Research on the Spatiotemporal Evolution Characteristics and Capital Driving Factors of Sustainable Economic Development in Northwest China

Abstract

Frontier research has focused more on the sustainable economic development (SED) of developing countries, with slightly less attention paid to the SED of underdeveloped regions in developing countries, especially without analyzing their driving factors from the dual perspectives of domestic and international capital. In fact, as an important factor in economic growth, capital is also the core driving force for achieving SED. This paper takes the five provinces (autonomous regions) of northwest China (NC) as a case study based on unbalanced panel data from 26 cities from 2000 to 2020 and employs the dynamic degree and MSAR model to analyze the spatiotemporal evolution characteristics of SED; the results indicate that the SED of NC exhibits an irregular pattern overall, the differences between provinces at the same time point are relatively small, and the same applies to prefecture-level cities. The spatial agglomeration characteristic of SED in NC is evident, with significant room for improvement. Additionally, this paper evaluates the impact of capital on the SED of NC through the two-way fixed effects model and the mediation effect model and finds that foreign direct investment (FDI) inflows have a “first-rising-then-declining” impact on the SED of NC, while financing constraints (FC) play a promoting role in the SED of NC. Furthermore, FDI inflows subsequently impact the SED of NC by affecting regional FC, with FC acting as a mediating variable in the influence of FDI on the SED of NC. In terms of policy implications, NC should make rational use of FDI, control the pace of easing FC, and precisely promote the SED of NC from a capital-driven perspective. Other underdeveloped areas in developing countries can take this as a reference to promote SED.

1. Introduction

Since the concept of sustainable development was mentioned by the Chinese government, the sustainable development of regional economies has gradually become one of the research topics in academia [1], and these studies mainly discuss path selection for sustainable economic development (SED) [2]. As a window for China to open up to the West, northwest China (NC) is located at the forefront of the Silk Road Economic Belt and has important strategic geopolitical and economic significance. However, NC is underdeveloped in terms of social and economic development and ecological environmental construction. The Chinese government has provided policy support to NC. Therefore, as a typical representative of underdeveloped regions in developing countries, NC has also garnered the attention of some scholars regarding its SED [3]. In addition, SED is prominently characterized by the improvement of total factor productivity (TFP) [4]. TFP is an important tool for analyzing the sources of economic growth, which can help identify whether economic growth depends on input-based growth or efficiency-based growth and thus judge the sustainability of economic growth. Therefore, this paper will use the regional TFP as an indicator to measure the level of SED in NC.

Furthermore, influenced by multiple uncertainties such as the COVID-19 pandemic and trade frictions, the outlook for global economic development remains unclear. The inflow and utilization of FDI in the field of capital elements [5] and the alleviation of financing constraints (FC) [6] have once again become the focus of academic research. Therefore, this paper will take the five provinces (autonomous regions) in NC as research samples, measure the level of SED in NC, display the pattern of SED, and analyze the evolution process and mechanism of SED in order to reveal the spatial differentiation and temporal trend of SED and gain a deeper understanding of the evolution laws of SED. Meanwhile, this paper attempts to explore the capital-driven factors of SED from the perspectives of FDI and FC.

The marginal contribution of this paper lies in the following: Firstly, in terms of research methods, the existing literature mostly explores the relationship between capital and SED based on empirical econometric methods [7]. This paper integrates various research methods from geography and economics, first following the geographical research approach of patterns, processes, and mechanisms, reporting the evolution process of SED, and analyzing its evolution mechanism. Then, it combines econometric methods to test the relationship between capital and SED, enriching the theoretical research on the relationship between the two. Secondly, in terms of research perspective, existing research on SED is abundant [8]; however, there is insufficient attention paid to the underdeveloped regions of developing countries. This paper, taking 26 cities across five provinces (autonomous regions) in NC as the research sample, analyzes the path of SED in NC through empirical tests, thereby enriching the relevant studies on SED in underdeveloped areas in developing countries. Thirdly, in order to avoid the economic changes caused by the COVID-19 pandemic, this paper uses the data of NC from 2000 to 2020 as the research sample and finds that FDI and FC are the driving factors of SED, thus confirming the positive role of capital in SED. In addition, this paper attempts to analyze the inflow and utilization of FDI, as well as the alleviation of FC in NC, against the backdrop of a global economic situation turnaround. It proposes strategies for the SED of NC, with the hope of providing a reference for the economic development of underdeveloped areas in other developing countries.

The remaining paper is structured as follows. The second part is the literature review, which reviews the research basis of the thesis theme. The third part is research design, which introduces the methodology and data. The fourth part includes represented facts, which report the evolutionary characteristics of SED in NC. The fifth part includes the empirical results, and the sixth part discusses the results in detail. The seventh part summarizes the full paper and provides policy implications.

2. Literature Review

2.1. Literature Review on Impact of FDI on SED

The impact of FDI on the SED of a host country’s economy has always been a focus of attention for academia and policy makers [9,10], but there is no unified statement on whether FDI has a positive or negative effect on SED. Firstly, the existing literature generally suggests that FDI has a positive impact on SED. For example, some scholars point out from the perspective of economic growth that FDI promotes regional sustainable development through capital transfer, technology spillover, and structural optimization [11]. They believe that FDI enhances the economic quality of host countries by promoting technological innovation and industrial upgrading [12]. Some scholars also propose from the perspective of balanced economic development that FDI contributes to income inequality in emerging-market countries [13]. Secondly, a small number of scholars believe that FDI has a negative impact on SED. For example, some scholars, from the perspective of environmental pollution and based on the pollution paradise hypothesis, believe that FDI may bring about environmental pollution and resource consumption problems, and that a large influx of FDI into heavily polluted industries is not conducive to SED [14]. Some scholars, from the perspective of regional development, believe that FDI may exacerbate regional development imbalances [15]. Based on the above analysis, the impact of FDI on SED is uncertain. This paper believes that appropriate FDI inflows can help promote SED, while excessive FDI may lead to environmental pollution and be detrimental to SED. Based on this, this paper proposes the following hypothesis.

