Unveiling the Regional Differences and Convergence of Urban Sprawl in China, 2006–2019

Abstract

There is an obvious imbalanced regional development among eastern, central, and western China. This is also a fundamental problem that policy makers and planners need to address. Specific to urban development, we wondered whether there were regional differences in urban sprawl and whether this trend was under control. By using the urban sprawl index (USI), this paper investigated the spatiotemporal pattern of urban sprawl from 2006 to 2019, and its regional difference and convergence among eastern, central, and western China. It finds that the cities with high, medium, and low sprawl in the east and west regions are distributed with a clear geographical pattern, while the distribution in the central region has no intuitive geographical features. Also, the proportion of cities with high sprawl in the eastern region is more than that in the other regions, with low sprawl in central China and medium sprawl in the western region. Moreover, urban sprawl in all three regions showed a downward trend, but this process was fluctuating and had obvious phase characteristics. It can be concluded that there is a convergence trend in urban sprawl in China over the research period, and the club convergence effect exists in the eastern, central, and western regions.

1. Introduction

Since the proposition of the neoclassical growth theory in the mid-1960s by Solow [1], the disparities in regional growth have become a hot topic attracting the attention of many scholars. There is a consensus that imbalanced regional development will lead to serious social problems, and correspondingly reduce the welfare generated by high growth [2,3,4]. Therefore, reducing regional growth differences became an important issue for regional development in many countries. Urban sprawl is a byproduct of regional growth phenomena, and scholars have different views on it. Some scholars have highlighted the negative effects of urban sprawl, such as energy consumption and air pollution [5,6], environmental degradation and wildlife loss [7], economic inefficiency and agricultural decline [8,9], and unequal public services as well as social segregation [10,11]. In contrast, others have emphasized the benefits of urban sprawl, such as affordable housing, free parking, free movement, sufficient space, and yards and neighborhoods with green areas [12,13]. Meanwhile, sprawl in the suburbs provides space for the renewal of inner cities, leading to the update of existing infrastructure, which provides a good opportunity for the economic recovery of the core areas [14]. Actually, the pros and cons of urban sprawl depend largely on the extent of the sprawl. Excessive urban sprawl may lead to “urban disease”, but it is difficult to generate an agglomeration economy if the urban scale is too small [15]. Since 1978, China has experienced rapid economic reforms. At the same time, the Chinese government began to reflect on the failure of the balanced regional economic development strategy of the previous 30 years, and proposed a regional unbalanced development strategy, which was first implemented in the eastern areas, such as Zhengjiang and Guangdong provinces [16,17]. This strategy not only significantly increased the urbanization rate of Chinese cities from 17.92% in 1978 to 60.6% in 2019, but also increased the regional differences among the east, middle, and west. This process has led to the rapid expansion of cities, which inevitably leads to different levels of urban sprawl in various regions. Meanwhile, the country also faces serious challenges arising from unbalanced regional development and intensifying social injustice, which may threaten national unity and social stability [18]. The Chinese government began to consciously reduce regional differences. For example, China began to implement the western development strategy in 2000 to narrow the regional disparities. In 2014, China liberalized the restrictions on the settlement of small cities which further promoted the development of the central and western regions. These policies have made some achievements. Scholars have confirmed that the inequality between regions has decreased since 2004 [18]. However, the central and western regions of China have limited capacity to undertake industries, and the supply of infrastructure and public services is insufficient, which is not conducive to urban sprawl. Therefore, it is necessary to further evaluate the regional differences and changes in urban sprawl in China, which will provide a basis for solving the social and economic problems brought about by it.

Previous studies have highlighted the spatial pattern and driving factors of urban sprawl in China. There is still a lack of convergence analysis of urban sprawl among regions. Therefore, it is still unclear whether the inequality of urban sprawl in the east, middle, and west has been alleviated under the efforts of a series of policies of the Chinese government. Clarifying the above issues is of great significance for formulating regional development policies in the future. This paper aims to address this issue, including analyzing the spatiotemporal distribution, regional differences, and convergence of urban sprawl in China.

The remainder of this paper is structured as follows. Section 2 is the literature review including the prevailing definition and calculation of urban sprawl, and related convergence research. Section 3 briefly introduces the study area and explains the data as well as methodology. Section 4 presents the regional differences in urban sprawl in China. Section 5 further analyzes the convergence of urban sprawl in China. Conclusions are given in Section 6.

