Spatiotemporal Evolution and Influential Factors of Rural Poverty in Poverty-Stricken Areas of Guizhou Province: Implications for Consolidating the Achievements of Poverty Alleviation

Abstract

China has made remarkable reductions in absolute poverty. However, pressing questions remain of how to consolidate the existing achievements of poverty alleviation and prevent rural households from regressing back into poverty, especially in continuously poverty-stricken areas. This paper explores the spatiotemporal evolution of rural poverty and its influential factors under different poverty standards in three continuously poverty-stricken areas in Guizhou Province using 2003–2017 panel data and the spatiotemporal analysis method. The results show that decreasing poverty is an apparent spatial heterogeneity and there are area differences in the different research periods. The degrees of the average annual poverty reductions of all research counties were between 2.98–10.09%, 10.60–26.96%, and 11.46–43.19% in 2003–2007, 2008–2010, and 2011–2017, respectively; the poverty-stricken counties with high rates of poverty reduction are often adjacent to the nonpoverty-stricken counties. Influential factors vary in different areas over time, which is the result of the combination of leading influential factors (economy) and various influencing factors (natural location, social development, and education/labor) with regional characteristics. Although the effect of natural location on rural poverty in the research area is decreasing, its impact should not be ignored. Considering the complexity of poverty’s influential factors and the dependence on poverty alleviation policies, it’s essential for China to continue to strengthen its targeting of poverty in the continuously poverty-stricken areas, especially the counties in the inner areas. Devoting to building the coordination of regional development and ecological environment protection is an important way to achieve sustainable development goals with higher quality in the continuously poverty-stricken areas.

1. Introduction

Ongoing concerns for sustainable development have risen over the years since the millennium development goals (MDGs) formulated 8 sustainable development goals in 2000, and 17 sustainable development goals (SDGs) formed the 2030 Development Agenda adopted in 2015. Poverty alleviation is the foremost agenda of the sustainable development goals in many countries; meanwhile, “end poverty in all its forms everywhere” has been the greatest global challenge [1,2]. Approximately 80% of poor people live in rural areas across the world, which makes reducing rural poverty a crucial development target and the common mission of humanity [3]. As the economy rapidly develops and poverty alleviation policies are implemented, China has made remarkable achievements in poverty alleviation. In addition, China entered a new era, with its elimination of absolute rural poverty under current poverty standards (the new poverty line of 2300 RMB per year based on 2010 constant prices) at the end of 2020 [4]. However, this situation does not mean that China has completely lifted itself out of poverty or solved all the problems of regional poverty; Chinese society still faces many challenges—including housing, education, employment, and medical care—in alleviating poverty and consolidating the achievements of poverty alleviation [4,5,6]. Approximately 577 million people in China’s rural areas will remain in relative poverty for a considerable time [6]. Moreover, the poverty-stricken populations are mainly distributed in continuous poverty-stricken areas, which are deep mountain areas, ethnic minority areas, ecologically fragile areas, areas with inadequate public facilities and public services, and even some areas with a high incidence of noncommunicable diseases [7,8]. At the same time, the current COVID-19 pandemic certainly poses a challenge to the process of poverty alleviation. Global extreme poverty is expected to increase in 2020 for the first time in over 20 years to as many as 150 million by 2021, according to the World Bank. Therefore, research on the spatiotemporal evolution and mechanism of rural poverty remains important and necessary, especially for preventing rural households who have emerged from poverty from regressing back into it, consolidating the achievements of regional poverty alleviation, and eliminating all forms of poverty in China’s post-poverty alleviation era.

Previous studies have been conducted on the many factors that influence poverty, confirming the interrelatedness and downward or upward spiraling processes among them, combined with different theoretical and quantitative analysis models. The results of some of these studies have evidenced the contribution to alleviating poverty of agricultural cooperatives [3], foreign direct investment [9], infrastructure investment [10], social capital [11], land-based investment [12], economic diversification, tangible and intangible assets (e.g., education and health) [13], public expenditures on early childhood education and care [14], improvements in human well-being and income [15], land transfer intensity [4], labor transfer [16], growth and income inequality [17], infrastructure improvement and residents’ living standards [18], ecosystem services value [19], mobile internet use [20], land liquidation and consolidation [6,21], technology, financial support and development [22,23], mobility and transport accessibility [2,15,24], and vegetation vitality [25]. Of the other solutions that have been extensively researched, priority has been given to policies that degrade land [26,27] and provide insufficient geographical capital [28,29], which are bad for poor people.

Targeted poverty alleviation is a complex system engineering approach, just as poverty is a complicated multidimensional phenomenon. Specifically, affected by various factors—such as economic development, resource endowment, natural factors, public facilities, education, medical care, and poverty alleviation policies—the dynamic change process of rural poverty has presented regional heterogeneity [30,31]. However, of particular interest in most of these studies is any single factor or a category of influential factors that affect the poverty alleviation goals. Local cases are lacking in the analyses of the multidimensional impact mechanism of long-term series, especially because poverty standards are constantly changing, which is key to regions emerging from poverty as a sustainable development goal.

Regional poverty is an external manifestation of the non-equalizing nature of a –human–society system caused by many influential factors, which shows an unbalanced distribution in China. Fourteen continuously poverty-stricken areas were designated by the Chinese government as the main battleground for poverty alleviation from 2011 to 2020 [32]. Most of the poverty-stricken counties were determined based on natural geographic and socioeconomic conditions. In addition, approximately 70% of those counties are limited by severe topographic conditions (average slope exceeding 10°) [33,34]. In many mountainous areas, the remote geographical location is an important factor affecting residents’ ability to enjoy services and infrastructure and is generally considered the main reason for high poverty [35,36]. Generally, poverty in poverty-stricken counties is impacted by natural, economic, and social elements. Unfortunately, the impact of some elements is difficult to overcome, which also determines the difficulty and uncertainty of rural revitalization after poverty alleviation in such regions. Complex and multiple influential factors could shape the poverty level among the counties, which means that the spatial distribution characteristics of different continuously poverty-stricken areas differ notably. A clear understanding of the spatiotemporal patterns of rural poverty and its influential factors in continuously poverty-stricken areas could help eliminate all forms of regional poverty so the country can enter a new era of sustainable regional development. Fortunately, the spatiotemporal analysis method [29,37] makes it possible to reveal the evolution of the characteristics and mechanisms of regional poverty.