H1. 

The impact of FDI on SED exhibits a nonlinear, inverted, U-shaped relationship.

2.2. Literature Review on Impact of FDI on SED Through FC

FC, as an important obstacle to enterprise development, cannot be ignored in terms of their impact on SED. This section aims to explore the impact of FDI on FC and the impact of FC on SED. Firstly, we discuss here the impact of FDI on FC. Previous studies have shown that FDI alleviates FC for enterprises by reducing information asymmetry and technology spillovers in the credit market [16]. Therefore, most scholars believe that FDI inflows can bring capital, optimize capital allocation caused by credit distortions in a host country, reduce the cost of capital, and thereby alleviate FC caused by financial inefficiency [17]. A minority of scholars argue that FDI inflows may displace financing for domestic firms, having a negative impact on their financing and thus exacerbating FC [18]. Other scholars have pointed out that the impact of FDI inflows on FC in a host country is subject to other conditions and is therefore uncertain [19]. Secondly, we discuss here the impact of FC on SED. Most existing scholars use TFP to measure SED and generally believe that FC have a suppressive effect on TFP [20]. Some literature points out that FC have other impacts on SED; improvements in FC do not enhance SED [21], and FC may even contribute to SED [22]. There is a nonlinear relationship between FC and SED [23]. Based on the above analysis, FDI can affect FC, which in turn can effectively enhance SED. Based on this, this paper proposes the following hypothesis.

H2. 

FDI affects SED by acting on FC.

3. Research Design

3.1. Regional Overview of Study

In terms of the study area, this paper focuses on the five provinces (autonomous regions) in NC: Shaanxi, Gansu, Qinghai, Ningxia, and Xinjiang. The total administrative area of the five provinces (autonomous regions) in NC is about 3.043 million square kilometers, accounting for 31.7% of the national land area. In 2023, the GDP was CNY 7388.99 billion, accounting for only 5.89% of the national GDP, and the average growth rate was 4.91%. The permanent population in 2023 exceeded 100 million, accounting for about 7% of the total national population. NC is not only a bridge area for trade between China and central Asian countries but also an important part of the domestic section of the “Silk Road Economic Belt”. The five provinces (autonomous regions) in NC have repeatedly received policy support from national strategic plans such as the Western Development and the Belt and Road Initiative. They are rich in energy and mineral resources, have vast land, have unique geographical positions, and have unique conditions for opening up to the West. Due to the fact that NC is a typical representative of underdeveloped regions in developing countries, this paper will take NC as the research sample to analyze the spatiotemporal evolution trend of its SED and analyze the driving factors of the SED from the perspectives of FDI and FC. Against the background of the global economic situation turning point, exploring the dynamic characteristics and driving mechanisms of SED in NC is both a summary of its historical development experience and a prospect for future development paths.

3.2. Methodology

This paper will use stochastic frontier analysis (SFA) to measure the SED of NC, then analyze the spatiotemporal evolution characteristics of the SED based on the dynamic degree and MSAR model, and finally test the impact of FDI and FC on the SED based on a two-way fixed effects model and mediation effect model. Therefore, this paper will focus on introducing SFA, the dynamic degree, the MSAR model, the two-way fixed effects model, and the mediation effect model.

3.2.1. Stochastic Frontier Analysis

Regional TFP measurement methods are diverse. This paper is based on the conventional calculation method, namely the SFA method, to measure the TFP that can characterize SED. First, referring to the method of Kumbhakar (2000) [24], a translog form of the production function is specified:

$$\begin{array}{l}\ln Y_{it}={\beta }_0+{\beta }_k\ln K_{it}+{\beta }_l\ln L_{it}+{\beta }_{t}t+{\displaystyle \frac{1}{2}}{\beta }_{kk}\ln K_{it}^2+{\displaystyle \frac{1}{2}}{\beta }_l\ln L_{it}^2\\ +{\displaystyle \frac{1}{2}}{\beta }_{tt}{t}^2+{\beta }_{kl}\mathrm{tln}{K}_{it}+{\beta }_{lt}t\ln L_{it}+{\beta }_{kl}\ln K_{it}\ln L_{it}-u_{it}+{\epsilon }_{it}\end{array}$$

(1)

In Equation (1), $u_{it}$ represents the inefficiency term, and ${\epsilon }_{it}$ is the random error term. At the same time, referring to the distribution setting by Battese and Coelli (1992) [25], $u_{it}=u_i{e}^{-{\eta }_i}$; ${\epsilon }_{it}~N\left(0,{\sigma }_{\epsilon }^2\right)$; and $cov\left(u_{it},{\epsilon }_{it}\right)=0$. $Y_{it}$ is the level of output, $K_{it}$ is capital input, and $L_{it}$ is labor input.

Following the function setup and calculation as described, and referring to the research process of Fuentes et al. (2001) [26] and Orea (2002) [27], the cumulative growth rate of TFP can be obtained through Equation (2).

$$TF{P}_{it}=\exp \left({\beta }_0+{\beta }_{t}t+{\displaystyle \frac{1}{2}}{\beta }_{tt}{t}^2+{\beta }_{kt}t\ln K_{it}+{\beta }_{tt}t\ln L_{it}\right)\times T{E}_{it}$$

(2)

Building on the research of Fuentes et al. (2001) [26], the TFP for the current year can be calculated by accumulating the growth rates of TFP from previous years. In this context, the output variable is the gross regional product, and the input variables include labor and capital, where labor is the total number of urban employees and capital is the capital stock calculated using the perpetual inventory method.

3.2.2. Dynamic Degree

In accordance with the research content of this paper, based on the provincial-level data of SED calculated for the five provinces (autonomous regions) in NC, the dynamic degree calculation formula was utilized to measure the magnitude and rate of change, which were used to analyze its convergence or divergence characteristics. The detailed calculation formula is as follows:

$$D={\displaystyle \frac{U_b-U_a}{U_a}}\times {\displaystyle \frac{1}{T}}\times 100\%$$

(3)

In Equation (3), D represents the dynamic degree of the SED of the research subject, U_a and U_b represent the initial and terminal values of the SED level of the research subject, and T represents the length of the research period.

