Railway Yearbook (CRY, 2019), and the website of China National Railway Group Corporation (2020) (CENSD, 2020), and collected and compiled using Python crawler technology.

The control variables (government intervention, living standard of residents, construction of infrastructure, density of social activities, and industrial development) were obtained from the China Economic network statistical database (CENSD, 2020) and the mediating variables (industrial structure) were obtained from the 2019 China County Statistical Yearbook (CCSY, 2019). Data for all indicators were calculated per capita, except for PM2.5 indicator data, which were annual average concentrations. All monetary indicators were deflated. All indicators were log-transformed except for the evaluation indicator of high-quality economic development and the opening of high-speed rail.

Table 3 demonstrates that the average value of the high-quality economy of counties in the Yangtze River upstream area is 0.09, implying that there are distinctions in high-quality economy across counties; the average value of opening HSR for China between 2006 and 2018 was about 0.06 with a standard deviation of 0.05; the mean value of government intervention was 0.34 with a standard deviation of 0.35; the mean value of living standard of residents was 11,000 with a standard deviation of 9,701; the mean value of telephone number was 48100 with a standard deviation of 47,900; the mean value of intensity of economic and social activities was 0.03 with a standard deviation of 0.02; the mean value of development of industrial enterprises was 55.16 with a standard deviation of 60.95; the mean value of topographic relief was 1.84 with a standard deviation of 1.16; the mean value of Industrial structure was 0.36 with a standard deviation of 0.09.

In Table 3, we also list Skewness, kurtosis, variance inflation factor (VIF), which represent a measure of the direction and degree of skewness of statistical data distribution and a statistic compared with the normal distribution, avoiding multicollinearity, respectively. In order to prevent the estimation error caused by multicollinearity, we calculate the variance inflation factor (VIF) of each variable and reveal that the maximum VIF value of variables is 2.54, which is far less than 10, indicating strong independence between variables. Therefore, there is no serious multicollinearity problem between the variables.

Before the regression analysis, we conduct correlation analysis with Pearson test on each variable and demonstrate that all correlation coefficients are less than 0.7, indicating that there is a correlation between various variables without considering other factors (Table 4).

Stationary tests with Augmented Dickey-Fuller (ADF) and Philips Perron (PP) methods were performed for each time series of each variable, first on levels and then on the first differences. We can see that ADF and PP demonstrate that except for intensity of economic and social activities (Den) variable, we reject the null hypothesis of a unit root (0.05 significance level) for all other variables, while in the first difference we reject the null hypothesis for all variables (Table 5).

1) Logit Model Construction and Propensity Score Estimation. Propensity Score Matching (PSM) is effective in addressing the problems of sample selection bias and heterogeneity bias. Thus, the treatment and control groups were identical in terms of their trend characteristics. Therefore, the relevant observable variables are matched to the counties of the treatment and control groups. For the sake of more accurately matching between the treatment and control groups, referring to Lin et al. (2020), the logit model for estimating the propensity score is established as follows: L o g i t T r e a t e d = 1 = β 0 + β i X i + ε i (11)

T r e a d e d represents the virtual variable operating the HSR, and X i represents the matching variable. Table 6 shows part of the PSM estimation results. The estimated ATT is 0.0053, with a 10% significance level, demonstrating that the sample estimation effect is good. In addition, according to the propensity score results of all samples after Logit model regression, it can be found that most matching variables display significance at the 10% level.

2) Matching Balance and Cosupport Tests. Aiming at the differences before and after the matching of control variables, a balance test can be performed on the treatment and control groups to investigate whether the PSM-DID model is effective. After matching, the testing results showed that the differences between the treatment and control groups tended to be remarkably decreased, and the absolute value of the standard deviation of each pairing index was significantly less than 10%. Therefore, the use of PSM here was reasonable and met the requirements of randomized trials.

Another precondition for PSM is that the “common support hypothesis” must be satisfied. To satisfy this assumption, this study removes individuals with propensity scores close to the two tails in the counties of the treatment and control groups, and only uses these matched samples to assess HSR’s influence on high-quality economies. Figure 3 demonstrates that all sample-matching effects are good and the assumption is satisfied.

The plot about the common range of values for the propensity scores of the experimental and control groups, the vertical axis can be interpreted as the sample size. As can be seen in Figure 3, the majority of the propensity score values for the experimental and control groups lie within the common range of values. Only individual samples are outside the common range of values (green section). As a result, only a small number of samples are lost when propensity score matching is performed. Therefore, the common support hypothesis is satisfied.

Before using the above model for regression analysis, this study draws from Beck et al. (2010) to perform parallel trend tests to confirm that the change trend of the high-quality economy in the treatment and control groups before operating HSR is similar. Otherwise, the changes in the high-quality economy of counties in the upper Yangtze River area may not be caused by implementing HSR. Figure 4 shows that the parallel trend assumption holds in this case.

In the graph, the vertical axis is the policy effect, which indicates the level of high-quality economic development of the county in that year, and the horizontal axis is before and after the policy implementation. As can be seen from Figure 4, there is no significant difference in the level of high-quality economic development between the experimental and control groups before the operation of HSR, which is consistent with the parallel trend hypothesis. And after the implementation of the policy, especially after the third year of the policy implementation, the level of high-quality economic development in the county began to improve, so the model results estimated in this paper can be a more realistic response to the policy effect.