2. Literature Review

2.1. Definition and Characteristics of Urban Sprawl

Before discussing the definition of urban sprawl, it is necessary to distinguish the concepts of urban expansion and urban sprawl. There is a consensus that urban expansion is the expansion of urban land, especially construction land [19], and urban sprawl is a special kind of urban expansion, often associated with negative expansion and excessive resource consumption [14]. However, the theory and practice have not yet proved that urban sprawl is harmful without any benefit; understanding urban sprawl from a negative perspective lacks sufficient evidence, and does not meet the needs of social and economic development in backward regions [20]. Thus, increasingly studies have chosen to use urban sprawl to describe urban development. On the one hand, urban expansion characterized by increasing land area cannot answer the question of whether the urban development model is competitive [21]. On the other hand, urban sprawl can not only express the land expansion, but also reflect the characteristics of population and economic activities, which provides a comprehensive understanding of urban development.

The definition of urban sprawl remains ambiguous. Galster et al. regarded urban sprawl in terms of the processes of land use behavior and land development and proposed eight measuring indicators, which include concentration, clustering, proximity, mixed-use, kernel-density, density, clustering, and centrality [22]. Yue et al. defined urban sprawl as the encroachment of urban lands to non-urban lands occurring beyond the built-up area [23]. Similarly, Grigorescu et al. pointed out that urban sprawl is a low-density, unplanned, and discontinuous growth at the urban boundary [24]. Empirical studies in Sri Lanka and France have also verified the low-density development model of urban sprawl [25,26]. Following these ambiguous definitions, existing research has identified the characteristics of urban sprawl from different perspectives. The major one is that the growth rate of land use area exceeds that of population, which builds a foundation for quantifying urban sprawl [27,28,29,30,31,32]. Moreover, some scholars have identified the characteristics of urban sprawl from the spatial configuration, and generalized as it circle layer, leapfrog, commercial ribbon, and scattered shape [33,34].

2.2. Measurement of Urban Sprawl

Generally, urban sprawl calculation methods have been gradually enriched, evolving from a single indicator to a multi-dimensional index. With the progress of remote sensing, the map-based method has been gradually applied in many countries and regions, including China [35], Spain [36], and Ghana [37]. Comparatively, the density-based method is more abundant in terms of population density and land density. Malpezzi believed that urban sprawl can be expressed and quantified using a series of special parameters, among which the simplest one is the average density of urban built-up areas [38]. Fallah et al. developed the sprawl index based on the average population density [39]. As well, some scholars have made explorations of an entropy-based method, such as Bhatta et al. [19,40,41]. However, this approach has been opposed by Nazarnia et al. for its inappropriate application in urban sprawl measurement [42].

Apart from these, Fulton and Pendall proposed the urban sprawl index (USI) by comparing the growth rates of land area and population [43]. Meanwhile, with the development of remote sensing and geographic information technology, extensive land and population data are available to researchers, which provide a better method for the measurement of urban sprawl. For example, by using nighttime light data and LandScan population data, Zhang and Pan established an urban sprawl index that is capable of capturing the detailed differences in inner-city areas [27]. Feng et al. used the same method to measure the spatial siphon and spillover effects of urban sprawl in China [29]. In addition, Wang et al. further added population density as a counter indicator in the calculation of urban sprawl [28].

In order to comprehensively sketch the degree of urban sprawl, scholars have developed multi-dimension methods. For example, Jaeger and Schwick added per capita building area to the previous index and put forward the weighted urban proliferation (WUP) to examine urban sprawl [44]. Tian et al. made an attempt to add availability of public facilities and transportation accessibility [45]. Although this method can reflect the characteristics of urban sprawl more accurately, it is difficult to apply in many small- and medium-sized cities due to the lack data. Moreover, because the indicators are not uniform, it is difficult to compare these results.

2.3. Convergence of Urban Sprawl

Scholars believe that regional differences in urban sprawl hinder sustainable socio-economic development [46]. Some research has cast light on the convergence of urban sprawl. Wei et al. found land sprawl in Chinese cities tends to converge within cities, but not among provinces and regions [47]. Chang et al. used the Dagum Gini (DG) coefficient and stochastic convergence to test the differences in urban sprawl in China, and found there is an inherent stochastic convergence in some regions [46]. In addition, there are also studies on the convergence of urban sprawl and land use expansion in specific regions of China, such as the Yangtze River Delta region [48] and Jiangsu Province [49]. In addition, provincial-level development policies [47], industrial development [48], investment in real estate, and per capita disposable income [49] have proven to play an important role in the convergence of urban sprawl.