3.2.3. MSAR Model

The P-order Autoregressive (AR) model based on time series, as a linear model, fails to capture the structural change characteristics of variables. Therefore, to examine the nonlinear transition process of time series variables, following the research approach of Krolzig (1997) [28], an unobservable regime state variable is incorporated into the AR model, resulting in the following MSAR model:

$$y_t-\mu (s_t)={\varphi }_1(s_t)(y_{t-1}-\mu (s_t))+\dots \dots +{\varphi }_p(s_t)(y_{t-p}-\mu (s_t))+{\epsilon }_t, {\epsilon }_t|I_{t-1}\sim iid N(0,{\sigma }^2)$$

(4)

$$y_t=v(s_t)+{\varphi }_1(s_t)y_{t-1}+\dots \dots +{\varphi }_p(s_t)y_{t-p}+{\epsilon }_t, {\epsilon }_t|I_{t-1}\sim iid N(0,{\sigma }^2)$$

(5)

Equations (4) and (5) represent the MSMA(M)-AR(p) and MSIA(M)-AR(p) models, respectively, with the former indicating that the mean and coefficients vary with the regime state and the latter indicating that the intercept and coefficients vary with the regime state. In Equations (4) and (5), y is the time series variable, μ is the mean, v is the intercept, p is the lag order, ${ \epsilon }_t$ is the error disturbance term following a normal distribution,${\varphi }_i$ denotes the autoregressive coefficients, and $I_{t-1}$ is the information set in the period t − 1. This paper intends to model the annual average level of SED in the five provinces (autonomous regions) of NC from 2000 to 2020 using the aforementioned MSAR model and to analyze the temporal changes in the characteristics of SED in the five provinces (autonomous regions) of NC.

3.2.4. Two-Way Fixed Effects Model and Mediation Effect Model

One of the research objectives of this paper is to analyze the capital-driven mechanism of SED in NC through empirical means. Therefore, a two-way fixed effects model was established for econometric analysis. The specific econometric model was set up as follows:

$$se{d}_{it}={\alpha }_0+{\theta }_{1}X+{\displaystyle \sum {\beta }_i}contro{l}_{it}+cod{e}_i+yea{r}_t+{\epsilon }_{it}$$

(6)

As this paper will also explore the possible transmission mechanisms between capital and the SED of NC, following the research of Liu et al. (2023) [29], a mediation effect model was also established on the basis of Equation (6) to test the transmission mechanism:

$$M_{it}={\alpha }_0+{\phi }_{1}M+{\displaystyle \sum {\beta }_{i}contro{l}_{it}+cod{e}_i+yea{r}_t}+{\epsilon }_{it}$$

(7)

$$se{d}_{it}={\alpha }_0+{\alpha }_{1}X+{\alpha }_2{M}_{it}+{\displaystyle \sum {\beta }_{i}contro{l}_{it}}+cod{e}_i+yea{r}_t+{\epsilon }_{it}$$

(8)

In the aforementioned model, $i$ represents regional entities, $t$ represents years, $sed$ represents the level of regional SED, $X$ represents the independent variable of regional capital, $M$ is the mediating variable, $control$ is the control variable, $code$ is the unobservable fixed effect of regional entities, $year$ is the time fixed effect, ${\epsilon }_{it}$ is the random disturbance term. Referring to the research of Petersen (2009) [30], based on the two-way fixed effects model regression, “clustering” treatment is applied to the regions, thereby generating robust standard errors for the coefficient estimates.

3.3. Data

Based on the consistency of the data, this paper intends to use data from five provinces and regions in NC since 2000 for research. However, this paper found that there were unconventional changes in the data after 2021. Therefore, according to the data characteristics of FDI and SED, in order to reduce the potential impact of the COVID-19 pandemic on this study, this paper selected data from 2000 to 2020 for research. Specifically, for the spatiotemporal evolution analysis using the dynamic degree and MSAR models, provincial-level data from the five provinces (autonomous regions) in NC for the years 2000 to 2020 were used for discussion, based on the fit of the data. In order to reflect the reliability of the research conclusions when testing the driving mechanism, a sample of 26 prefecture-level cities in the five provinces (autonomous regions) of NC was adopted, totaling 231 unbalanced panel data. The data at the provincial level were mainly derived from the “China Macroeconomic Database”, with some annual data obtained through statistical bulletins on the national economic and social development of various provinces. The data from the prefecture-level cities come from the “China City Statistical Yearbook”.

4. The Evolutionary Characteristics of SED in NC

This section follows the research approach of “pattern, process, and mechanism”, introduces the measurement methods and results of SED, and, based on the numerical values of SED at the provincial level of the five provinces (autonomous regions) of NC, analyzes the spatial differences in SED in each province from a macro-perspective by the dynamic degree. Furthermore, the MSAR model is utilized to dissect the SED’s temporal change characteristics, and the spatiotemporal evolution mechanism is presented through visual maps.

4.1. Measurement of SED in NC

Current research on regional SED is still in the exploratory stage, and there is a diversity in the measurement of regional SED, with no consensus yet formed. This paper posits that for regional development, the pursuit of high-quality growth ultimately needs to be linked with regional production efficiency. In other words, the core requirement of regional SED is the steady enhancement of the region’s TFP. Therefore, this paper will utilize TFP to measure the SED of the regions. It is worth noting that, based on existing research [31], although the concept of TFP cannot fully encompass the connotations of SED, it does reflect the input–output level and innovation capacity of an individual region. The magnitude of its value is directly related to the quality of regional development. That is, the magnitude of TFP is consistent in direction and degree with the level of SED. Therefore, in the current context where the definition of SED is not clear, using TFP as a measure of SED is a good second-best-choice criterion.

