Numerous studies have examined regional economic growth patterns and economic disparities, with the results often revealing clear differences between regions. At the national level, studies based on provincial or city data have generally discovered absolute convergence between provinces and cities in China, that is, the economic growth of developing areas has been faster than that of developed areas (Qi et al., 2013; Yin et al., 2016). However, according to county-level data, the gap in economic development among Chinese counties is expanding. Counties in areas such as the southeast coast and Beijing-Tianjin-Hebei agglomeration have generally exhibited rapid economic growth, whereas counties in the central, south-central, and southwestern parts of China have generally exhibited slower economic growth (Zhou et al., 2014). At the regional level, the economic disparities are declining at the county level in economically developed regions, such as the Yangtze River Delta and Fujian Province (Fang et al., 2017; Zhao et al., 2020), whereas less-developed areas, such as Yunnan Province, are clearly exhibiting a trend for polarization in county economic growth, with this trend presenting as a typical core-periphery pattern with weak spatial fluctuation (Chen et al., 2017). Liu and Xue (2021) also discovered in their study on the Loess Plateau that the absolute difference, fluctuation of the relative difference, and total difference in economic development in this region were all steadily increasing.

As poverty has increasingly been alleviated, scholars have focused more on the pattern of economic growth in destitute areas. The analysis conducted by Pan and Jia (2014) into national-level poor counties in the years 2000-2010 revealed an imbalance in the overall economic development level of the poor counties; moreover, the economic gap between counties expanded, with already strong counties exhibiting stronger growth and weak counties exhibiting weaker growth. However, considerable convergence-club characteristics were discovered in poor counties. In his research on the Wuling Mountain area, Ding (2014a) discovered that the economic gap had first increased and then decreased. Geng et al. (2016) explored the economic development level of counties in Tibetan areas of Sichuan Province and concluded that the economy of the region was characterized by strong spatial centralization and clear solidification of poverty.

Scholars have generally paid more attention to economic development on the national, urban regional, or provincial levels, with research on economic growth in destitute areas, especially at the national level, being insufficient; moreover, the studies that have been conducted have revealed inconsistent findings. The economic growth trends in 14 contiguous poverty-stricken areas considered the main battlefields of poverty alleviation and the spatial performance in these areas have not yet been systematically analyzed.

The determinants of economic growth have been the focus of numerous studies. Scholars have conducted considerable research on place-based factors such as capital investment, industrial structure, and government’s fiscal expenditure. The Solow growth model decomposes economic growth into productivity growth related to capital, labor, and all factors. High investment accelerates capital accumulation through capital expansion by increasing the total factor productivity, thus promoting economic growth (Romer, 1986). Empirical studies have confirmed the role of capital investment in promoting economic growth (EGFST, 2005; Wang et al., 2015a). However, several studies have reported that the promotional role of investment diminishes gradually over time (EGFST, 2005). An unbalanced investment structure may even stunt economic growth (Qiu et al., 2020). Financial capital plays a critical role in economic growth. Adopting a theoretical perspective, McKinnon (1973) and Shaw (1973) suggested that financial development can promote economic growth through “finance repression” and “financial deepening.” This theory was confirmed for the context of China (Zhao and Lei, 2010). Regarding industrial structure, factor inputs theoretically flow from low-productivity sectors to high-productivity sectors; thus, transformation of an industrial structure is the main impetus for economic growth (Gan et al., 2011). Empirical studies have generally confirmed the positive influence of the secondary sector, especially technology-intensive industries, on economic growth (Sun, 2006; Zhao and Tang, 2015). However, the influence of the tertiary sector on economic growth is relatively weak, and a tendency toward excessive servitization leads to a structural economic slowdown (Zhou et al., 2020a). In terms of government interventions, the new growth theory, which introduces government public expenditure into the Arrow-Kurz model, was proposed, and theoretical derivation indicated that an increase in public expenditure would significantly promote economic growth (Barro, 1990). According to public finance theory, government expenditure is mainly financed through taxes; however, the advantages of tax financing are disputed, with increases in tax rates inhibiting economic growth (Yan and Gong, 2009). Just as theoretical disagreements have occurred, empirical analyses of the relationship between government intervention and economic growth have resulted in divergent conclusions (Barro and Redlick, 2011; Wang et al., 2015a).

Researchers have increasingly realized in recent years that geo-structural factors such as geographic environment, market location, and accessibility, not only place-based factors, play crucial roles in economic growth (Shearmur and Polèse, 2007). A suitable local environment is conducive to attracting enterprises and labor and is considered a key factor in promoting economic growth (Spolaore and Wacziarg, 2013). Felkner and Townsend (2011) confirmed that geographic variables such as elevation, soil fertility, and rainfall variation have significant influences on enterprise agglomeration in Thailand. Posada et al. (2018) also reported in their study on Spain that precipitation is a critical factor affecting regional employment growth. Proximity to markets also plays a major role in economic growth. Rappaport and Sachs (2003) discovered that proximity to a coastline was a key factor in explaining local productivity and economic growth in the United States. However, improving the accessibility of transport does not necessarily promote economic growth. Because of the backwash effect, improvements in accessibility may also lead to loss of regional economic factors due to people’s attraction to large cities, which is not conducive to economic growth outside the cities (Condeço-Melhorado et al., 2014). Therefore, the relationship between market entry and economic growth remains unclear.