3. Study Area, Data, and Methods

3.1. Study Area and Data Sources

A total of 268 Chinese prefecture and above-level cities were selected as the study areas (Figure 1). The division of eastern, central, and western regions was first seen in China’s Seventh Five Year Plan in 1986. In 1997, the central government decided to establish Chongqing as a province-level municipality and incorporate it into the western region. In 2000, the central government included Inner Mongolia and Guangxi Province in the western development strategy. So far, the current regional pattern of China has been formed. In addition, because of changes in administrative divisions, the scope of municipal districts among Chinese cities was not fixed from 2006 to 2019. In order to avoid this problem, the scope of municipal districts of Chinese cities in 2019 was employed to measure the changes in population and built-up areas in each year. The map of Chinese cities was adopted from the Map Technology Review Center of the Ministry of Natural Resources, the review number is GS (2019) 1686, and the download address is http://bzdt.ch.mnr.gov.cn (accessed on 1 August 2022). To a certain extent, the urban built-up areas do not equal inhabited areas, as some places are described as empty cities or ghost cities. The actual built-up areas, where human social and economic activities occur intensively, are identified as those with a night light brightness value greater than 1000 and a population density greater than 1000 people/km².

In addition, the nighttime lighting dataset PANDA (the first gigaPixel-level humAN-centric viDeo dAtaset) released by the National Tibetan Plateau Scientific Data Center of China was used. The download address is https://www.gigavision.cn (accessed on 1 August 2022). These data have a higher quality (the average root mean square error (RMSE) is 0.73, the determination coefficient (R²) is 0.95, and the linear slope at pixel level is 0.99). Moreover, their correlation with socio-economic indicators is better than that of all existing products [50]. As for population data, the LandScan High Resolution Global Population Dataset released by Oak Ridge Laboratory is the highest-resolution population distribution dataset currently available in the world [51]. Other data were derived from the corresponding statistical yearbooks, including the China Urban Statistical Yearbook and the China Urban Construction Statistical Yearbook. It should be noted that this research spans a long period of time, and there are data missing for certain cities. The missing ones are supplemented with the data from adjacent years of the city or estimated using the linear interpolation method.

3.2. Methods

(1)

Measuring urban sprawl index

With the development of remote sensing and computer technology, the quantitative investigation of urban sprawl has aroused the interest of scholars. It can better compare the differences in sprawl between cities or regions, and also better compare with other indicators to analyze the relationship between urban sprawl and society, the economy, and ecology. There is a consensus that the quantitative method of urban sprawl measurement should be simple, low-data-requiring, reliable, and robust [52]. Therefore, the human-land-based method is employed in this research. Referring to the existing literature [27,28,29], the calculation method of the urban sprawl index is as follows.

$$USI=\sqrt{SP\ast SA}$$

(1)

where USI is the urban sprawl index, which is between 0 and 1. The closer its value is to 1, the higher the sprawl level. SP indicates the degree of low population density, and SA is the degree of low land use. Their calculation formulas are:

$$SP=0.5\ast PL-PH+0.5$$

(2)

where PL is the proportion of the urban population density lower than the national average in the total population density of the city, and PH is the proportion of the urban population density higher than the national average in the total population density of the city. In this paper, the average population density in China is based on the year 2006, and the value is 5246 persons/km². SP is between 0 and 1, with values closer to 1 indicating that the population tends to develop in low density.

$$SA=0.5\ast AL-AH+0.5$$

(3)

where AL is the proportion of the urban land area lower than the national average in the total area of the city; AH is the proportion of the urban land area higher than the national average in the total area of the city. The average urban land area in China is 41,966 km² based on the data in 2006. SA is between 0 and 1. The closer its value is to 1, the more land use tends to develop at low intensity.