Referencing Shi et al. (2018), this study uses the PSM-DID method to perform stepwise regression to assess HSR’s effect on the high-quality economy of counties in the upstream Yangtze River area. Table 7 presents the benchmark findings with the PSM-DID models. Column 1) includes only core explanatory variables, and columns (2)–(6) are the regression outcomes obtained by gradually increasing the control variables one by one.

The regression results demonstrate that the coefficients are positive at the 1% level. These findings accord with the expectations of this study, suggesting that opening HSR tends to have a significantly positive influence on the high-quality economy of sample counties. Thus, Hypothesis one is supported. From the column-by-column comparison, it appears that, with the gradual addition of control variables, the coefficients of the core explanatory variable show an overall decreasing trend, from the initial 0.021 to 0.008, and the goodness of fit R₂ continues to increase. This finding reveals that after controlling for the other factors, opening the HSR, which influences the high-quality economy revealed by the regression results of the PSM-DID model, is more accurate.

Some scholars, such as Xu et al. (2015), Fan et al. (2016), Zhang (2017), and Niu and Xin, (2021), apply DMSP/OLS night satellite lighting data to measure regional economy as a proxy to avoid being affected by price factors in different regions at different times. This study selects each county’s average nighttime brightness data to replace the variable data of GDP, which measures economic growth, and uses the entropy weight method (EWM) to re-estimate the high-quality economy of counties. Table 8 presents our empirical findings.

As expected, core explanatory variables are markedly positive across different specifications, proving the above conclusion’s stability, namely, HSR significantly boosts the high-quality economy of counties. The signs of the control variables with regression models also coincide with expectations in the literature.

The placebo test in the DID model usually includes two methods: changing the time point of the policy and randomizing the treatment group. This study consults Fan and Tian, (2013), Zhang and Tao, (2016), and Ji and Yang, (2020) to conduct counterfactual tests and artificially change the HSR operating time. Specifically, taking the opening time by 2 years, 3 years, and 4 years in advance as a dummy year, we construct three virtual variables of “pseudo-high-speed rail opening” and artificially construct a new processing group to re-estimate. Table 9 presents the regression results.

The consequences of our dummy variables are not significant, indicating that if a counterfactual test is conducted to artificially advance the HSR opening time, a high-quality economy cannot be significantly improved. This further verifies that operating HSR substantially promotes a high-quality economy in the sample counties.

Referencing Lin and Tan (2019), terrain relief was chosen as the instrumental variable. Topographic relief satisfies the two properties of “correlation” and “exogenousness” as an instrumental variable, so it is effective and feasible to select it as an instrumental variable. Two-stage OLS and GMM methods were used for the evaluation. Tables 10 and 11 show the estimation results. With or without the control variables, the X coefficient of the core variable appears significantly positive. This implies that operating HSR enhances the high-quality economy of counties in the upstream Yangtze River area. Our research conclusion is robust and alleviates possible endogeneity problems.

Referring to Shen and Li, (2021), lagged variables of high-quality economic development are introduced into the regression model as explanatory variables, a dynamic panel data model is constructed, and the system GMM is used to perform regression analysis. Table 12 shows the GMM estimation outcomes of the dynamic panel one-stage system. The findings reveal that implementing HSR can remarkably promote the high-quality economy of sample counties. After the control variables were gradually added, the regression results remained stable. This method effectively alleviates the possible endogeneity problems.

Taking the average value of high-quality economy with level data acquired from the descriptive statistics of variables as the dividing point, we divide the county-level administrative regions into “regions with high-quality economy higher than average” and “regions with high-quality economy lower than average” according to their development level and conduct regression and heterogeneity test separately. The regression results are listed in Table 13. HSR tends to facilitate the high-quality economy of counties whose development level is higher than average, while it has a certain adverse influence on counties whose high-quality economies are lower than average.

A possible explanation is that when the level of a high-quality economy is lower than average, these counties often have a weak economic development foundation and insufficient support facilities, although HSR implementation seems to connect regions with different development levels and improve accessibility between regions. In addition, because of insufficient national support policies, lack of human resources, and relatively backward industrial infrastructure, implementing HSR tends to cause a large labor force to flow to developed counties, and the siphon effect is obvious. The radiation-driven effect of well-developed areas on these undeveloped areas cannot be quickly manifested in the short term, which will adversely affect areas with a lower-than-average levels of high-quality economy upon HSR implementation. Hypothesis two was confirmed in this case.

To further explore the heterogeneity of HSR’s influence on the high-quality economy of counties, this study regresses counties in different provinces in this area. Table 14 shows the empirical results. Our findings reveal that implementing HSR remarkably improves the high-quality economy of counties or districts in Qinghai and Hubei Provinces (up to Yichang). HSR has a positive but not significant effect on Sichuan, Yunnan, and Guizhou Provinces, and a relatively prominent negative suppressive influence on high-quality development in Chongqing counties.

A possible explanation is that owing to apparent regional differences in the efficiency of capital and labor redistribution, HSR’s influence on the high-quality economy of counties across different provinces is also heterogeneous. For example, because of the construction and development of the economic circle of twin cities in the Chengdu-Chongqing region, it has certain regional and economic advantages, and the transportation infrastructure has been relatively perfect. The promotion of high-quality economic advances by HSR partially is partially reduced to a certain extent. Therefore, we cannot conclude that HSR’s advent has certain positive or negative influences on the high-quality economy of counties in the Sichuan-Chongqing area. Hypothesis two was thus confirmed in this study.