This paper utilized SFA to measure the SED levels of the five provinces (autonomous regions) in NC and their prefecture-level cities. Based on the calculated values of regional SED from 2000 to 2020, combined with the quartile quantile points, specifically the 25%, 50%, and 75% quantile points, they were discretized into K categories (K = 4). The specific division criteria were as follows: (1) low level [0.113, 0.757]; (2) lower-middle level (0.757, 1.449]; (3) upper-middle level (1.449, 2.121]; (4) high level (2.121, 2.920]. For the ease of comparative analysis, this paper only presents the provincial SED levels of NC for the years 2001, 2010, and 2019 (Figure 1). According to the calculation results combined with Figure 1, the following can be known: (1) The SED status of the five provinces (autonomous regions) in NC exhibited unstable characteristics, and further, there is still some room for improvement in the level of SED in NC. (2) Shaanxi, Ningxia, and Gansu, which are closer to the central region, had a slightly better development state of their SED level than Xinjiang and Qinghai, which are in the more western areas.

Similarly, this paper measured the SED levels of prefecture-level cities in the five provinces (autonomous regions) of NC from 2000 to 2020, and after dividing according to the quartile quantile points, it presents the SED levels for the years 2001, 2010, and 2019 (Figure 2). Analysis of the figure indicates the following: (1) From a temporal perspective, compared to the years 2001 and 2019, the SED level of various cities in NC in 2010 was relatively superior. Detailed research found that between 2010 and 2013 the state of SED was at a better level. It is believed that in the post-financial crisis era, the central and local governments stimulated the economy by increasing funds, which led to an improvement in the level of SED. In other words, an increase in capital may have contributed to the enhancement of SED in NC. (2) From a spatial perspective, there was relatively little difference in the SED levels among the prefecture-level cities within the five provinces (autonomous regions) of NC and between provinces. Analysis shows that the geopolitical and economic environments of the five provinces (autonomous regions) of NC and their prefecture-level cities were relatively similar; hence, the levels of SED also exhibited convergence.

4.2. Spatial Difference Analysis Based on Dynamic Degree

Based on the dynamic degree calculation formula presented earlier, this paper calculated the SED levels of the five provinces (autonomous regions) in NC and presents the trend of the dynamic degree changes (

Table 1. SED level and dynamic degree evolution of five provinces.

Year	Shaanxi		Gansu		Qinghai		Ningxia		Xinjiang
	Values	Dynamic Degree	Values	Dynamic Degree	Values	Dynamic Degree	Values	Dynamic Degree	Values	Dynamic Degree
2000	2.07	-	2.09	-	1.56	-	1.60	-	1.92	-
2001	1.21	−41.44	0.59	−71.60	1.25	−19.93	1.13	−29.11	0.62	−67.91
2002	0.49	−38.06	0.95	−27.35	0.64	−29.46	0.68	−28.61	0.69	−31.99
2003	1.10	−15.72	1.01	−17.23	1.42	−3.09	0.93	−13.94	0.60	−22.88
2004	1.77	−3.63	1.73	−4.30	2.16	9.57	2.33	11.45	2.02	1.32
2005	1.99	−0.76	2.00	−0.86	2.92	17.32	2.55	12.04	2.19	2.73
2006	2.28	1.64	2.35	2.11	2.02	4.82	2.14	5.70	2.06	1.21
2007	1.68	−2.72	0.70	−9.46	0.72	−7.67	0.73	−7.76	0.80	−8.37
2008	0.46	−9.72	0.16	−11.57	0.36	−9.66	0.11	−11.61	0.68	−8.05
2009	0.73	−7.20	0.39	−9.04	0.65	−6.48	1.10	−3.46	0.27	−9.56
2010	1.81	−1.28	1.68	−1.97	1.45	−0.74	1.32	−1.73	1.39	−2.78
2011	2.55	2.11	2.78	3.01	2.85	7.44	2.73	6.46	2.73	3.83
2012	2.27	0.78	2.12	0.13	1.93	1.96	2.56	5.06	1.92	−0.01
2013	1.54	−1.99	1.62	−1.71	1.18	−1.88	1.21	−1.84	1.51	−1.64
2014	0.43	−5.66	0.87	−4.16	0.64	−4.21	1.11	−2.19	0.78	−4.26
2015	0.94	−3.64	0.30	−5.71	0.43	−4.83	0.51	−4.53	0.97	−3.32
2016	1.06	−3.05	1.20	−2.67	1.81	0.99	1.34	−1.01	1.72	−0.67
2017	2.74	1.91	1.90	−0.53	2.25	2.58	2.40	2.98	2.35	1.31
2018	2.71	1.71	2.73	1.71	2.71	4.06	2.67	3.75	2.72	2.31
2019	2.03	−0.10	2.16	0.18	1.44	−0.41	2.14	1.79	1.75	−0.46
2020	0.36	−4.14	1.32	−1.84	0.99	−1.84	0.76	−2.63	0.51	−3.68

). The analysis reveals the following: (1) In terms of numerical values, the differences in SED levels among the five provinces (autonomous regions) in NC were relatively small, and their SED levels did not show a continuous growth trend but instead exhibited a fluctuating pattern. (2) Regarding the dynamic degree, due to the fluctuating changes in the SED levels of each province, the dynamic degree values alternated between positive and negative, further indicating an unstable trend in the SED levels of each province. (3) Overall, the spatial differences in SED levels among the provinces were small, and they exhibited irregular temporal change characteristics.

4.3. Time-Series Change Analysis Based on the MSAR Model

This paper constructed an MSIAH(3)-AR(2) model for the SED levels of the five provinces (autonomous regions) in NC, which meant that the intercept, coefficients, and variance were regime-dependent, with a second-order lag in a three-regime model to describe its evolutionary process. Since the level of SED is a positive indicator, the three regimes were defined as the low-level regime, medium-level regime, and high-level regime (Figure 3). From the figure, the following can be seen: (1) Overall, the SED level of the five provinces (autonomous regions) in NC alternated between the three regimes, showing a fluctuating trend. (2) Stage by stage, from 2002 to 2013, the SED level alternated between the low-level and high-level regimes, which is observed to be due to the economy being in a period of rapid growth, with insufficient attention paid to the quality of development. (3) From 2014 to 2020, the SED level was always maintained in the medium-level regime, which is observed to be because China began to pay attention to the quality of economic growth during this stage, and the northwest region, due to its relatively weak economic development foundation, could only maintain a medium level of development quality in this stage.