(2)

Convergence analysis model

Convergence analysis models are mainly divided into two categories: σ convergence and β convergence. σ convergence refers to the USI variance or discrete coefficient of different cities tending to decrease over time. β convergence is to investigate the trend in USI in different cities from the perspective of growth rate, and the formula is as follows. If the coefficient of β convergence is negative, cities with low USI value sprawl faster than cities with high USI.

$$\ln \left(\frac{US{I}_{\dot{i},t+1}}{US{I}_{i,t}}\right)=\alpha +\beta \ln US{I}_{i,t}+{\displaystyle \sum }_{k=1}^f{\theta }_k{X}_{k,i,t}+{\epsilon }_{i,t}$$

(4)

where $US{I}_{\dot{i},t+1}$ and $US{I}_{i,t}$ represent the urban sprawl level of city i in year t + 1 and year t respectively, and $\ln \left(\frac{US{I}_{\dot{i},t+1}}{US{I}_{i,t}}\right)$ indicates the growth rate of urban sprawl. $\alpha$ and $\beta$ represent a constant term and the convergence coefficient, respectively. Meanwhile, $X_{k,i,t}$ is the kth control variable of city i in year t and $\theta$ is their coefficient. $f$ means the number of control variables.

Based on Formula (4), a spatial econometric model with an inverse distance as the spatialized weighted matrix was established to fully consider the spatial dependence among adjacent cities. This model is characterized by a spatial lag in terms of the dependent variable, independent variable, and error term. Different spatial econometric models have been established with combinations of different lags [53], and the basic one is shown in Formula (5), where W is the spatial weight matrix.

$$\ln \left(\frac{US{I}_{\dot{i},t+1}}{US{I}_{i,t}}\right)=\alpha +\beta \ln US{I}_{i,t}+{\displaystyle \sum }_{k=1}^f{\theta }_{k}W{X}_{k,i,t}+{\epsilon }_{i,t}$$

(5)

According to the existing literature, ten influencing factors from the natural, economic, and social dimensions are selected as the control variables. Specifically, natural factors include topography and coastline. Previous studies on Chinese cities have shown that complicated topography could restrict population mobility and urban construction [54]. The topography can be calculated using Nunn and Puga’s approach [55]. While considering the impact of sea affinity on population agglomeration, the straight-line distance to the nearest coastline is used to reflect its location conditions by reference to the method of Liu et al. [56]. Economically, the GDP (gross domestic product) of each city is introduced to examine the impact of urban economic development on the growth rate of urban sprawl. Meanwhile, the two ratios of the added value of the tertiary industry to GDP and foreign investment to GDP are added in the model. When it comes to social factors, public services, especially education conditions, have an impact on population agglomeration [57]. Thus, the number of full-time university teachers in the city is used to indicate the level of public service. The development of science and technology that is embodied as the indicator of scientific expenditure in GDP also influences urban expansion. Furthermore, government intervention in economic activities, per capita road area, and urbanization rate are usually controlled in empirical research on urban development [58]. The statistical descriptions of the variables are shown in

Table 1. Summary statistics of the variables.

Variables	Interpretation of Variables	Mean	SD	Min	Max
lnUSI	urban sprawl index	−0.85284	0.165233	−1.4198	−0.56345
lnTerrain	the relief amplitude of landform	0.620498	1.553331	−3.7575	3.150394
lnCoastline	the straight-line distance from each city to the nearest coastline	7.238634	1.636031	1.859935	9.166862
lnGDP	gross domestic product	14.37422	1.11476	12.21549	17.72753
IS	industrial structure	43.91392	10.95389	20.28	72.17
FDI	the degree of opening up	0.018926	0.018542	0.000199	0.087189
lnPS	public service	7.547359	1.37624	3.401197	10.84556
STL	scientific and technological level	0.002415	0.002289	0.000062	0.012717
lnGI	government intervention	−1.12821	0.360196	−2.0992	−0.47575
lnProad	per capita road area	2.301607	0.573459	0.760806	3.697344
UR	urbanization rate	0.475496	0.174171	0.1407	0.9344

Note: n = 3484.

, and the workflow schema is shown in Figure 2.