In this context, this study examines the contiguous poverty-stricken areas in Guizhou Province to understand the spatiotemporal evolution of rural poverty and the influential dimensional factors on the effect of rural poverty under different standards using a spatiotemporal analysis and suggests efficient solutions for those regions to consolidate the achievements of regional poverty alleviation. The remainder of the study is organized as follows. Through a brief review of the previous literature, Section 2 demonstrates background information on the study area, data sources, and theoretical framework for the spatiotemporal dynamics of rural poverty and influential factors. The empirical analyses are presented in Section 3. Section 4 discusses the formation mechanism of rural poverty in the contiguous poverty-stricken areas. A summary of the findings and limitations of this study, as well as future research directions, are offered in Section 5.

2. Materials and Methods

2.1. Study Area

Sixty-five counties in Guizhou Province that belonged to three contiguous poverty-stricken areas (Wuling, Wumeng, and the rocky desert area of Dian–Gui–Qian) were selected as the study area (Figure 1). Among them, 15 counties were distributed in the Wuling area, 10 counties belonged to the Wumeng area, and 40 counties were in the rocky desert area of Dian–Gui–Qian. In addition, poverty degrees and causes are different for three contiguous poverty-stricken areas. Guizhou Province is located in Southeastern China and is between 103°36′–109°35′ E and 24°37′–29°13′ N. It covers an area of 17.62 × 10⁴ km², accounting for about 1.8% of the total area of China. The geomorphic types of Guizhou Province are mainly plateau, basin, and hill, of which the area of mountain and hill accounts for 92.5% of the total area of the province, and the average altitude is about 1100 m. In particular, Guizhou Province is in the most concentrated fragile karst area among the three major karst areas in the world [8]. Due to the ecologically fragile and mountainous natural environment, consequent land degradation, and underdeveloped infrastructure, Guizhou had the largest number and the most extensive and deepest poverty level of rural poor people in China: 7.45 million of the rural poor population were distributed in this province in 2013 [29,34]. Therefore, Guizhou Province is typified by poverty, ecological fragility, and underdevelopment. The poverty-stricken counties are representative and have commonalities (the distribution of poverty-stricken populations: central and western regions of China, mountainous areas, and ethnic minorities concentrated areas) with other areas. Based on the poverty standards formulated by the Chinese government in 2004, 2008, and 2011, this paper divides the research period into three stages: 2003–2007, 2008–2010, and 2011–2017.

2.2. Data Sources and the Selection of the Influential Factors

Poverty incidence (PI), which is the ratio of the poor to the total population in a region and is the most widely used measurement index for poverty, was used to characterize regional poverty. It reflects the breadth of regional poverty and the scale of the poor population. The original PI data and influential factors included social and economic statistics and geographical data that were obtained from different sources from 2003 to 2017. (1) The PI data for the 65 poverty-stricken counties were obtained from the Statistical Yearbook of Guizhou Province and the Leading Group Office of Poverty Alleviation and Development of Guizhou Province. (2) The data of the digital elevation model (DEM) was downloaded from the Geospatial Data Cloud website (http://www.gscloud.cn, accessed on 23 August 2022) with a spatial resolution of 30 m. (3) The annual normalized difference vegetation index (NDVI) was acquired from the Resource and Environment Science and Data Center (http://www.resdc.cn/DOI, accessed on 23 August 2022). (4) The geographic coordinates for the city and county government were derived from the Baidu map. (5) The social and economic statistics were collected from the Statistical Yearbook of Guizhou and China County Statistical Yearbook.

The establishment of a relatively comprehensive framework of influential factors may be an effective, sustainable development instrument to prevent relapses into poverty, especially when the policy- and decision-making influence in the value chain is changing. In light of the study area characteristics, all the influential factor selections and their definitions and measurements are consistent with the previous literature and expert knowledge. The influential factors that were explored in this study include the natural endowment, location, economy, education, labor capital, and social development (

Table 1. Description of poverty influential factors in continuous poverty-stricken areas in Guizhou.

Influential Factors	Indicator	Abbreviation
Natural endowment	Elevation	P1
	Slope	P2
	Ratio of slope areas above 15°	P3
	Ratio of slope areas above 25°	P4
	Ratio of slope areas above 30°	P5
	Topographic relief	P6
	NDVI	P7
Location	Distance to provincial capital city center	P8
	Distance to the city center	P9
Economy	Per capita regional gross domestic product (GDP)	P10
	Ratio of output value of primary industry in GDP	P11
	Ratio of output value of secondary industry in GDP	P12
	Ratio of output value of tertiary industry in GDP	P13
	Per capita public revenue	P14
	Per capita public expenditure	P15
	Per capita net income of peasants	P16
	Per capita grain output	P17
	Per capita retail sales of social consumer goods	P18
	Per capita deposit balance of financial institutions	P19
Education	Per capita education expenditure	P20
	Ratio of students to teachers in junior and senior high school	P21
	Ratio of students to teachers in primary school	P22
Labor capital	Rural employees in primary industry as ratio to total rural employees	P23
	Rural employees in secondary industry as ratio to total rural employees	P24
	Rural employees in tertiary industry as ratio to total rural employees	P25
Social development	Per capita investment in fixed assets	P26
	Number of junior and senior high schools per ten thousand people	P27
	Number of primary schools per ten thousand people	P28
	Number of fixed telephone users per ten thousand people	P29
	Number of industrial enterprises per ten thousand people	P30
	Number of beds in medical and health institutions per ten thousand people	P31
	Number of social welfare adoptive agencies per ten thousand people	P32
	Number of beds in social welfare adoptive agencies per ten thousand people	P33
	Total power of agricultural machinery	P34