4.4. Spatiotemporal Evolution Mechanism Analysis

First, the level of SED in NC is closely related to the macroeconomic growth of China. (1) From 2000 to 2013, during the period of rapid economic growth in China, the level of SED in NC exhibited an irregular pattern of change. From 2014 to 2020, as China’s economy entered a period of high-quality growth, the level of SED in NC gradually improved. The transformation of China’s economy from rapid growth to high-quality growth was closely related to and reflects the temporal change characteristics of the SED level in NC, shifting from irregular fluctuations to a stable state of medium-level development quality. (2) The interactive change characteristics between China’s macroeconomy and the high quality of NC indicate that, on the one hand, faster economic growth is not necessarily better, and rapid economic growth does not necessarily lead to SED. On the other hand, it confirms the correctness and necessity of the Central Committee’s judgment on the transformation from rapid economic growth to high-quality growth.

Second, the level of SED in the provinces of NC exhibited two-dimensional convergence. (1) In terms of spatial dimension, there was a general small difference among the provinces (autonomous regions), and they exhibited a synchronous change trend. On the one hand, the level of SED in each province and region fluctuated between 0.3 and 3, without any region showing a distinctive advantage. On the other hand, in specific years, the numerical differences in the level of SED among the provinces (autonomous regions) were small, such as in 2008 when the global financial crisis led to a general low level of SED in all provinces. It is believed that the spatial dimension’s change characteristics are in line with the “First Law of Geography”. (2) In terms of the temporal dimension, the level of SED in each province and region was also relatively close each year. Although the level of SED in each province and region showed an irregular pattern of change, the change characteristics among the provinces (autonomous regions) showed consistency, which was prominently reflected in the similar fluctuation trends of SED in all provinces. A comprehensive spatiotemporal distribution characteristic revealed that the SED levels of adjacent provinces were more convergent and similar, and the SED levels of different provinces at the same point in time also showed convergence.

Third, the change in the level of SED in NC was significantly influenced by marketization factors. (1) After China joined the WTO in 2001, influenced by its deepening degree of openness, China’s economy and finance were impacted, and the level of SED in NC decreased in 2002 but has shown an upward trend since then. (2) In 2008, affected by the global financial crisis, the SED in NC also regressed, but as the adverse effects of the financial crisis gradually faded, the level of SED in NC showed an improving trend. (3) After 2013, as China’s economy entered a new normal and the superimposed situation of the three periods posed new challenges to the SED in NC, the overall level of SED in the region rose amidst fluctuations.

5. The Relationship Between FDI, FC, and SED in NC

Another research objective of this paper is to explore the capital-driven factors of SED in NC, namely, to analyze the impact of FDI (external capital) and FC (internal capital) on the SED of NC from two dimensions. Based on the basic viewpoints of Marxist philosophy, external factors take effect through internal factors. Therefore, this section will discuss the process by which FDI (external capital) affects the SED of NC through FC (internal capital).

5.1. Empirical Design

5.1.1. Variable Measurement

The dependent variable is the regional SED (sed), which was measured and analyzed in the previous sections. The independent variable is FDI (fdi). Following the common practice in the existing literature, this paper will measure the independent variable by the actual investment amount of foreign capital (in USD ten thousand). However, according to the “China City Statistical Yearbook”, the measurement standards for actual foreign investment include both “the whole city” and “the urban district”. In line with the research theme, this paper intends to use the logarithm of the actual investment amount of foreign capital (in USD ten thousand) for the whole city to measure the independent variable, FDI (fdi).

The mediating variable is regional FC (fc). FC are generally studied at the micro-level of enterprises, referring to the level of external financing, and there are relatively few existing measures for regional FC. Therefore, referring to the study by Wang et al. (2022) [6], this paper uses the regional financial repression index as a proxy variable for regional FC.

Based on the relevant literature and data availability [32,33], control variables were selected as follows: (1) the informationization level (infor), measured by the ratio of the total postal and telecommunications business volume at the end of the year to the GDP; (2) industrial structure (secgdp), measured by the proportion of the value added by the secondary industry to the GDP; (3) the human capital level (hum), measured by the ratio of the number of students in regular higher education institutions to the total population at the end of the year; (4) the population agglomeration degree (popu), measured by the ratio of the total population at the end of the year to the administrative area; (5) the fixed asset investment level (fix), measured by the logarithm of the total fixed asset investment; (6) infrastructure construction (infra), measured by the ratio of the annual freight volume to the total population at the end of the year; and (7), considering the impact of lag effects and interaction effects, this paper adds the interaction term between FDI and one-period lagged regional SED (fdihtfp) as a control variable.

Table 2. Variables’ description.

Type	Variable Name	Symbol	Measurement Method
Dependent Variable	Regional SED	sed	The logarithm of the SFA-measured TFP
Independent Variable	FDI	fdi	The logarithm of the actual foreign investment amount
	FDI2	fdisq	The square term of FDI
Mediator Variable	FC	fc	Using financial repression as a proxy variable
Control Variables	Informationization level	infor	(postal business income + telecommunications business income)/GDP
	Industrial structure	secgdp	The value added by the secondary industry/GDP
	Human capital level	hum	The number of students in regular higher education institutions/total population
	Population agglomeration degree	popu	The total population at the end of the year/administrative area
	Fixed asset investment level	fix	The logarithm of the total fixed asset investment
	Infrastructure construction	infra	The annual freight volume/total population at the end of the year
	Time effect	year	The dummy variable for the year
	Regional effect	code	The dummy variable for the region

provides a description of the variables related to the impact of FDI on the regional SED.

Table 3. Descriptive statistics.