4. The Spatiotemporal Distribution and Regional Differences of Urban Sprawl in China

Using the Jenks natural breaks method to classify the USI in 2006, and based on this classification standard, cities were divided into three categories: high-sprawl cities (USI = 0.48–1), medium-sprawl cities (USI = 0.42–0.48) and low-sprawl cities (USI = 0–0.42). Visualizations through Arcgis 10.4 are shown in Figure 3. It can be found that, in eastern China, the cities with high values of USI are mainly located in Heilongjiang, Jilin, and Shandong Provinces. The USI value of southeastern coastal cities is generally low. From the perspective of temporal–spatial evolution, the urban sprawl in the northeast and southeast regions of eastern China has gradually declined, and increased in central regions. In central China, the spatial distribution of urban sprawl has no obvious geographical pattern. In term of temporal–spatial evolution, the sprawl of cities around Wuhan shows a trend of increasing first and then decreasing, such as Xianning and Huanggang. Moreover, the USI in some cities in the southwest of central China has shown a continuous downward trend. In western China, the urban sprawl is generally high in the northern and southern regions, and low in the middle parts. As for temporal–spatial changes, the urban sprawl in northern cities of western China such as Hulun Buir and Chifeng continues to decline. Cities in the central west, such as Jiuquan, Yulin, and Qingyang, showed a trend of decline first and then growth. Meanwhile, it is relatively stable in the southwest with little change.

In order to further understand the differences in urban sprawl among regions, we analyzed the changes in regional average USI (Figure 4) and the proportion of three types of cities in their respective regions (

Table 2. The proportion of three types of cities in different regions.

Proportion of Cities (%)	High Sprawl			Medium Sprawl			Low Sprawl
Year	2006	2012	2019	2006	2012	2019	2006	2012	2019
Eastern region	36.4	27.9	39.8	38.9	34.7	24.5	24.7	37.4	35.7
Central region	33.7	31.1	31.1	29.8	20.7	31.1	36.5	48.2	37.8
Western region	35.6	24.6	28.7	38.3	43.8	47.9	26.1	31.6	23.4

). Figure 4 shows that the average USI during the research period showed a downward trend; that is, the USI values of all cities and regions in 2019 were lower than those in 2006, which indicates that Chinese cities have become relatively compact. In detail, the average values for eastern cities in 2006 and 2019 were 0.46028 and 0.44541, respectively, a decrease of 3.34%. As for central cities, the values were 0.44368 and 0.43819 in 2006 and 2019, respectively, a 1.25% reduction. When it comes to western cities, the USI declined by 2.27% in research period, with values of 0.44998 and 0.44001 in 2006 and 2019, respectively. This illustrates that the urban sprawl in the east decreased the fastest, followed by that in the west, and the central cities was the lowest.

This process is also rolling, that is, the USI showed a unilateral downward trend from 2006 to 2010, followed by an increase first and then a decreasing trend from 2010 to 2013, and a stable trend from 2013 to 2016, as well as an upward-fluctuating trend from 2016 to 2019. In detail, during 2006–2010, the USI of eastern cities decreased by 4.07% on average, and that of cities in the central and west shrunk by 2.11% and 2.44%, respectively. It can be seen that the eastern cities shrunk faster during this period, while those in the west and the middle were similar. However, the USI fluctuated sharply between 2010 and 2013, showing a trend of increasing first and then decreasing, more obvious in the eastern cities. Among them, the USI of eastern cities rose by 5.45% at first, but it dropped rapidly to the level of 2010 in only one year. This increasing tendency was 2.25% and 1.38% in central and western cities, respectively, and all cities rapidly dropped to the level of 2010. From 2013–2016, the USI was stable at 0.42–0.43. After 2016, USI showed a rising but volatile trend. In this stage, the eastern and western regions spread rapidly, while the central regions were relatively stable.

According to the classification results of the Jenks natural breaks method of USI, cities were divided into three categories: high-sprawl cities (USI = 0.48–1), medium-sprawl cities (USI = 0.42–0.48), and low-sprawl cities (USI = 0–0.42). On this basis, we further compared the regional differences in urban sprawl in China. Table 2 reflects the proportion of the three types of cities in their respective regions.

According to Table 2, there was no significant difference among the three regions in the proportion of high-sprawl cities in 2006. However, in 2012, the proportion in the east and west was significantly lower than that in the middle. By 2019, the proportion of high-sprawl cities in the eastern region exceeded that of the central region again. In terms of medium-sprawl cities, their proportion in the western region was significantly higher than that in the other two regions, while it showed a downward trend in the eastern cities and a decline first and then growth in the central cities. As for low-sprawl cities, the central region enjoys the highest proportion, followed by the east, and finally the west. The above results prove the regional differences in urban sprawl. However, whether these differences are reduced is still nebulous. Therefore, convergence analysis is needed.