). Natural endowment, composed of natural resources, congenital advantages, and other determinants, the most direct and insurmountable influential factor, plays an important role in agricultural development, especially in continuously poverty-stricken areas [2,29]. The location represents the nearness of each poverty-stricken county to more developed areas, which reflects the external contact capacity of the counties. The economy includes indicators that are relatively more direct factors for judging regional poverty and embodies the counties’ overall development [2,37]. Various studies have shown that education is one of the most important dimensions of poverty reduction, and some of the poverty-stricken counties have disadvantaged education conditions, which is also the main factor affecting the intergenerational transmission of poverty [11,38]. Poverty is closely connected to income, and poor people can drastically improve their living standards by participating in agricultural and nonagricultural activities to secure higher incomes [16]. In addition, regional development relies on a lack of an effective labor supply; thus, sufficient labor capital is conducive to reducing poverty. Social development, which reflects the basic standards of infrastructure construction and social welfare institutions, has been proven to play a crucial role in poverty alleviation [11,18]. These influential factors directly or indirectly shape and affect regional poverty, so regional rural poverty exhibits a variety of complex characteristics. Generally, these influential factors cover most of the indicators that are considered poverty causes in much of the literature [2,6,8,29].

2.3. Methodology

2.3.1. Spatiotemporal Analysis of Rural Poverty

Rural poverty is a dynamic process that demonstrates many different characteristics that change over time and space, which reflect changes in poverty alleviation policies. Therefore, the Mann–Kendall (MK) statistical test method, local indicators of spatial association (LISA) analysis, and one-way analysis of variance (ANOVA) were used in this paper to explore the spatiotemporal dynamics of rural poverty in the poverty-stricken counties in Guizhou Province.

(1) Mann–Kendall (MK) statistical test;

The MK test is widely applied to assess whether time series data exhibit an upward or downward trend [39,40]. The samples tested in this method do not need to follow a specific distribution and can be suitable for nonnormally distributed data without the interference of a few outliers. Assuming that the PI value of the poverty-stricken counties is (x1, x2, …, xn), test statistic S is given by Equation (1):

$$S={\displaystyle \sum _{i=1}^{n-1}{\displaystyle \sum _{j=i+1}^n\mathrm{sgn}(x_j-x_i)}}$$

(1)

$$\mathrm{sgn}(x_j-x_i)=\{\begin{array}{l}1x_j-x_i>0\\ 0x_j-x_i=0\\ -1x_j-x_i<0\end{array}$$

(2)

where n is the length of the time series of the PI data. When n ≤ 10, the significance value can be obtained from a tabulation; when n >10, then the variance of S can be defined as $\mathrm{var}(S)=n(n-1)(2n+5)/18$. The equation of the standardized statistic Z_MK is as follows:

$$Z_{MK}=\{\begin{array}{l}(S-1)/\sqrt{\mathrm{var}(S)}S>0\\ 0S=0\\ (S+1)/\sqrt{\mathrm{var}(S)}S<0\end{array}$$

(3)

If $\left|Z_{MK}\right|>Z_{1-\alpha /2}$ at a significance level of α, then a significant trend exists in the time series PI data. In particular, a positive value of Z_MK indicates a markedly upward trend and vice versa.

(2) Local indicators of spatial association (LISA);

Local spatial autocorrelation analyses can reflect the spatial variation and heterogeneity of geographical objects; therefore, a LISA analysis was chosen to explore the local spatial differences in the value of the average annual poverty reduction (AAPR) under three different research periods [41,42]. For the LISA analysis, all of the AAPR values were categorized into four classes in the local Moran’s I index across the poverty-stricken counties: high–high (H-H) and low–low (L-L), representing high (low) AAPR values surrounded by high (low) AAPR values, and high–low (H-L) and low–high (L-H), representing high (low) AAPR values surrounded by low (high) AAPR values.

(3) One-way analysis of variance (ANOVA).

An ANOVA, a statistical method widely used to test significant differences between treatment groups [43], was applied to explore the differences between the 3 continuously poverty- and nonpoverty-stricken areas in Guizhou Province. This method calculates the F statistic using statistical inference, and then the F test is performed. The sums of the squared variables (S_T) consist of two parts: the sums of the squared variations (S_A) between groups caused by the control variables, and the sums of the squared error in the group caused by the random variables.

$$\begin{array}{l}{S}_A={\displaystyle \sum _{i=1}^k{n}_i}{({\overline{x}}_i-\overline{x})}^2\\ S_E={\displaystyle \sum _{i=1}^k{\displaystyle \sum _{j=1}^{n_i}{(x_{ij}-{\overline{x}}_i)}^2}}\\ S_T=S_A+S_E\end{array}$$

(4)

where k is the horizontal number and n_i is the number of tests performed at the ith level. The formula of F statistic is:

$$F=\frac{\frac{S_A}{k-1}}{\frac{S_E}{n-k}}$$

(5)

where n represents the total number of tests.