The aforementioned literature on the drivers of economic growth mainly focuses on the national, regional, or urban scale; research targeting destitute areas has been relatively limited. Few studies have investigated policies related to poor counties, government fiscal expenditure, and transportation locations. Using panel data of 993 counties in China from 2005 to 2015, Huang (2018) examined the impact of the establishment of national-level poor counties on local economic development. The results revealed that the establishment of such counties had had a significant and continuous promotional effect on local economic development and that the longer time had passed since establishment of national-level poor counties, the stronger was the promotive effect. In a study into the underdeveloped Appalachia region in the United States in the 1980s and 1990s, Mencken and Tolbert (2005) discovered that federal investment in infrastructure increased the productivity of private capital and labor, thereby promoting regional economic growth. Transportation location is a critical factor affecting economic growth in destitute areas, but the direction of its effect remains unclear. On the basis of data for 14 contiguous destitute areas in China, Wang et al. (2015b) observed that the degree of road transport superiority had a significant promotional effect on economic growth, with each 1% increase in superiority resulting in 0.193% higher economic output. Daniels and Minot (2021) reached similar conclusions in a study of rural Uganda. However, Chen and Sun (2017) discovered in their research on the Beijing-Tianjin-Hebei region of China that the siphon effect of the big cities led to slower economic growth in areas closer to Beijing, Shijiazhuang, and other cities than other cities in this region.

Notably, studies on the drivers of economic growth have mostly employed a global perspective and have not paid sufficient attention to the spatial heterogeneity in influencing factors. In fact, a few researchers have found spatial differentiation in the effects of various factors. In terms of financial development, for example, Zhang and Qin (2009) considered provincial data from the eastern and western parts of China and reported that financial development made a significant contribution to economic growth; however, in central and northeastern parts of the country, financial development did not promote economic growth. In terms of market location, Wang and Gao (2014) also found in their study on Inner Mongolia that because of the proximity of central Inner Mongolia to core cities and the location conditions of the Beijing-Tianjin-Hebei region, the economy of central Inner Mongolia has developed rapidly since the reform and opening-up and that this region has replaced eastern Mongolia as an economic hot spot. However, the aforementioned studies investigated large units, such as regions and provinces, and lacked detailed investigation of smaller spatial units, such as counties. Such investigation was the main purpose of the present study.

The study area comprised the 14 contiguous destitute areas identified in the Outline for Poverty Alleviation and Development in Rural China (2011-2020), namely the Liupan Mountain area, Qinba Mountain area, Wuling Mountain area, Wumeng Mountain area, Dian-Gui-Qian Rocky Desert area, Western Yunnan Border Mountain area, Southern Da Hinggan area, Yanshan-Taihang Mountain area, Luliang Mountain area, Dabie Mountain area, Luoxiao Mountain area, Tibet, the Tibetan areas of four provinces, and three prefectures within southern Xinjiang. These 14 areas are located across 22 provincial-level regions and 680 counties and account for 73.1% of poor counties in China. In addition, the 14 areas cover 4,046,200 km², which represents 42.52% of China’s total land area.

As illustrated in Figure 1, the study area is concentrated in central and western China. Topographically, the area is dominated by mountains and hills, the surface shape is fragmented, water and soil resource constraints are prominent, and the overall capacity of resources and the environment is relatively low. The 14 areas are also commonly the sites of geological disasters and are vulnerable to soil erosion and rocky desertification. They are not only vulnerable areas in terms of their ecosystem but also crucial barriers for national ecological security (Guo and Jiang, 1995; Zhou et al., 2020b). As the main battlefields of China’s period of battling poverty, these areas have an immense area, a large population, and relatively lagging economic development. These used to be the regions with the largest concentration of poor people in China. In 2011, when the central government designated the 14 areas as the main battlefields in the fight against poverty, the per capita GDP, local fiscal revenue, and net income of farmers in the contiguous destitute areas were far below the national and western China average levels - only 49%, 44%, and 73% of the western China averages, respectively. The population in poverty in the contiguous destitute areas accounted for more than 70% of the national population in poverty, and the average poverty incidence rate was 28.4%, which was 15.7% higher than the national average of 12.7% (Ding, 2014b). The 14 areas are thus not only critical in terms of geography but also key in social and economic significance. An in-depth analysis of the spatio-temporal characteristics of and factors influencing their economic growth was therefore conducted.