5. The Convergence Analysis of Urban Sprawl in China

In this section, the convergence analysis model is employed to further investigate whether urban sprawl in China has marginal decline and finally tends to be stable. In

Table 3. Convergence analysis of USI among all cities.

		Model Ⅰ	Model Ⅱ	Model Ⅲ	Model Ⅳ
	β	−0.286 ***	−0.279 ***	−0.280 ***	−0.285 ***
		[0.0185]	[0.0190]	[0.0188]	[0.0187]
Natural factors	lnTerrain		−0.0925 **	−0.0897 **	−0.0749 *
			[0.0419]	[0.0404]	[0.0400]
	lnCoastline		0.0590	0.0742 *	0.0996 **
			[0.0421]	[0.0404]	[0.0396]
Economic factors	lnGDP			0.0160 ***	0.00354
				[0.00483]	[0.00504]
	IS			0.000973 ***	0.000364 **
				[0.000138]	[0.000160]
	FDI			−0.00725	0.0319
				[0.0911]	[0.0874]
Social factors	lnPS				0.00721 **
					[0.00345]
	STL				1.295 **
					[0.522]
	lnGI				−0.0122 ***
					[0.00441]
	lnProad				0.00782 **
					[0.00302]
	UR				0.0505 ***
					[0.0152]
Convergence rate (%)		0.0259	0.0251	0.0252	0.0258
Half-life (year)		26.76	27.61	27.50	26.86
	Constant	−0.246 ***	−0.609 **	−0.996 ***	−1.100 ***
		[0.0157]	[0.283]	[0.289]	[0.281]
	Observations	3484	3484	3484	3484
	R-squared	0.170	0.173	0.190	0.205

Note: *** p < 0.01, ** p < 0.05, * p < 0.1.

, Model I show the analysis results of the absolute β convergence. Models Ⅱ to Ⅳ are outcomes of the conditional β convergence analysis after gradually adding natural, economic, and social factor variables. The β coefficients of all models are negative, which indicates that there is a convergence effect on the sprawl of Chinese cities. That is to say, cities with a lower degree of sprawl show a latecomer advantage and catch-up effect, and the difference in expansion narrows among regions. Apart from that, the convergence rate and the half-life in absolute β convergence are 0.0259 and 26.76, respectively, which are similar to that of condition β convergence. It indicates that it will take 26–27 years for cities with low sprawl to catch up to cities with high sprawl following the current trend.

Due to spatial spillover effects, regional differences and geographical location play an indispensable role in the mechanism of urban sprawl. In view of this, a spatial econometric model was established with an inverse distance as the spatialized weighted matrix. The premise of the spatial econometric models is that variables are spatially autocorrelated. Therefore, the Moran’s I of the dependent variable was calculated, as presented in

Table 4. Spatial autocorrelation test.

	Moran’s I	z-Value	p-Value
All cities	0.216	104.086	0.000
Eastern cities	0.282	55.181	0.000
Central cities	0.225	34.537	0.000
Western cities	0.177	24.716	0.000

. The analysis results showed that the convergence of urban sprawl had spatial autocorrelation.

To increase the robustness of the results, SLM, SEM, and SDM models were used for statistical testing based on the inverse distance space weight matrix. The results in

Table 5. Results of spatial econometric models.

		SLM	SEM	SDM
				Main	W
	Β	−0.290 ***	−0.266 ***	−0.299 ***	0.172 ***
		[0.0128]	[0.0105]	[0.0131]	[0.0579]
Natural factors	lnTerrain	−702.5 ***	−781.3 **	−635.5 ***
		[48.28]	[343.2]	[102.4]
	lnCoastline	702.6 ***	781.3 **	635.5 ***
		[48.28]	[343.2]	[102.4]
Economic factors	lnGDP	−0.00397	−0.00610	−0.00508
		[0.00504]	[0.00425]	[0.00504]
	IS	−5.99 × 105	−0.000123	−0.000126
		[0.000174]	[0.000159]	[0.000175]
	FDI	−0.0541	−0.100 *	−0.0475
		[0.0755]	[0.0539]	[0.0753]
Social factors	lnPS	0.000565	−0.00537	0.000236
		[0.00309]	[0.00369]	[0.00309]
	STL	0.0299	0.0574	0.0604
		[0.462]	[0.298]	[0.460]
	lnGI	−0.00803 *	−0.00806 **	−0.00756 *
		[0.00447]	[0.00325]	[0.00446]
	lnProad	0.00534	−0.000693	0.00406
		[0.00328]	[0.00286]	[0.00330]
	UR	0.0172	0.00137	0.0141
		[0.0143]	[0.0119]	[0.0143]
Convergence rate (%)		0.0263	0.0237	0.0273
Half−life (year)		26.35	29.24	25.38
	Observations	2948	2948	2948
	Log likelihood	4653.6306	6307.5650	4657.8081
	Pseudo R2	0.0006	0.0004	0.0006