2.3.2. Geographically and Temporally Weighted Regression (GTWR)

Given the need to explore the spatiotemporal patterns of the influential factors in the continuously poverty-stricken areas in Guizhou Province, the GTWR method was used for a time series analysis. The GTWR has been widely applied to study spatiotemporal heterogeneity in different areas and can more directly (vs. other methods) demonstrate the geostatistical relationship between factors [44]. Specifically, as a spatiotemporal nonstationary regression model with a weighted matrix integrating spatial and temporal information, a GTWR can effectively solve the problem of considering time and space in heterogeneity at the same time. The GTWR equation is detailed as follows:

$$Y_i={\beta }_0(u_i,v_i,t_i)+{\displaystyle \sum _k{\beta }_k(u_i,v_i,t_i)X_{ik}}+{\epsilon }_i$$

(6)

where $(u_i,v_i,t_i)$ represents the space–time geographical coordinates of county i, ${\beta }_0(u_i,v_i,t_i)$ represents the regression constant corresponding to the same county, ${\beta }_k(u_i,v_i,t_i)$ is the regression coefficient of the kth explanatory variable of county i, $X_{ik}$ represents the values of the kth explanatory variable of county i, and ${\epsilon }_i$ is the error term of the model function. The estimated parameter can be calculated as follows using a locally weighted least squares estimation.

$$\widehat{\beta }(u_i,v_i,t_i)={\left[X^{T}W(u_i,v_i,t_i)X\right]}^{-1}{X}^{T}W(u_i,v_i,t_i)Y$$

(7)

where $W(u_i,v_i,t_i)=diag({\alpha }_{i1},{\alpha }_{i2},\cdots ,{\alpha }_{in})$ is the diagonal matrix, and n is the number of observations. All the calculations were performed in MATLAB 2016b. A tenfold cross-validation was used to obtain the best bandwidth. Due to regional and periodical differences, the effect between the factors was not static and similar. Therefore, the selection of influential factors with the help of a stepwise regression analysis cannot be ignored before the inquiry analysis. A schematic diagram of the variance decomposition of two and three categories of variables is shown in Figure 2. Taking the variance decomposition of two categories of variables as an example, for dependent variable Y, the variable’s interpretation includes: the independent influence part [a] and [c] of two explanatory variables X1 and X2, the combined influence part [b], and the unexplained residual variance part [d]. Therefore, variable influence can be solved by the following formula, where, R² represents the adjustment coefficient of the GTWR model. The variance decomposition of the three categories of variables is similar. The calculation method for three categories of variables is similar to this.

$$\{\begin{array}{l}a+b=R_a^2\mathrm{of}(\mathrm{Y}\sim \mathrm{X}1)\\ b+c=R_b^2\mathrm{of}(\mathrm{Y}\sim \mathrm{X}2)\\ a+b+c=R_{ab}^2\mathrm{of}\left[\mathrm{Y}\sim (\mathrm{X}1,\mathrm{X}2)\right]\\ 1-a-b-c=R_c^2\mathrm{of}\mathrm{Residuals}\left(\mathrm{R}\right)\end{array}$$

(8)

2.3.3. Variance Decomposition

Variance decomposition [45], a classical statistical method used to uncover simplifying structures in a large set of factors for multivariate analysis, was applied to explore the interpretive degree of various influential factors on rural poverty in the continuously poverty-stricken areas in Guizhou Province. The implementation of a variance decomposition mainly includes two steps: (1) a regression analysis is performed on a single factor or combined factors separately; (2) the net influence and the combined influence of various factors are acquired from the single or multiple regression.

3. Results

3.1. Spatiotemporal Dynamics of Rural Poverty

3.1.1. Analysis of the Change in Rural Poverty over the Years

Figure 3 shows the PI of the poverty-stricken counties tested with the MK statistical test in 2003–2007 and 2011–2017. A significant downward trend was observed for most of the poverty-stricken counties in the two research periods, indicating that most of the poverty-stricken counties obviously reduced their rural poverty. The PI (10.5%, 11.7%, 11.3%, 10.7%, and 9.3%) in Taijiang County first increased and then decreased from 2003 to 2007, which indicates that the poverty alleviation effect was not significant during this research period. The change in rural poverty in poverty-stricken counties is evidence of the effectiveness of regional poverty alleviation, which also serves as a reference for the poverty reduction strategy of “implementing policies according to the county”. These results indicate that although the poverty standards and poverty alleviation policies in different research periods are changing, the rural poverty of most of the poverty-stricken counties in Guizhou Province showed a significant decreasing trend from 2003 to 2017.

3.1.2. Spatial Pattern of the Rural Poverty Decreasing

To further reveal the spatiotemporal evolution of rural poverty in the poverty-stricken counties, the rate of the average annual poverty reduction was introduced to demonstrate the degree of poverty reduction in the three research periods, and a LISA analysis was used to explore the spatial autocorrelation of the poverty reduction based on GeoDa software. Figure 4 shows the decrease in poverty exhibited different characteristics in the different research periods. The poverty-stricken counties had degrees of average annual poverty reduction between 2.98–10.09%, 10.60–26.96%, and 11.46–43.19% in 2003–2007, 2008–2010, and 2011–2017, respectively. Specifically, more than half of the poverty-stricken counties had rates of average annual poverty reduction that ranged from 7.04% to 8.41% in 2003–2007, the counties with “L-L” levels were located in southeastern Guizhou Province, and the “H-H” counties were scattered in the north-central part of the rocky desert area of the Dian–Gui–Qian area in Guizhou Province. Most of the poverty-stricken counties had rates of average annual poverty reduction that ranged from 12.73% to 16.97% in 2008–2010, and the “L-L” counties were distributed southwest of the rocky desert area of Dian–Gui–Qian in Guizhou Province, which indicated a low degree of poverty reduction. The spatial heterogeneity of the decrease in poverty in the poverty-stricken counties was obvious from 2011 to 2017, with the “L-L” counties mainly distributed in southeastern Guizhou Province and the “H-H” counties mostly located in northeastern Guizhou Province. By comparison, the three research periods demonstrated an increasing trend of average annual poverty reduction, with an increasing gap among the poverty-stricken counties. In addition, the poverty-stricken counties close to the nonpoverty-stricken counties tend to have greater poverty reduction.