Data from 2011 to 2018 were retrieved from the China County Statistical Yearbook and China Statistical Yearbook. Because the China County Statistical Yearbooks for 2011 and 2012 do not report county GDP data, we supplemented the data with data from the China Provincial Statistical Yearbook. Vector data on county administrative divisions were obtained from the National Basic Geographic Information Database of the National Bureau of Surveying and Mapping. Data on geographic environment such as precipitation were obtained from the Resource and Environment Science Data Center of the Chinese Academy of Sciences (http://www.resdc.cn/Default.aspx). Ideally, transportation location data beginning in 2011 would have been used; however, because of the difficulty of obtaining these data, the 2012 data were used. In addition, road network data from the 2013 China Road Mileage Map Subseries (China Map Press) were obtained by vectorizing the road traffic and county area data. The corresponding transportation distance is the shortest possible distance between corresponding locations in each county when using various roads. Of the 680 counties, only 669 were included because data on socioeconomic variables for certain county-level administrative units were unavailable.

In the GWR model, county economic growth (average annual growth rate of per capita GDP from 2011 to 2018) was taken as the dependent variable. Regarding the selection of independent variables, we incorporated place-based factors and geo-structural factors into the model in accordance with the approach in a related study (Shearmur and Polèse, 2007), and the economic characteristics of underdeveloped areas were fully considered. Specifically, the place-based factors were as follows: initial economic level, industrial structure, capital investment, financial development, and government intervention. In addition to considering market location, we considered geo-structural factors in the choice of geographical conditions. Because the formerly destitute areas in China are mostly located in mountainous and hilly regions, the areas have certain similarities in terms of topographical conditions but clearly heterogeneous precipitation conditions. Given that precipitation is considered a factor with a crucial effect on economic growth (Spolaore and Wacziarg, 2013), the average annual precipitation was employed to reflect an area’s geographical environment. Moreover, because of the low level of economic development in formerly destitute areas and agriculture remaining a major sector of the regional economy, we added the level of agricultural mechanization to the GWR model. To examine potential multicollinearity issue, we calculate the variance inflation factor (VIF) for each variable. The corresponding results are presented in Table 1. The VIF of each variable was considerably lower than 3, which indicated that no multicollinearity problem existed.

China’s formerly destitute areas have relatively small economies, and their economic development levels are generally low. In 2018, the GDP of the 14 contiguous destitute areas was RMB 5.37 trillion, representing only 5.97% of the Chinese GDP. This ratio is substantially lower than the household population share of these 14 areas (17.62%). The average GDP of counties in the formerly destitute areas was RMB 8.031 billion, which was only 25% of the national average for counties (RMB 31.579 billion). The GDP per capita in the formerly destitute areas was RMB 25,679, which was only 39.8% of the national average (RMB 64,521). A large economic gap thus exists between formerly destitute areas and other areas in China, suggesting that relative poverty alleviation is a long-term task for the government.

At the county level, great disparities were discovered in the overall patterns of economic size and development level in the formerly destitute areas (Figure 2). In terms of economic size, counties with the largest economic sizes in 2018 were mainly located in the Dabie, Qinba, Wuling, and Wumeng Mountain areas. The GDP of these counties were mostly higher than RMB 20 billion. The total economic output of counties in Tibet, the Tibetan areas of four provinces (except for Golmud City), and three prefectures of southern Xinjiang was generally smaller than RMB 5 billion. In terms of economic size per capita, with the exception of a few counties in Tibet and Tibetan areas of four provinces, the per capita GDP of the counties in 2018 was less than RMB 40,000; such low GDP per capita was thus relatively widely distributed and not attributable to some areas having extremely high or low GDP per capita. The data indicated that 651 counties had a per capita GDP lower than the national average (RMB 64,521), and these counties accounted for 97.3% of the contiguous destitute areas in China. The per capita GDP level of counties in formerly destitute areas was generally low. The intercounty gap in per capita GDP was smaller than was the intercounty gap in economic size.

Since the central Chinese government identified 14 contiguous destitute areas as the main focus in the battle against poverty, the economies of these areas have generally grown rapidly. From 2011 to 2018, the per capita GDP in formerly destitute areas increased by RMB 12,614, with the average annual growth rate being 10.54%, higher than the national average of 9.14%. The Chinese government’s poverty alleviation policy has thus had remarkable effects. In 2011, the per capita GDP in formerly destitute areas was 36.1% of the national average, but in 2018, it reached 39.8% (Figure 3). Thus, the economic gap between the formerly destitute areas and the country has a whole has narrowed.

The economic gaps between counties in formerly destitute areas first narrowed and then slowly increased. By using the coefficient of variation and Theil index of per capita GDP at the county level, we noted a clear decline in the difference in per capita GDP between counties from 2011 to 2016; however, the difference increased after 2016. Overall, however, the economic disparity at the county level in formerly destitute areas was considerably narrower in 2018 than in 2011; thus, a convergence trend was observed in the county-level economy.