Variable	sed	fdi	fc	infor	secgdp	hum	popu	fix	infra
Mean	−0.201	0.399	0.380	0.027	49.840	182.200	−4.294	15.000	25.310
Standard Deviation	0.741	0.623	0.684	0.019	12.610	241.900	1.052	1.256	25.890
Minimum	−1.767	−1.015	−2.426	0.003	18.200	0.000	−7.663	12.270	1.273
Maximum	2.076	2.117	5.634	0.135	89.590	1261.000	−2.516	18.600	247.200
10th Percentile	−1.177	−0.429	−0.284	0.008	35.340	18.240	−5.180	13.340	5.136
25th Percentile	−0.769	−0.085	−0.022	0.016	41.360	33.120	−4.681	14.000	9.435
50th Percentile	−0.215	0.433	0.298	0.023	48.950	70.120	−4.239	14.950	17.160
75th Percentile	0.341	0.864	0.797	0.031	57.010	217.600	−3.706	15.960	28.150
90th Percentile	0.736	1.151	1.237	0.045	66.100	527.400	−3.048	16.520	60.650
Observations	231	231	231	231	231	231	231	231	231

presents the descriptive statistics of the variables related to the impact of FDI on the SED of NC. As shown in the table, the three core variables of sed, fdi, and fc exhibit normal distribution, indicating that the distribution of data is symmetric around the mean. Therefore, although the sample size of this paper is relatively small, the normal distribution characteristics of the core variables help ensure that the model estimation is effective.

5.1.2. Basic Fact Analysis

This section uses scatter plots to illustrate the relationship between FDI, FC, and SED. Figure 4 is a fitting graph of FDI and regional SED, with the horizontal axis representing FDI and the vertical axis representing regional SED. Figure 5 is a fitting graph of FDI and FC, with the horizontal axis representing FDI and the vertical axis representing regional FC. The figures indicate the following: (1) Overall, FDI shows a positive correlation with the SED of NC and a negative correlation with FC, meaning that the inflow of FDI contributes to the SED of NC and helps alleviate FC. (2) It is not clear whether the relationship between FDI and the SED of NC is linear or nonlinear. Therefore, it is necessary to further determine this through regression equations.

5.2. Empirical Results

5.2.1. Test of Driving Factors

According to

Table 4. Test on the driving factors of SED in NC.

Model	(1)	(2)	(3)	(4)	(5)	(6)
Variable	sed	sed	sed	sed	sed	sed
fdi	0.0104 *	0.1892 ***			0.1968 ***	0.1984 ***
	(0.0061)	(0.0251)			(0.0243)	(0.0243)
fdisq		−0.0117 ***			−0.0125 ***	−0.0126 ***
		(0.0016)			(0.0016)	(0.0016)
fc			0.0493 **	0.0383	0.0632 ***	0.0427 *
			(0.0182)	(0.0291)	(0.0165)	(0.0255)
fcsq				0.0036		0.0068
				(0.0075)		(0.0064)
control variables	YES	YES	YES	YES	YES	YES
year	YES	YES	YES	YES	YES	YES
code	YES	YES	YES	YES	YES	YES
observations	231	231	231	231	231	231
R-squared	0.984	0.987	0.984	0.984	0.988	0.989

Note: *, **, ***, respectively, indicate significance at the 10%, 5%, and 1% statistical levels. The values in parentheses represent the robust standard errors of the regression coefficients. The same format will be used for subsequent tables.

, firstly, Models (1)–(2) report the regression process of the linear and squared terms of FDI. The results show that FDI inflow presents an inverse U-shaped relationship with the SED of NC, indicating that FDI inflow has a non-uniform characteristic of first promoting and then inhibiting the SED of NC. Thus, hypothesis H1 is proven. Secondly, Models (3)–(4) report the regression process of the linear and quadratic terms of FC. The results show that only the linear term has a good estimation result, indicating that FC have a promoting effect on the SED of NC. Thirdly, Models (5)–(6) report the comprehensive regression results of FDI and FC. The comparison finds that Model (5) performs better, further confirming the positive impact of FC on the SED of NC. Integrating the regression results from Table 4, both FDI and FC significantly impact the SED of NC. This indicates that external capital and internal capital can effectively drive the SED of NC. However, the impact of FDI is nonlinear, while the impact of FC is linear.

The stepwise regression results (Table 4) show that the significance and the magnitude of the coefficients for FDI and its squared term, as well as the FC term, remain relatively stable, indicating that the model has a certain degree of robustness. However, to ensure the reliability of the driving factor baseline regression results, this paper also conducted a series of robustness and endogeneity tests (

Table 5. Robust test of driving factors for SED in NC.

Model	(1)	(2)	(3)	(4)
Variable	sed	sed	sed	sed
fdi	0.1639 **	0.1526 ***	0.1798 ***	0.1677
	(0.0609)	(0.0234)	(0.0238)	(0.1390)
fdisq	−0.0117 ***	−0.0096 ***	−0.0115 ***	−0.0107
	(0.0034)	(0.0015)	(0.0015)	(0.0087)
fc	0.0734	0.0553 ***	0.0512 **	0.0616 ***
	(0.0468)	(0.0153)	(0.0162)	(0.0182)
control variables	YES	YES	YES	YES
year	YES	YES	YES	YES
code	YES	YES	YES	YES
observations	112	221	231	231
R-squared		0.990	0.984	0.587

Note: **, ***, respectively, indicate significance at the 5%, and 1% statistical levels. The values in parentheses represent the robust standard errors of the regression coefficients.

): Firstly, re-estimation using a different estimation method was conducted. Model (1) was re-estimated using the dynamic GMM regression model, and the variables remained relatively stable. Secondly, control variables were added. In Model (2) of Table 5, the logarithm of the actual GDP was added as a control variable, and the overall regression level did not change significantly, still showing robustness. Thirdly, the explained variable was replaced. In the baseline regression, the TFP representing SED was measured by the regional GDP as the output variable, with capital stock and total employment as input variables. In Model (3) of Table 5, the SED indicators were measured by the regional GDP as the output variable, with the capital stock and the number of urban employed personnel as input variables, and the estimation results were still robust. Fourthly, re-regression using two-stage least squares with instrumental variables was conducted. In this paper, the “average of FDI in all years except the current year” was chosen as the instrumental variable for the 2SLS estimation. The results can be seen in Model (4) of Table 5. Compared with the baseline regression results, endogeneity issues have a relatively minor impact on the relevant conclusions of this paper, which also indicates that the panel two-way fixed results have a certain degree of robustness.