3.1.3. Difference Analysis of Three Continuous Poverty-Stricken Areas

China has 14 continuously poverty-stricken areas designated by the Chinese government in 2011 according to poverty measurement indicators (e.g., per capita net income of farmers), three of which are distributed in Guizhou Province. Whether there are differences among these areas in Guizhou Province is worth further exploration. Before using an ANOVA, the homogeneity of variance test (Appendix A) was carried out between the continuously poverty-stricken areas and the nonpoverty-stricken areas. The 2017 data failed to pass the inspection that the significance level needs to be greater than 0.05. The results of the ANOVA (Appendix B) indicated that there were significant differences between the three continuously poverty-stricken areas and the nonpoverty-stricken areas, with a significance level of less than 0.05 from 2003 to 2016. To further explore the difference between these areas, the least significant differences (LSD) method was used to make multiple comparisons of the PI average between the three continuously poverty-stricken areas and the nonpoverty-stricken areas. The results (Appendix C) based on LSD show that the three continuously poverty-stricken areas and the nonpoverty-stricken areas exhibited significant differences except for the Wuling and Wumeng areas. As shown in Figure 5, those two areas exhibited less difference mostly because their PI average was close, and their change trend was the same. These differences and similarities mean these areas should be treated separately when exploring the change and influential factors of rural poverty in the different continuously poverty-stricken areas.

3.2. Spatiotemporal Dynamics of the Influential Factors on Rural Poverty

Because each influential factor had different importance in the three continuously poverty-stricken areas in the different research periods, a stepwise regression analysis was used to select the influential factors before constructing the GTWR model. Considering the length of the paper, the coefficients of each influential factor are displayed in the form of a list (

Table 2. Maximum, minimum, and mean values of the coefficients of the influential factors in the three continuously poverty-stricken areas.

2003–2007	Wuling area	Influential factor	P7	P8	P16	P27	P30	P31
		Maximum	−0.32	0.83	−2.10	−0.36	−0.49	0.68
		Minimum	−0.40	0.65	−1.93	−0.58	−0.51	0.52
		Mean value	−0.37	0.73	−2.02	−0.49	−0.48	0.62
	Wumeng area	Influential factor	P9	P15	P16	P28
		Maximum	−0.87	3.98	−5.41	−2.76
		Minimum	−1.01	3.65	−5.60	−3.14
		Mean value	−0.92	3.77	−5.52	−2.92
	Rocky desertification of Dian-Gui-Qian area	Influential factor	P4	P8	P10	P15	P16	P22	P24	P30
		Maximum	10.52	−0.62	−1.10	0.87	−2.35	−0.55	−25.14	−0.33
		Minimum	8.88	−0.75	−1.48	0.73	−2.56	−0.82	−28.71	−0.48
		Mean value	9.50	−0.84	−1.25	0.80	−2.45	−0.70	−27.17	−0.39
2008–2010	Wuling area	Influential factor	P13	P16	P22
		Maximum	−9.16	−4.65	−0.98
		Minimum	−12.68	−5.00	−1.30
		Mean value	−10.39	−4.76	−1.13
	Wumeng area	Influential factor	P1	P9	P16	P34
		Maximum	1.24	0.67	−5.81	−2.10
		Minimum	1.04	0.41	−6.32	−2.44
		Mean value	1.11	0.51	−6.04	−2.29
	Rocky desertification of Dian-Gui-Qian area	Influential factor	P4	P10	P16	P29	P34
		Maximum	20.46	−1.13	−4.90	−0.33	−1.06
		Minimum	16.58	−1.32	−5.28	−0.59	−1.54
		Mean value	18.65	−1.20	−5.07	−0.45	−1.33
2011–2017	Wuling area	Influential factor	P10	P14	P16	P24	P26	P28
		Maximum	−7.26	4.77	−6.71	29.45	4.89	3.06
		Minimum	−8.38	3.02	−7.30	22.45	3.58	2.54
		Mean value	−7.57	3.55	−7.04	25.00	4.37	2.89
	Wumeng area	Influential factor	P5	P16	P27	P34
		Maximum	−121.78	−8.29	−2.19	−2.74
		Minimum	−128.53	−8.68	−2.46	−3.08
		Mean value	−125.18	−8.49	−2.35	−2.84
	Rocky desertification of Dian-Gui-Qian area	Influential factor	P1	P5	P10	P16	P26	P30	P34
		Maximum	−2.70	65.74	−5.75	−9.89	4.38	4.64	−1.03
		Minimum	−3.28	55.09	−6.77	−11.18	3.11	2.79	−1.92
		Mean value	−3.00	60.33	−6.26	−10.46	3.69	3.68	−1.51

), and the maps (Figure 6) of the local R² distribution of the three continuously poverty-stricken areas are given as an example in 2003, 2008, and 2011.

• 2003–2007;

Wuling area: As indicated by the data in Table 2, the influential factors of NDVI (P7), distance to the provincial capital city center (P8), per capita net income of peasants (P16), number of junior and senior high schools per ten thousand people (P27), number of industrial enterprises per ten thousand people (P30), and number of beds in medical and health institutions per ten thousand people (P31) had a high degree of interpretation, with an adjusted R² of 0.9427 for the GTWR model, which indicates that the main influential factors included the economy, location, natural endowment, and social development. Specifically, the coefficients of P30, P27, P16, and P7 showed significant negative values in all the studied counties, which indicates an abundantly positive effect on poverty alleviation in the Wuling area with the increase in the number of industrial enterprises and the net income of farmers. However, the increase in the number of junior and senior high schools and the NDVI could not effectively reduce rural poverty. In addition, the long distance to the provincial capital center and the improvement of medical facilities had a poor effect on rural poverty reduction in this period, as exhibited by the positive value of the coefficient on P8 and P31. The results of the local R² distribution demonstrated that the counties with a better explanation of rural poverty mainly included Yanhe Tujia autonomous area, Sinan, and Zheng’an County.