Regarding the spatial growth pattern, clear spatial differentiation was observed in the economic growth of each county illustrated in Figure 4. The counties with the strongest per capita GDP growth were mainly located in central and southern Tibet, the eastern Tibetan areas of four provinces, a few areas such as Golmud and Delingha in Qinghai, the Qinba Mountain area, and peripheral counties in the Wuling Mountain area; the per capita GDP of these counties grew by more than RMB 10,000 over the study period. By contrast, the counties located in the central and northern Tibetan areas of four provinces, Liupan Mountain area, southern Da Hinggan area, three prefectures of southern Xinjiang, and Yanshan-Taihang Mountain area had smaller positive economic increments. The counties in Tibet, the northern part of the Dian-Gui-Qian Rocky Desert area, and the western part of the Wuling Mountain area generally experienced rapid economic growth. By contrast, the economic growth rate of counties in the Tibetan areas of four provinces, the Liupan Mountain area, the Yanshan-Taihang Mountain area, the southern Da Hinggan area, and the Dabie Mountain area was relatively slow, and the economy of some counties in the Tibetan areas of four provinces even shrank. Overall, western and southern counties grew faster economically than central and northern counties did.

We further analyzed the spatial autocorrelation model of per capita GDP and growth rate in formerly destitute areas at the county level from 2011 to 2018. The global Moran’s I for these years was 0.261, 0.196, 0.220, 0.249, 0.233, 0.268, 0.265, and 0.275, respectively, and the p value passed the 1% significance threshold. This result demonstrated that the spatial distribution of economic development in counties within formerly destitute areas exhibited a significant positive correlation, that is, counties with higher economic levels were clustered, as were those with relatively underdeveloped economies. The global Moran’s I exhibited an overall upward trend over time, indicating that the degree of agglomeration of county economies increased continually. The global Moran’s I of the GDP growth rate per capita from 2011 to 2018 was 0.313, and the 1% significance test was passed, indicating that county economic growth also had a significant positive spatial correlation. The faster-growing counties were spatially clustered, as were the slower-growing counties. The economic growth of neighboring counties was characterized by mutual influence and interdependence.

The Getis-Ord G_i^* was used to analyze the local autocorrelation characteristics of the per capita GDP and its growth in counties, and a distribution map of cold and hot spots was obtained. Overall, the pattern of cold and hot spots in terms of county economies changed considerably from 2011 to 2018. In 2011, the hot spots were mainly located in western and northern Qinghai Province and Lhasa in Tibet. These areas were either provincial capitals or resource-based areas with large economies, a low population, and high per capita GDP. The cold spots were mainly located in the Wumeng Mountain, western Wuling Mountain, northern Dian-Gui-Qian Rocky Desert, and western Liupan Mountain areas (Figure 5a). In 2018, both economical hot and cold spots increased greatly in size. Except for the western part of Qinghai, the hot spots in southern Tibet increased markedly. In addition, the eastern and northern parts of the Qinba Mountains became new hot spots. The cold spot areas extended from the western Liupan Mountain area to the eastern Tibetan areas of four provinces. The cold spot areas in the Wuling Mountain and Dian-Gui-Qian Rocky Desert areas decreased considerably, and the cold spot areas in the Wumeng Mountain area increased. Furthermore, some counties in three prefectures of Xinjiang became new economic cold spots (Figure 5b).

The economic growth of the counties in formerly destitute areas exhibited clear spatial agglomeration. The hot spots of economic growth were mainly located in central and southern Tibet (Naqu, Lhasa, and Shannan), the northern Dian-Gui-Qian Rocky Desert area (Guizhou), the western Wuling Mountain area (Chongqing and Guizhou), the Western Yunnan Border Mountain area, and the central Qinba Mountain area (Shaanxi). Spatial agglomeration was observed for these counties, and they were found to have had faster economic growth than other destitute regions. The cold spots of economic growth were mainly located in the central and northern parts of the Tibetan areas of four provinces (Qinghai), the western Liupan Mountain area, the Yanshan-Taihang Mountain area, and the southern Da Hinggan area. The economic growth in these areas was slow and clearly spatially dependent.

The analysis presented in section 4 revealed that the economic growth of counties in formerly destitute areas of China had significant spatial autocorrelation and heterogeneity. The ordinary least squares (OLS) method disregards geographical variation in the relationships between dependent and independent variables. By contrast, the GWR model can effectively solve local variation problems on the basis of location. To verify the superiority of the GWR model, regression analysis of the data was conducted using the OLS and GWR models. The projection coordinates of the county seat of each county were selected as the geographical coordinates. The fixed Gaussian function was used as the weight function, and the bandwidth was determined using the Akaike information criterion (AIC). The results of the OLS versus GWR model comparison are presented in Table 2. The R² obtained using the GWR model was 0.6589, considerably higher than that obtained using the OLS model (0.2996). The AIC value obtained using the GWR model was -1609.5924, significantly smaller than that obtained using the OLS model (-1490.1014). The residual sum of squares (RSS) obtained using the GWR model was also significantly lower than that obtained using the OLS model. Overall, the results indicated that compared with the OLS model, the GWR model had superior fit when describing economic growth in formerly destitute areas.