Note: *** p < 0.01, ** p < 0.05, * p < 0.1.

show that the convergence coefficients of the three models are negative, which indicates that the convergence trend of Chinese cities is still obvious while controlling for the spatial spillover effect. Meanwhile, the coefficients of β in Table 2 and Table 4 are similar, indicating that the analysis results are clearly robust. In addition, after adding the spatial econometric model, the convergence rate and half-life had no significant changes compared with the results in Table 2. Therefore, it can be concluded that the urban sprawl in China has obvious convergence, and the cities with slower sprawl have a catch-up effect. The catch-up time is about 26 years, if the current trend is maintained.

After verifying the convergence of urban sprawl in China, we further checked the club convergence effect in each region (as presented in

Table 6. Convergence analysis of different regions.

		Eastern Cities				Central Cities				Western Cities
		OLS	SLM	SEM	SDM	OLS	SLM	SEM	SDM	OLS	SLM	SEM	SDM
	β	−0.302 ***	−0.292 ***	−0.300 ***	−0.301 ***	−0.276 ***	−0.406 ***	−0.418 ***	−0.416 ***	−0.303 ***	−0.525 ***	−0.462 ***	−0.572 ***
		[0.0275]	[0.0190]	[0.0195]	[0.0195]	[0.0328]	[0.0361]	[0.0371]	[0.0371]	[0.0266]	[0.0321]	[0.0365]	[0.0328]
Control variables		Control	Control	Control	Control	Control	Control	Control	Control	Control	Control	Control	Control
Convergence rate (%)		0.0276	0.0265	0.0274	0.0275	0.0248	0.0400	0.0416	0.0413	0.0277	0.0572	0.0476	0.0652
Half-life (year)		25.11	26.15	25.29	25.20	27.94	17.32	16.66	16.78	25.02	12.11	14.56	10.63
	Constant	−0.843 *				−2.428 **				0.349
		[0.479]				[1.102]				[0.306]
	Observations	1534	1298	1298	1298	1001	539	539	539	932	511	511	511
	Log likelihood		1932.1087	1931.4517	1933.7315		877.7208	876.7439	878.3240		961.7371	1039.3148	971.8178
	R-squared	0.226				0.186				0.240
	Pseudo R2		0.0007	0.0004	0.0007		0.0008	0.0008	0.0008		0.0006	0.0004	0.0006

Note: *** p < 0.01, ** p < 0.05, * p < 0.1.

). The results of OLS in different regions showed that the coefficients of β were significant and negative at the level of 1%, which was still the case after adding the spatial econometric model, indicating that there is a club convergence effect of urban sprawl in eastern, central, and western China. However, in terms of convergence rate, there is a certain heterogeneity among the three regions. The half-life of the central and western regions decreased significantly in contrast to their OLS analysis result after adding the spatial econometric model. This means that convergence in the central and western regions has remarkable spatial spillover effects. That is, the western region has the fastest convergence speed, followed by the central region, while the eastern region has the slowest convergence speed.

6. Conclusions

The unbalanced regional development strategy of the Chinese government has led to regional differences in urban construction, although a great deal of research and policy is trying to narrow this gap. However, for urban sprawl, it is far from enough. This is because there is a prejudice against urban sprawl and most people believe that all cities should control their sprawl. However, as we mentioned above, this position lacks sufficient evidence. Although excessive urban sprawl may lead to “urban disease”, it is difficult to generate an agglomeration economy if the urban scale is too small. Meanwhile, unlike urban expansion, which is limited to a single land perspective, urban sprawl provides us with a richer dimension to understand cities, such as the perspective of human–land relationship. Therefore, it is meaningful to analyze the regional differences and convergence tendency of urban sprawl in China, which will help us to have a more comprehensive understanding of China’s urban and regional development, and provide planners and managers with a decision-making basis. In this context, we measured the regional differences and convergence of urban sprawl in China.