Wumeng area: The influential factors of distance to the city center (P9), per capita public expenditure (P15), per capita net income of peasants (P16), and the number of primary schools per ten thousand people (P28) had a 96.42% interpretation for rural poverty during this period. The coefficients of P15, P16, and P28 showed negative values, and P15 showed positive values, indicating that the increase in the net income of farmers, the improvement of primary school infrastructure, and the decrease in per capita public expenditure helped to greatly reduced poverty. However, better location conditions, such as being closer to the city center, were not conducive to rural poverty alleviation during this period. The spatial distribution of the local R² showed a better interpretation for Qixingguan District and Dafang, Nayong, and Qianxi counties.

Rocky desert area of Dian–Gui–Qian: The influential factors of the ratio of slope areas above 25° (P4), distance to the provincial capital city center (P8), per capita regional gross domestic product (GDP)(P10), per capita public expenditure (P15), per capita net income of peasants (P16), the ratio of students to teachers in primary school (P22), rural employees in secondary industry as a ratio to total rural employees (P24), and a number of industrial enterprises per ten thousand people (P30) had a relatively high degree of interpretation, with an adjusted R² of 0.7279. The coefficients of P24, P8, P16, P30, P22, and P10 showed negative values, and the coefficients of P4 and P15 were positive, indicating that the increases in rural employees in the secondary industry, net income of farmers, number of industrial enterprises and per capita regional GDP, and the decreases in the ratio of the slope area above 25° and per capita public expenditure had a great positive effect on rural poverty reduction. However, the decrease in the distance to the provincial capital city center and the increase in primary school teachers could not effectively reduce rural poverty during this period. The spatial distribution of the local R² showed that the areas distributed in southeastern Guizhou Province had a better interpretation of rural poverty.

• 2008–2010;

Wuling area: The influential factors of the ratio of the output value of the tertiary industry in GDP (P13), per capita net income of peasants (P16), and the ratio of students to teachers in primary school (P22) had an 86.55% interpretation for rural poverty during this period. The coefficients of P13, P16, and P22 showed negative values, which indicates that the increases in the net income of farmers and the ratio of the output value of the tertiary industry to the GDP had a significant effect on poverty reduction. However, the increase in primary school teachers could not effectively alleviate rural poverty during this period. The spatial distribution of the local R² showed a better interpretation in Wanshan District and Sinan and Zheng’an counties.

Wumeng area: The influential factors of elevation (P1), distance to the city center (P9), per capita net income of peasants (P16), and total power of agricultural machinery (P34) had a high degree of interpretation, with an adjusted R² of 0.7279. The increase in the net income of farmers and the improvement in agricultural machinery were conducive to rural poverty reduction, as exhibited in the negative values of P16 and P34. Better location conditions, such as being near the city center and having a low elevation, can effectively alleviate rural poverty, as exhibited through the positive values of P1 and P9. The results of the local R² distribution show that the counties with a better interpretation of rural poverty mainly included Weining Yi and Hui and Miao autonomous, Zhijin and Qianxi counties, and Qixingguan District.

The rocky desert area of Dian–Gui–Qian: The influential factors of the ratio of slope areas above 25° (P4), per capita regional gross domestic product (GDP) (P10), per capita net income of peasants (P16), number of fixed telephone users per ten thousand people (P29), and total power of agricultural machinery (P34) had a relatively high degree of interpretation, with an adjusted R² of 0.7525. The increase in the net income of farmers and regional GDP, the improvement in telecom technologies and agricultural machinery, and the decrease in the ratio of the slope area above 25° had significant effects on rural poverty reduction, demonstrated by the negative values of P10, P16, P29, and P34 and the positive value of P4. The spatial distribution of local R² showed that the areas located in southeastern Guizhou Province had a better interpretation of rural poverty, while the areas distributed in southwestern Guizhou Province had a relatively poor interpretation.

• 2011–2017.

Wuling area: The influential factors of per capita regional gross domestic product (GDP) (P10), per capita public revenue (P14), per capita net income of peasants (P16), rural employees in secondary industry as a ratio to total rural employees (P24), per capita investment in fixed assets (P26), and a number of primary schools per ten thousand people (P28) had an 88.29% interpretation for rural poverty during this period. The increase in the net income of farmers and regional GDP effectively reduced rural poverty, demonstrated by the negative values of P10 and P16. The coefficients of P14, P24, P26, and P28 showed significant negative values, indicating that the increases in public revenue, number of primary schools, rural employees in secondary industries, and investment in fixed assets were not conducive to alleviating rural poverty. The spatial distribution of the local R² showed a better interpretation of the Yuping Dong and Yanhe Tujia autonomous areas and Dejiang County.

Wumeng area: The influential factors of the ratio of slope areas above 30° (P5), per capita net income of peasants (P16), number of junior and senior high schools per ten thousand people (P27), and total power of agricultural machinery (P34) had a high degree of interpretation, with an adjusted R² of 0.9391. The coefficients of P5, P16, P27, and P34 were negative, which indicates that the increases in the net income of farmers and the number of junior and senior high schools, and the improvement of agricultural machinery had a significant effect on poverty alleviation. However, the decrease in the ratio of the slope area above 30° could not effectively alleviate rural poverty during this period. The results of the local R² distribution showed that the counties with a better interpretation of rural poverty mainly included Qixingguan District, Tongzi, Zhijin, and Qianxi County.