Figure 6 shows the spatial distribution of the goodness of fit between the indicators selected using the GWR model and the actual county economic growth based on the processing results obtained using the GWR model. The R² for most counties was in the range of 0.517 to 0.856, indicating that the nine indicators selected in this study had strong ability to explain economic growth in formerly destitute areas. R² was highest for the areas in southwestern China (including the Dian-Gui-Qian Rocky Desert area, Wuling Mountain area, Wumeng Mountain area, and Western Yunnan Border Mountain area), the Qinba Mountain area, and the northern Tibetan areas of four provinces, and R² was lowest for counties in western Tibet, the eastern Tibetan areas of four provinces, and the Liupan Mountain area, corresponding to weak explanatory power of the indicators. This low R² may be attributed to these areas’ high altitude and harsh environments, which hamper life in general and economic production (Tian et al., 2018). The geographical environment indicators selected in this study failed to fully reflect the geographical conditions, resulting in a low degree of fit.

Each indicator in the GWR model yielded a specific regression coefficient regarding the economic growth of each county, and the results obtained for these indicators are presented in Table 3. Figure 7 displays the distribution map of regression coefficients obtained using the GWR model. The positive and negative results for the regression coefficients revealed that each indicator had both positive and negative effects on economic growth, which suggested that the influential factors had clear spatial instability and heterogeneity. The effects of tertiary industry growth and financial development level on economic growth were mainly positive. By contrast, the effects of initial economic level and market location were mainly negative. The proportions of positive and negative values for the other indicators were approximately equal. Moreover, the indicators exhibited significant spatial differences. According to the mean absolute value of the regression coefficients, growth of secondary industry had the strongest effect on economic growth, followed by initial economic level, whereas the impact of other variables was relatively weak.

Initial economic level had a negative and positive effect on economic growth in 77.88% and 22.12% of the formerly destitute counties, respectively. A negative effect indicates that counties with lower initial economic levels grew more rapidly. A trend of economic convergence among counties was discovered, which is consistent with the trend in urban economic growth at the national level but contrary to the trend in county economic growth (Zhou et al., 2014; Li et al., 2016). This result may be related to the time and scope of these studies. In the formerly destitute areas, poverty alleviation promoted economic growth and narrowed economic disparities. For example, regarding the Wuling Mountain area, a study revealed that the intraregional economic gap first increased and then decreased between 2000 and 2011 (Ding, 2014a). The positive effect of initial economic development level indicated that counties with a poorer economic base grew more slowly. These counties were located in the southern part of Tibet and in the border areas of Sichuan, Tibet, and Yunnan (Figure 7a), which is consistent with findings of previous studies for Yunnan and the Tibetan areas of Sichuan Province (Geng et al., 2016; Chen et al., 2017). These areas are a major part of China’s “three regions and three states” that have extreme poverty, and the prospect of managing relative poverty in these areas remains daunting.

Industrial structure factors had various effects on economic growth. The positive effect of secondary industry growth was predominant, and the positive and negative effects of tertiary industry growth were approximately equivalent. In total, 92.08% of the counties were positively affected by secondary industry growth; thus, the greater the increase in the proportion of the secondary industry was, the faster was the economic growth of the county. This result indicated that an increase in industrialization level strongly drove the economic growth of formerly destitute areas, which is consistent with the findings of studies that have investigated economic growth at the national or provincial level in China (Zhao and Tang, 2015; Chen et al., 2017). In terms of the spatial distribution of the regression coefficients, the counties for which secondary industry growth had strong impacts were mainly located in the Tibetan areas of Sichuan Province, the Qinba Mountain area, and the Luliang Mountain area (Figure 7b). Tertiary industry growth had a positive effect on 54.56% of the counties and a negative effect on 45.44% of the counties. A positive effect indicated that the greater the proportion of tertiary industry was, the faster the economy grew. The counties with high impact values were mainly located in the Tibetan areas of Sichuan Province, the Western Yunnan Border Mountain area, the Yanshan-Taihang Mountain area, and the Luliang Mountain area (Figure 7c). We surmised that this result was related to the relatively developed local tourism industry in these areas. In addition, the findings of another study on the Tibetan areas of Sichuan Province are consistent with our findings (Tian and Li, 2018); that study reported that counties in the Tibetan areas of Sichuan Province with a higher proportion of tertiary industry had a more developed tourism-related economy. Furthermore, the present study discovered that the economic development level of the service sector was the main factor contributing to economic differences in the area. The areas for which the negative impact was strong were mainly located in Tibet, the western and eastern Tibetan areas of four provinces, the Dian-Gui-Qian Rocky Desert area, the eastern Wuling Mountain area, and the Dabie Mountain area. The increase in the proportion of tertiary industry in these areas was not conducive to economic growth possibly because the tertiary sector in these areas is dominated by traditional life service industries, which have low labor productivity. The scale of the economy for these services is relatively small (Fang et al., 2012), and expansion of the service sector was detrimental to the overall level of regional economic development.