The empirical results show that the asymmetric inverse U-shaped relationship between FDI and the SED of NC indicates that the inflow of FDI can effectively promote the SED of NC in the initial stage, but when the inflow of FDI reaches a higher level, it is detrimental to the SED of NC, meaning that FDI has a “first-enhancing-then-inhibiting” effect on the SED of NC. The analysis suggests the following: in the initial stage, the inflow of FDI not only compensates for the capital gap caused by the insufficient supply of funds in NC but also promotes the optimization and upgrading of the industrial structure in NC through international trade, thus fostering the SED of NC in multiple dimensions. Moreover, the inflow of FDI is often accompanied by technological spillovers, which play a significant role in promoting technological progress in NC. Technological progress is a prerequisite for regional TFP and SED. In addition, technological spillovers lead to the optimization of resource allocation, the growth of industrial output, and the transformation of economic growth methods, providing the necessary economic foundation for the SED of NC. Therefore, the inflow of FDI promotes the SED of NC through capital inflow, international trade, and spillover effects in the initial stage. In the middle and later stages, the inflow of FDI competes with northwest Chinese enterprises for factors and product markets, and the intensification of competitive pressure makes FDI detrimental to the SED of NC. Furthermore, in general, foreign investors are unwilling to transfer technology and experience to the host country. As a result, when NC introduces FDI, it may face technological protection from foreign investors, creating barriers to technology transfer, making the positive spillover effects of FDI less apparent. Even in the middle and later stages, the level of productivity in NC may not be able to match the spillover effects of FDI inflow, thereby suppressing the SED of NC.

In addition, the empirical results also indicate that FC contribute to the SED of NC in the current stage. Generally speaking, most existing literature believes that FC, resulting from the “difficulty and high cost of financing”, have a negative impact on the SED of the regional economy. However, this paper finds that FC have a positive promoting effect on the SED of NC. The analysis suggests the following: (1) although FC suppress the channels of capital funding for the economic development of NC, they stimulate the competitive potential of the region. Enterprises within the region, in order to survive, strengthen their innovation capabilities, concentrate limited funds to tackle technical challenges, and promote the SED of enterprises through the innovation of research and development, thereby achieving the SED of NC. (2) Under the influence of FC, the economic and financial strength of the region cannot guarantee the capital conditions required for economic growth fields. Enterprises within the region can optimize internal management and intensive production paths to improve the utilization rate of funds. In the short term, they improve their own level of SED through non-financing means, thereby achieving the overall SED of NC. (3) To a certain extent, FC are an internal institutional arrangement at the government level, and this institutional arrangement is to take into account the overall situation to ensure the steady progress of regional SED from a macro-perspective. Therefore, it is not surprising that FC have a positive promoting effect on the SED of NC.

5.2.2. Test of Impact Mechanism

The aforementioned analysis indicates that both external and internal capital factors can effectively influence the SED of NC. However, further consideration reveals that FDI inflows may affect the SED of NC by impacting regional FC. Therefore, this section will verify the existence of the transmission chain through mediation effect testing. The dependent variable is the SED of NC, the independent variable is FDI inflow, and the mediating variable is FC. In

Table 6. A test of the intermediary effect of FDI on the SED of NC through FC.

Model	(1)	(2)	(3)
Variable	sed	fc	sed
fdi	0.1892 ***	−0.1204	0.1968 ***
	(0.0251)	(0.1094)	(0.0243)
fdisq	−0.0117 ***	0.0138 **	−0.0125 ***
	(0.0016)	(0.0070)	(0.0016)
fc			0.0632 ***
			(0.0165)
control variables	YES	YES	YES
year	YES	YES	YES
code	YES	YES	YES
observations	231	231	231
R-squared	0.987	0.721	0.988

Note: **, ***, respectively, indicate significance at the 5%, and 1% statistical levels. The values in parentheses represent the robust standard errors of the regression coefficients.

, Models (1) to (3) report the mediation effect test results of FDI affecting the SED of NC through FC, and the analysis is as follows: Firstly, in the three regression equations, the coefficients of both the core independent variable and the mediating variable are significant at the 5% level. According to the mediation effect test process, it is considered that FC are the mediating variable for FDI to affect the SED of NC, and FC play a strong concealing role in the spillover effect of FDI on the SED of NC. That is to say that the indirect effect is in the opposite direction to the direct effect. In other words, FDI inflows affect the SED of NC by influencing FC, and this transmission channel exists. Thus, hypothesis H2 is proven. Secondly, the squared term of FDI has a significant positive impact on FC, indicating a U-shaped relationship between FDI and FC. The analysis suggests that initial FDI inflows effectively alleviate FC, but when the inflow of FDI is excessive, the spillover effect of FDI shows diminishing marginal utility, and the adverse impact of FDI on FC gradually emerges.

The empirical results of the impact mechanism test show that FC serve as an intermediary variable in the process by which FDI affects the SED of NC. The following is observed: (1) After the initial inflow of FDI, in the direct impact area of capital supply, FDI flows in through forms such as joint ventures and mergers and acquisitions, increasing the capital supply in NC and effectively reducing FC there. Moreover, the introduction of FDI can force the reform of regional financial markets, improve the financing environment distorted by policies, alleviating FC in NC and thus influencing its SED. In addition, in the indirect spillover areas such as technology transfer and management experience, enterprises in NC can improve their own operational performance, corporate reputation, and asset levels by imitating and learning under the spillover effect of FDI, enhancing their internal financing capabilities, which is conducive to alleviating FC and further influencing the SED of NC. (2) As FDI continues to flow in, local governments in NC, in order to continuously attract investment, provide excessive preferential policies to FDI enterprises, thereby exerting a squeeze effect on the financing of local enterprises in NC, intensifying the degree of local FC. On the other hand, in the product market, FDI enterprises produce more competitive products through advanced technology and other advantageous conditions, occupying the market share of local enterprises in NC, compressing their profit margins, and thus reducing the capital sources of local enterprises, increasing FC. Under the dual influence of these factors, the continuous inflow of FDI exacerbates FC in NC and continues to affect the SED of NC.