The rocky desert area of Dian–Gui–Qian: The influential factors of elevation (P1), the ratio of slope areas above 30° (P5), per capita regional gross domestic product (GDP) (P10), per capita net income of peasants (P16), per capita investment in fixed assets (P26), number of industrial enterprises per ten thousand people (P30), and total power of agricultural machinery (P34) explained 84.02% of the rural poverty during this period. The increases in the net income of farmers and regional GDP, the improvement of agricultural machinery, and the decreases in the ratio of the slope area above 30° and the investment in fixed assets had a significant effect on rural poverty reduction. However, the low elevation and the increase in industrial enterprises had a poor effect on rural poverty reduction during this period. The results of the local R² distribution demonstrate that the counties with a better interpretation of rural poverty were mainly located in southeastern Guizhou Province.

3.3. Analysis of the Variance Decomposition Results of the Influential Factors

The GTWR model was introduced into the variance decomposition. Through a comprehensive analysis of the influencing factors of the research periods in the three continuously poverty-stricken areas, two or three categories that represent relatively independent dimensions affecting poverty were selected for the variance decomposition, as shown in

Table 3. Statistics of the variance decomposition in the three continuously poverty-stricken areas.

Year	Area	Category	Variable	R2	Adjusted R2	Variance
2003–2007	Wuling area	Natural location (X1)	P7, P8	0.3891	0.3721	a = 0.0329 b = 0.4215 c = 0.1053 d = 0.2856 g = 0.0438 f = 0.0547 e = 0.0009 R = 0.0573
		Economy (X2)	P16	0.7552	0.7518
		Social development(X3)	P27, P30, P31	0.2369	0.2047
	Wumeng area	Natural location (X1)	P9	0.0883	0.0693	a = 0.1224 b = 0.8534 c = 0.1276 d = −0.1037 g = −0.0912 f = −0.0796 e = 0.1302 R = 0.0409
		Economy (X2)	P15, P16	0.7973	0.7887
		Social development (X3)	P28	0.1057	0.0870
	Rocky desertification of Dian-Gui-Qian area	Natural location (X1)	P4, P8	0.2462	0.2384	a = 0.1032 b = 0.3576 c = 0.0036 d = 0.0858 g = 0.1283 f = −0.0440 e = 0.0934 R = 0.2721
		Socio-economy (X2)	P10, P15, P16, P30	0.6720	0.6651
		Education/Labor (X3)	P22, P24	0.1897	0.1813
2008–2010	Wuling area	Economy (X1)	P13, P16	0.8189	0.8060	a = 0.8594 b = −0.0534 c = 0.0595 R = 0.1345
		Education/Labor (X2)	P22	0.0287	0.0061
	Wumeng area	Natural location (X1)	P1, P9	0.2947	0.2425	a = 0.0617 b = 0.6067 c = 0.1206 d = 0.2922 g = −0.0086 f = −0.0607 e = −0.0507 R = 0.0358
		Economy (X2)	P16	0.8480	0.8426
		Social development (X3)	P34	0.0339	0.0006
	Rocky desertification of Dian-Gui-Qian area	Natural location (X1)	P4	0.1964	0.1894	a = 0.0264 b = 0.3840 c = 0.0269 d = 0.0791 g = 0.1522 f = 0.0061 e = 0.0778 R = 0.2475
		Economy (X2)	P10, P16	0.6984	0.6931
		Social development (X3)	P29, P34	0.2757	0.2630
2011–2017	Wuling area	Economy(X1)	P10, P14, P16	0.8136	0.8081	a = 0.1613 b = 0.0719 c = 0.0115 d = 0.5566 g = −0.0086 f = 0.0052 e = 0.0850 R = 0.1171
		Social development (X2)	P26, P28	0.7106	0.7049
		Education/Labor (X3)	P24	0.1018	0.0931
	Wumeng area	Natural location (X1)	P5	0.2292	0.2179	a = 0.1004 b = 0.4543 c = 0.1195 d = 0.2352 g = 0.1474 f = −0.0557 e = −0.0620 R = 0.0609
		Economy (X2)	P16	0.7782	0.7749
		Social development (X3)	P27, P34	0.1739	0.1492
	Rocky desertification of Dian-Gui-Qian area	Natural location (X1)	P1, P5	0.1619	0.1557	a = 0.0342 b = 0.2780 c = 0.0005 d = −0.0181 g = 0.4060 f = 0.0326 e = 0.1070 R = 0.1598
		Economy (X2)	P10, P16	0.7724	0.7709
		Social development (X3)	P26, P30, P34	0.5511	0.5461

. In general, the findings show that the category economy had a major influence on rural poverty in the three continuously poverty-stricken areas for the three research periods, with an explanation degree above 65%, implying that regional economic development played an important role in alleviating county rural poverty. Natural conditions, social development, and education/labor had various effects on the different counties in the different research periods. The natural location partly explained rural poverty in the Wumeng area and the rocky desert area of Dian–Gui–Qian. In addition, the interpretation degree of the natural location in the rocky desert area of Dian–Gui–Qian decreased over time, which indicates that the influence of natural factors on this area was decreasing. With regard to the Wuling area, the influence of the natural location appeared only from 2003 to 2007, with an explanation degree of 37%. The impact of education/labor has begun to contribute to rural poverty alleviation, and the influence degree was increasing.

The findings from the rocky desert area of Dian–Gui–Qian suggest that social development had an increasing impact on rural poverty, with an interpretation degree of approximately 26% in 2008–2010 to 55% in 2011–2017. The impact degree and effect of the various influential factors on the three continuously poverty-stricken areas yielded considerably different results over time, indicating that poverty alleviation policies need to be considered according to different regional characteristics.