Agriculture mechanization had a positive effect on economic growth in most counties. The counties for which agriculture mechanization had the strongest positive impact were mainly located in the Western Yunnan Border Mountain area, Tibetan areas of four provinces, Wuling Mountain area, Luoxiao Mountain area, and eastern Liupan Mountain area (Figure 7d). Agriculture mechanization contributed considerably to economic growth in these areas (Xu and Zhang, 2016). With the advancement of industrialization and urbanization, the cost of agricultural labor is continually rising. The shift to modernization, in which agricultural mechanization is used as a technical means, is crucial to minimizing labor costs and increasing efficiency and labor productivity and has become an important factor driving regional economic growth (Wei and Li, 2012). The counties for which the impact of agriculture mechanization was strongly negative were mainly located in three prefectures of southern Xinjiang and western Tibet. An increase in agriculture mechanization was detrimental to the economic growth of these regions. Specifically, local agriculture mechanization was characterized by insufficient agricultural equipment in Xinjiang, several problems were identified, including the use of small and inefficient machinery, repeated investment in poor machinery, and low utilization rates (Wei and Li, 2012). The excessive growth of investment in agricultural machinery has resulted in less investment in other factor inputs, which has not been conducive to overall economic improvement. Consequently, the effects of increasing agriculture mechanization in formerly destitute areas cannot be generalized. The local context of each region should be considered, and various development responses should be implemented.

The positive and negative effects of capital investment on economic growth in formerly destitute areas were approximately equivalent. The counties for which fixed asset investment had strong positive effects were mainly located in the western Liupan Mountain area, northern Tibetan areas of four provinces, southern Qinba Mountain area, and three prefectures of southern Xinjiang (49.18%). The faster capital investment grew, the faster the economy grew, which indicates that economic growth in the aforementioned areas was clearly driven by investment factors. The counties for which capital investment had strong negative effects were mainly located in the southern Da Hinggan area, Yanshan-Taihang Mountain area, Wuling Mountain area, and Luliang Mountain area (Figure 7e). In the aforementioned areas, the capital stock exceeded the level required for local economic development, and promoting economic growth through fixed asset investment was no longer effective (Zhang and Zhang, 2020). An ineffective investment structure can cause a decline in investment efficiency and inhibit economic growth. Manufacturing investment and real estate investment have significant effects on economic growth, whereas infrastructure investment does not (Qiu et al., 2020). By contrast, a study on the three provinces of northeast China revealed that fixed asset investment significantly inhibited the growth of county economies (Zhang and Zhang, 2020).

Financial development contributed to economic growth in 69.96% of the counties investigated. The counties strongly positively affected by financial development were mainly located in the Dian-Gui-Qian Rocky Desert area, Wumeng Mountain area, and Qinba Mountain area (Figure 7f). The growth of loans issued by financial institutions resulted in crucial financial support being provided to economic entities and the efficiency of resource allocation being improved in these areas; thus, regional economic growth was promoted in the aforementioned areas (Zhao and Lei, 2010). Economic growth in 30.04% of the counties was negatively influenced by financial development; that is, the more rapidly the number of loans issued by financial institutions increased, the slower the economy grew. These areas were mainly located in the Dabie Mountain area, three prefectures of southern Xinjiang, and western Tibet. Wang and Zhu (2018) reported that the efficiency of financial support was low in the aforementioned areas; thus, an increase in the number of loans had a limited effect on the economic development of poor counties in these areas. In addition, Ding (2012) observed that in three prefectures of southern Xinjiang, credit funds flowed mainly to other developed areas with low local utilization rates. Therefore, excessive financial loan growth resulted in a lower extent of local capital usage, which was detrimental to economic development.