6. Discussions

This study not only explores the spatial pattern distribution, spatiotemporal evolution process, and spatiotemporal evolution mechanism of SED in NC, but also empirically tests the driving factors of SED from the perspectives of FDI and FC and examines the relationship between FDI, FC, and SED. Therefore, this paper discusses the research results as follows.

Firstly, how should we view the nonlinear relationship between FDI and SED in underdeveloped regions in developing countries? This article finds that FDI inflows have an asymmetric, inverted, U-shaped relationship with SED in NC, but existing research mainly focuses on the linear impact of FDI on SED, including positive and negative effects [34,35]. Therefore, this study expands the research on the relationship between FDI and SED. In addition, for policy makers in developing countries, central governments need to control the valve of cross-border capital flows, maintaining a relatively controllable scale and flow rate. When introducing foreign capital, they should guide foreign investment to tilt towards underdeveloped areas as much as possible to weaken the unbalanced development caused by differences in foreign capital inflow. In terms of local governments in underdeveloped areas, they should reasonably formulate policies to introduce foreign capital with appropriate policy preferences and actively obtain production factors with core competitiveness represented by advanced technology to enhance their TFP, thereby achieving SED in underdeveloped areas in developing countries.

Secondly, how should we view the role of FC in the impact of FDI on SED in underdeveloped regions in developing countries? This study has found that FDI inflows can indeed alleviate FC in NC, but the existence of FC is also beneficial to the SED of NC, indicating that neither a lower number nor a higher number of FC are better. Previous studies generally believe that FC are an unfriendly variable that are detrimental to SED [36]. Therefore, this study expands the relevant research on the economic effects of FC and confirms that appropriate FC can help alleviate the negative effects caused by other insufficient factors. The specific inspiration for the SED of underdeveloped areas in developing countries is that it needs to maintain an appropriate level of FC and should not blindly pursue few FC. The appropriateness of FC should be judged in conjunction with the SED of the economy, that is, whether the current FC have stimulated the endogenous growth dynamics of the regional economy and promoted SED.

In summary, this study takes NC as the research sample, reaffirming the important role of capital factors in SED and providing new insights into the impact of FDI on SED and FC on SED. These findings have significant implications for policy makers in underdeveloped regions in developing countries, especially in the current context of SED. This study provides guidance on how to utilize capital to promote SED.

It is worth noting that although this study has important research significance, it must be acknowledged that it has certain limitations. First of all, the data period of this study was 2000–2020, and the potential impact of the COVID-19 pandemic on this study was not fully considered, so the time period of the data may have brought bias to the results. The conclusions of this study need to be further explored and expanded. Secondly, this paper only focuses on NC. Although the research results provide important insights, they may not be fully applicable to other countries and regions with different regional characteristics. Thirdly, this study uses regional TFP to measure SED, which may not fully capture the connotations and time-varying characteristics of SED. Therefore, the measurement of SED needs further optimization. Nevertheless, this study lays a solid foundation for understanding the capital-driven factors of SED and provides valuable avenues for future research to address these limitations.

7. Conclusions, Implications, and Prospects

This paper closely examines the relationship between FDI, FC and the SED of NC. Based on a review of the existing literature, it attempts to answer the following questions through mechanism analysis and empirical testing: What is the current state of SED in NC? What is the impact of FDI (external capital) and FC (internal capital) on SED? In the context of a global economic situation turning point, where should the SED of underdeveloped areas in developing countries head? Based on the above issues, this paper analyzes the spatiotemporal evolution characteristics of SED using data from NC from 2000 to 2020 and empirically tests the impact of FDI and FC on SED. The specific conclusions are as follows: Firstly, the levels of SED among the provinces in NC show small spatial differences and exhibit irregular temporal changes. Secondly, there is an asymmetric, inverse, U-shaped relationship between FDI and the SED of NC, indicating that the inflow of FDI can effectively promote the SED of NC in the initial stage. However, when the inflow of FDI reaches a higher level, it is not conducive to the SED of NC, that is, FDI has a “rise and then suppress” effect on the SED of NC. Thirdly, FC have a mediating effect in the process of FDI influencing the SED of NC, that is, FDI affects the SED of NC by acting on FC.

The research conclusions of this paper indicate that the SED of the regional economy is the result of the joint action of various factors; it is influenced not only by capital elements such as FDI and FC but also by regional characteristics such as the level of informatization, industrial structure, and the level of human capital. Therefore, this paper believes that under the background of the global economic situation turning point, underdeveloped areas in developing countries should properly handle the following issues: they should control the speed and direction of FDI flows, maintain an appropriate level of FC, concentrate advantageous funds to develop advantageous industries and regions first, and promote SED from multiple aspects. Specifically, underdeveloped regions in developing countries should seize the strategic opportunities of the global economic situation’s turning point in line with their own personalized characteristics. Under the guidance of the market and with government coordination, they should improve their economic and social environment, introduce and utilize foreign capital rationally, maintain appropriate FC for enterprises, and address the alleviation of FC in a solid and orderly manner, thus promoting SED from multiple angles.

This study has some limitations, including a narrow time range of data used, the uniqueness of the sample, and the singularity of the indicator measurement. Therefore, in the future, while ensuring stable and accessible data, we will expand the time frame of this study, attempt to extend the research area to underdeveloped regions in more developing countries, and adopt multiple dimensions to measure SED in order to deeply reveal the relationship between FDI, FC, and SED and provide a more reliable theoretical basis for policy makers.