4. Discussion

Over the last 15 years, three continuously poverty-stricken areas in Guizhou Province have progressed substantially in terms of poverty reduction and economic development. Our results indicate decreased poverty rates and spatial discrepancy at the county level between 2003 and 2017. The results indicate that China has made remarkable achievements in its commitment to poverty reduction and has built a strong foundation for overcoming poverty over the last decades [4,17,46]. However, our findings also confirm the imbalance of the county poverty reduction in the three continuously poverty-stricken areas, and a more significant discrepancy in poverty reduction is embodied by the spatiotemporal evolution of rural poverty. In particular, the spatial discrepancy in the rate of average annual poverty reduction of the three continuously poverty-stricken areas was obvious in the research period 2011–2017. The poverty-stricken counties with high decreasing rates were mainly distributed in the boundary counties of the continuously poverty-stricken areas, indicating that the developed counties had a significant radiation and driving effect on the poverty-stricken counties. The poverty-stricken counties with relatively low decreasing rates were mostly located in southeastern Guizhou Province, which is significantly affected by its terrain and fragile ecological environment. The natural location of the rocky desert area of Dian–Gui–Qian in the three research periods may be the reason for these results. It is thus clear that poverty reduction is a complicated systematic project. A comprehensive analysis of the spatiotemporal changes in poverty reduction is of great significance for exploring the essence of regional rural poverty and its influential factors and consolidating the achievements of poverty alleviation.

The characteristics specific to the three continuously poverty-stricken areas in Guizhou Province exhibited some similarities and differences in the influential factors and rural poverty. As elaborated in Section 3.2 and Section 3.3, the influential factors and their effect degrees in the continuously poverty-stricken areas were generally different over time. These differences revealed further details on the regional environmental characteristics and on the development and changes in the poverty alleviation policies when compared with the cross-sectional data analysis or some category of influential factors on which many studies focus. Our results also indicate that the spatiotemporal evolution of rural poverty in the three continuously poverty-stricken areas resulted from the combination of leading influential factors and various influencing factors with regional characteristics. As shown in Figure 7, the regional characteristics of three continuously poverty-stricken areas in Guizhou Province, summarized from Guizhou Province Chronicles Poverty Alleviation and Development, are largely consistent with the influential factors identified by the method used in this research. This result indicates the effectiveness of the research framework and method in this paper. For all the areas, the regional economy is the most critical influential factor affecting rural poverty, which means that the development of the regional economy guarantees sustainable poverty reduction. The effect of natural location on rural poverty is decreasing in the research area; however, its impact should not be ignored. Further analysis shows that the decrease in the impact of natural location factors is closely related to the policy of poverty alleviation and relocation. Social development, which represents residents’ comfort with social services and education/labor embodied in the reserve force for county development, has gradually strengthened its impact on rural poverty reduction. In the context of the regional environment, developing a regional characteristic economy, perfecting social security facilities, and formulating strategies for the coordinated development of the environment and economy can further consolidate the achievements of regional poverty reduction. For example, since the implementation of the targeted measures in poverty alleviation, the policies such as “one policy for one county”, “one industry for one county”, and “roads for all villages” have played a very important role in supporting the comprehensive poverty alleviation by 2020. Moreover, the dynamic mechanisms of natural location and rural poverty and the spatiotemporal difference in poverty reduction between the poverty-stricken counties under the background of targeted poverty alleviation need to be further explored in subsequent studies to provide guidance and support for sustainable development.

5. Conclusions

China has reduced absolute poverty and is contemplating a mission shift on consolidating the achievements of poverty alleviation, preventing regression back to poverty, and targeting relative poverty. In this context, it is important to assess the poverty reduction patterns that have evolved and the influential factors that have helped China reach this point. The main contribution of this paper, based on the spatiotemporal evolution of rural poverty in three continuously poverty-stricken areas in Guizhou Province, is a comprehensive understanding of the difference between the areas and the influential factors analyzed that impact rural poverty reduction in the different research periods.

We conclude that (1) most of the poverty-stricken counties showed a significant downward trend in rural poverty in the periods 2003–2007 and 2011–2017, indicating that those counties had obvious poverty reduction over time; (2) the three continuously poverty-stricken areas embodied different characteristics in the different research periods, with degrees of average annual poverty reduction of all research counties between 2.98–10.09%, 10.60–26.96%, and 11.46–43.19% in 2003–2007, 2008–2010, and 2011–2017, respectively, and the spatial heterogeneity of the decreasing poverty in the poverty-stricken counties was obvious from 2011 to 2017; (3) significant differences were observed in the PI average between the three continuously poverty- and nonpoverty-stricken areas except the Wuling and Wumeng areas; and (4) the influencing factors affecting the spatiotemporal evolution of rural poverty in the three continuously poverty-stricken areas in Guizhou Province at the different periods were different, and the degree and effect of each factor had significant spatial differences, which is consistent with the leading influential factors and various influencing factors with regional characteristics.

The spatiotemporal evolution of rural poverty and its influential factors in different continuously poverty-stricken areas can deepen our understanding of the mechanism of poverty and help to uncover further details about the change in spatial poverty. Major policy implications derived from our findings include the following actions: (1) steadily increase farmers’ income by developing characteristic industries and reshaping the human–land development pattern to improve the regional economy; (2) perfect social infrastructure—such as education, medical treatment, and welfare—which are the important influential factors for preventing the intergenerational transmission of poverty; and (3) strengthen the support of poverty alleviation policies in the counties in the interior of the continuously poverty-stricken areas, especially in eastern Guizhou Province. With the planned consolidation of the achievements of poverty alleviation and in view of the 2030 Sustainable Development Goals, sustainable poverty reduction in poverty-stricken counties remains pressing, and the results of this research indicate that the government should continue responding in the areas of education promotion, social welfare improvement. and sustainable development of the regional economy.