Government expenditure had a positive effect on economic growth in most counties. The faster government’s fiscal expenditure grew, the faster the economy grew. This effect was particularly pronounced in the Liupan Mountain area, western Qinba Mountain area, three prefectures of southern Xinjiang, and western Dian-Gui-Qian Rocky Desert area (Figure 7g). Government expenditure on infrastructure and public service provisions, tax concessions, and financial subsidies had positive effects on economic growth in the aforementioned regions (Markusen et al., 1996). However, government intervention had a negative effect on economic growth in 41.7% of the counties. These counties were mainly located in the southern Da Hinggan area, Yanshan-Taihang Mountain area, western Wumeng Mountain area, and southern Wuling Mountain area. As government expenditure increased, the rate of economic growth decreased in the aforementioned areas, which indicated that excessive government intervention in microeconomic activities and administrative allocation of resources disrupted market operation in these areas possibly because of low economic efficiency and stunted economic growth. A study on northeast China corroborated this conclusion; Mu et al. (2018) observed that for every 1% increase in government spending as a proportion of the GDP in northeast China between 1991 and 2016, the macroinvestment efficiency declined by 2.73%. Therefore, regulating government investment activities is essential for improving market operations and facilitating the guiding role of the market.

Precipitation had a significant positive effect on the economic growth of counties in the formerly destitute areas. Specifically, the counties’ economies grew faster when they had more annual precipitation. The positive influence of precipitation was strongest in the three prefectures of southern Xinjiang, southern Da Hinggan area, Yanshan-Taihang Mountain area, and Wumeng Mountain area. Water resource scarcity is severe in the contiguous destitute areas (Zhao et al., 2018). Additional precipitation helps alleviate water scarcity, thus improving irrigation conditions for agriculture and providing basic security for the development of nonagricultural industries, thus stimulating economic growth. The areas in which average annual precipitation had a negative effect on economic growth were mainly located in Tibet, the northern Tibetan areas of four provinces, the Liupan Mountain area, and the Luliang Mountain area. An increase in precipitation was detrimental to economic growth in these areas. We speculated that this result was related to the uneven spatio-temporal distribution of precipitation in the aforementioned areas. Such variation can easily lead to heavy precipitation and flooding or drought, which not only directly cause economic losses but also have a considerable negative effect on economic productivity and people’s lives (Liu et al., 2015).

Market location had a negative effect on the economic growth of 64.57% of the counties. The counties for which distance to the provincial capital or regional megacities had a strong negative impact were mainly located in the Wumeng Mountain area, Wuling Mountain area, western Qinba Mountain area, and western Dian-Gui-Qian Rocky Desert area. These areas experienced faster economic growth when the distance to megacities was smaller. Market opportunities from the provincial capital or regional megacities and the “spread effect” of the agglomeration economy were prominent in the aforementioned areas and resulted in the generation of a strong driving effect on the economies of the counties in this area. This result is consistent with the findings of relevant studies (Wang et al., 2015b; Holl, 2018). However, economic growth in 35.43% of the counties was positively influenced by distance to megacities; the closer a county was to the provincial capital or regional megacities, the slower its economy grew. The counties strongly affected by the aforementioned distance were mainly located in the Luoxiao, Yanshan-Taihang, northern Liupan, and eastern Qinba Mountain areas. These areas are subject to considerable competitive market pressure from large cities in terms of economic development, and the economic opportunities and population levels in the aforementioned areas have decreased because of the competitive pressure in provincial capitals and regional megacities. Thus, the backwash effects of large cities inhibited economic growth in the aforementioned areas. Proximity to core cities, such as Beijing and Shijiazhuang, accelerated the outflow of resources and a decline in developmental autonomy. In addition, the polarization effect of large cities was considerably stronger than their spillover effect. The “lamp shadow area” and metropolitan shadow area gradually expanded (Chen and Sun, 2017). Therefore, to solve key problems in regional development, an economic relationship should be cultivated between counties and large cities, with large cities functioning as leaders in economic development.

Using the regression coefficients of various factors influencing economic growth (other than initial economic level), we identified the dominant factors driving economic growth for each county (Table 4). Industrial structure played a critical role in the economic growth of formerly destitute areas. The growth of secondary industry and that of tertiary industry were the dominant driving factors in 259 counties, accounting for 77.4% of the total. Additionally, the economic growth of 54 and 51 counties was dominated by government expenditure and precipitation effects, respectively, accounting for 8.1% and 7.6% of all counties. These are crucial factors that cannot be ignored when analyzing economic growth. The number of counties for which the effects of agricultural mechanization, capital investment, market location, and financial development level were dominant was relatively small.

Furthermore, according to the dominant driving factors of economic growth in each county, the modes driving economic growth in the 14 formerly destitute areas are summarized in Table 5. The economic growth modes could be divided into six general modes. The secondary industry-driven mode, in which areas were positively affected by secondary industry growth, was the most common. This mode occurred in the Wuling Mountain area, Wumeng Mountain area, southern Da Hinggan area, and Luliang Mountain area. Growth of tertiary industry was another major factor driving county economic growth, but two opposite directions for this effect were found, namely, the tertiary industry-driven mode and tertiary industry-constrained mode; the driven mode occurred in the Luoxiao Mountain area, whereas the constraint mode occurred in the Dian-Gui-Qian Rocky Desert area, the Dabie Mountain area, and Tibet. The Western Yunnan Border Mountain area and Tibetan areas of four provinces were positively affected by both secondary and tertiary industry growth, and this third mode was termed the secondary- and tertiary industry-driven mode. Additionally, the economic growth of three prefectures in southern Xinjiang was water-resource driven, undoubtedly because this area is deep inland and experiences water shortages. The Liupan and Qinba Mountain areas were affected by secondary industry and government expenditure, termed the secondary-industry-government-driven mode. The growth of secondary industry was the core driving force of economic growth in formerly destitute areas, but tertiary industry played an equally important role. However, in some areas (such as the Dian-Gui-Qian Rocky Desert area, the Dabie Mountain area, and Tibet), excessive servitization of the economic structure led to slow economic growth. In addition, government expenditure and secondary industry promoted economic growth in the Qinba and Liupan Mountain areas.