In terms of spatial distribution and spatiotemporal evolution of urban sprawl, we found they are different in eastern, central, and western China. The cities with high, medium, and low sprawl in the east and west are relatively concentrated and have obvious geographical characteristics. However, the distribution of these three types of cities in the central region is chaotic, with no obvious geographical pattern. In addition, in the central and southern areas of the eastern region, the evolution of urban sprawl is in the opposite direction. The cities with high urban sprawl in the former are gradually increasing, while the cities with low urban sprawl in the latter are gradually increasing. The spatiotemporal evolution of urban sprawl in central China is mainly reflected in the reduction in urban sprawl around the core city (Wuhan). Moreover, we found the high-sprawl cities in the western region are mainly concentrated in the resource-based cities in the north and the economically backward cities in the south, with a decreasing trend. For example, Hulunbeier, Chifeng, Yulin, and Jiuquan in the northern region of western China are dominated by coal, steel, gold, and other resource mining, while Hechi, Baise, Guigang, and other ethnic areas in the south are particularly backward in economic development. Planners and managers should pay more attention to such cities.

The urban sprawl index (USI) showed a downward trend during the research period. This process has obvious stage characteristics. It can be roughly divided into four stages: 2006–2010, 2010–2013, 2013–2016, and 2016–2019, which respectively correspond to a unilateral downward trend, growth first and then a downward trend, a stable trend, and a wavelike rising trend. We found the USI in 2010–2013 fluctuated significantly. This phenomenon is reminiscent of the financial crisis that swept the world in 2008. The Chinese government issued a CNY4 trillion investment plan to expand domestic demand. In particular, CNY900 billion of this plan was used for affordable housing projects to increase support for housing construction. This policy has accelerated the expansion of urban land. At the same time, the population of China did not grow rapidly, which eventually led to significant sprawl. In addition, because of the lag of urban construction and policy effects, this exogenous impact was reflected after 2010 and began to weaken in 2012. Moreover, the proportion of cities with high, medium and low sprawl in each region is also significantly different.

We also found that there is a convergence effect in China’s urban sprawl; that is, the urban sprawl growth rate that was originally high slows down, while the urban sprawl speed that was originally slow becomes relatively fast. According to the current trend, it will take about 26 years for the latter to catch up with the former. In addition, the eastern, central, and western regions have club convergence effects. Meanwhile, convergence in the central and western regions has remarkable spatial spillover effects. As for convergence rate, the western region is the fastest, followed by the central region, and the eastern region is the slowest. However, the trend of urban development is not static. Under the influence of such natural, economic, and social factors, urban sprawl will tend to shrink or become polycentric, which is a prospect that needs to be studied in the future. How to promote regional coordinated development is a long-term problem faced by managers and planners. As far as urban sprawl is concerned, it has obvious advantages and disadvantages. Due to the existence of spatial spillover effects and regional linkage effects, the level of urban sprawl in different regions should not differ too much to ensure reasonable resource carrying capacity and an agglomeration economic effect. Our model shows that adjusting the industrial structure, improving the public service and scientific and technological level, strengthening infrastructure construction, and improving the level of urbanization are conducive to narrowing the differences in urban sprawl between regions. Meanwhile, the coefficient of government intervention is negative, indicating that the current policy still strengthens the differences among regions, so it is necessary to adjust the policy attitude in time. We therefore advise managers and planners to propose more specific control and encourage strategies to accelerate the convergence of urban sprawl in China, based on the different impacts of natural, social, and economic factors.

Undoubtedly, there are still some deficiencies in our research, and we look forward to filling them in future research. First of all, our research stays in the macro perspective, although the sprawl shape of each city is included in the step of calculating USI and can be visualized. However, it is still difficult for us to refine it to such an extent in this article. Apart from that, as mentioned in the literature, a polycentric urban structure and urban sprawl may exist at the same time. Their difference is that the former is planned and organized. Therefore, it is necessary to incorporate them into the same analytical framework in future research.