China’s poverty alleviation efforts had yielded substantial effects. In terms of modern poverty governance, accelerating the pace of economic growth in relatively poor regions and narrowing the economic gap between them and the rest of the country are critical for poverty alleviation. Identifying the spatio-temporal characteristics of and factors influencing economic growth in formerly destitute areas is of great significance for formulating future policy related to poverty alleviation and regional development. This study used a GWR model to decompose locally the regression coefficients of factors influencing economic growth in 14 contiguous destitute areas in China for determining the heterogeneous spatial characteristics of these factors and summarized the modes driving economic growth in different areas.

(1) The contiguous destitute areas had small economies and low levels of economic development. Moreover, a substantial economic gap exists between these areas and the rest of China. In 2018, the per capita GDP of the formerly destitute areas was RMB 25,679.2, only 39.8% of the national average. Since China identified 14 contiguous destitute areas as the main battlefields of poverty alleviation, the economies of these areas have generally grown rapidly; the average annual growth in per capita GDP between 2011 and 2018 in the aforementioned areas was 10.54%, higher than the national average of 9.14%. A trend of convergence in economic growth was observed among counties within these areas. Western and southern areas of the country experienced faster economic growth than central and northern counties. Spatial autocorrelation analysis revealed clear spatial agglomeration in county economies and their growth.

(2) GWR analysis indicated significant spatial heterogeneity in the effects of various indicators on the economic growth of counties in the formerly destitute areas. The effects of secondary industry growth and financial development on economic growth were mainly positive, whereas those of initial economic level and market location were mainly negative. The positive and negative effects of other indicators were approximately equivalent, with clear spatial differences being discovered. In terms of influence strength, secondary industry growth had the strongest impact on economic growth, followed by initial economic level, whereas other variables had relatively weak effects.

(3) This study discovered six modes through which economic growth was driven in the 14 formerly destitute areas. Among them, the secondary industry-driven mode was the most common and was found to have occurred in five areas including the Wuling and Wumeng Mountain areas. The next most prevalent driving modes were the tertiary industry-constrained mode, secondary- and tertiary industry-driven mode, and secondary-industry-government-driven mode. The tertiary industry-driven mode and water resource-driven mode were most prevalent in one area each, namely the Luoxiao Mountain area and three prefectures of southern Xinjiang, respectively.

The current empirical analysis revealed that spatial differences in the influence degree of various indicators, as well as the dominant factors affecting economic growth in different areas and counties, should be fully considered during the planning of economic development policy relating to formerly destitute areas; moreover, countermeasures should be taken in accordance with local conditions to improve the pertinence and scientific grounding of decision-making.

Therefore, differentiated growth strategies should be employed in accordance with the modes of economic growth in different areas. For example, in the Luliang Mountain area, the Wuling Mountain area, and other secondary industry-driven areas, policy makers should vigorously develop secondary industry in line with local resource conditions and improve the level of industrialization. For the Liupan Mountain area and other secondary-industry- government-driven areas, in addition to accelerating the pace of industrial development, policy makers should increase the government’s public expenditure, leverage the multiplier effect of government expenditure, and improve regional infrastructure and public services. For the three prefectures in southern Xinjiang, attention should be paid to water conservancy facilities and water resource allocation to alleviate the strong constraints of water resources on economic development.

In addition, the spatial heterogeneity of influencing factors should be fully considered to prevent negative effects on economic growth. For example, continuing to expand the scale of fixed asset investment and government fiscal expenditure may disrupt market operations, which may be counterproductive to economic growth in the southern Da Hinggan area and Yanshan-Taihang Mountain area. Improving the efficiency of credit use is more important than simply increasing lending in the Dabie Mountain area and three prefectures of southern Xinjiang. In Tibet and the three prefectures of southern Xinjiang, blindly promoting agricultural mechanization may have the negative effect. For the Yanshan-Taihang, Luoxiao, and Liupan Mountain areas, policy makers should not simply emphasize the improvement of external transport conditions and strengthen transportation connections with large cities but also rebuild economic relationships between large cities and formerly destitute counties as well as economic patterns involving complementary operations and positive interactions.