Introduction

The World Bank (2009) pointed out that successful urbanization in both developed and developing countries requires functional land markets, which are likely to facilitate geographic agglomeration of labor and capital. In the past two decades, there has been a renewed focus on the role of the urban land market in economic growth, urbanization, housing affordability, and spatial inequality (Li et al., 2020; Spence et al., 2008; Gao et al., 2020; Glaeser & Ward, 2009; Xiao et al., 2021; Wang et al., 2021; Wei et al., 2017). The growth of the land market becomes a critical trigger of urban development, and a healthy land market could contribute to sustainable urban development.

Economic development and urbanization have accelerated urban land sprawl and transactions and resulted in a vibrant land market in China, which has had profound impacts on housing prices and affordability (Chen et al., 2022a; Hu & Qian, 2017; Jafar et al., 2018; Li et al., 2020; Yang & Chen, 2014), industrial investment (Liu et al., 2013; Chen et al., 2018a; Chen et al., 2018b), and sustainable regional development (Wang et al., 2021; Wei, 2016). In 2020, total land conveyance fees reached 8.8 trillion yuan, while the value in 2005 was only about 0.5 trillion yuan (MLRC, 2005; MLRC, 2020), followed by a substantial increase in the urban land price. The reform of the Chinese urban land market has triggered heated debates on the reform of urban land management (Dowall, 1993; Fan et al., 2020; Xie et al., 2002), urban land expansion, structural change (Lin, 2007; He et al., 2016; Song et al., 2021), and urban land apportionment and conveyance (Deng, 2005; Yang et al., 2015) for a healthier urban growth model. A better understanding of the heterogeneous land market helps solve urban problems such as extremely high housing prices and benefits future sustainable urban development.

The early empirical studies on land prices were implemented in western countries, where market institutions have been operating for quite a long time (Colwell et al., 1998; Duranton & Puga, 2015). The recent attempts are exploring the determinants of land price in the Chinese context, especially in Beijing, Hangzhou, and Wuhan (Ding & Zhao, 2014; Hu et al., 2013; Liu et al., 2016b; Yang et al., 2015; Yang et al., 2020). However, these works analyze the land price in China from the perspective of urban economics and governance, with little attention to geography. Several dimensions are notable regarding the land’s geography, including the city’s landscape influencing land supply and the spatial effect. In recent years, the Chinese government published the “ecological redline” policy for environmental protection. The mountains, green space, and water are all potential landscapes within the “ecological redline”, which becomes a barrier to land supply (Bai et al., 2018). Thus, the distribution of natural resource influence both the volume of land supply and governments’ land supply strategy. Also, the Chinese hierarchical administrative system has been well discussed in the literature (Li et al., 2015; Wei et al., 2017), contributing to significant spatial heterogeneity in the regime. The heterogeneity also exists in different land-use types, while the industrial, commercial, and residential lands are sensitive to different determinants. However, existing knowledge is insufficient to unravel the complicated mechanisms of the land market in China, and further exploration is necessary.

Based on a land transaction dataset, this study fully considers the impacts of the regime and the heterogeneity in land-use types on the variation of land prices in China. The primary target of this study is to identify the price determinants of various urban land types from the perspective of both demand and supply sides, with particular attention to natural restrictions and contrived regulatory constraints (Rose, 1989; Saiz, 2010; Hu & Qian, 2017). This research is expected to provide twofold contributions to literature and implications to land-use planners and developers. 1. We would examine the spatial-temporal dynamics of the land market in China during different periods (2008–2015 and 2015–2020), highlighting the impacts of land-use policies on the land market; 2. The heterogeneity in how the land supply factors influence the land market regarding different time periods and land-use types will be revealed.

Literature review

Land demand, uneven regional development, and inequality in urban land prices

In a competitive urban land market, land prices are highly associated with the supply-demand relationship (Serra et al., 2004; Kheir & Portnov, 2016). According to the deprived land demand hypothesis (Tse, 1998), any increase in price and resultant demand for land can lead to more land supply. The relationship between land prices and economic development is further validated when they are influenced by similar factors (Davis & Palumbo, 2008; Wei et al., 2017). The empirical studies suggest that a new wave of land price increase was usually first observed in most developed metropolitan areas, for instance, Tokyo in Japan and Beijing and Shanghai in China, and then spilled over to adjacent areas and then to other metropolitan areas (Ando & Uchida, 2004; Zou et al., 2015). Thus, regional development becomes a crucial indicator of urban land prices.

In western countries, there has been a significant amenity-oriented turn in urban and geography research in recent ten years. In housing and land research, the accessibility to urban amenities has been highlighted in empirical studies. The most discussed one is the accessibility to public transportation (Tian et al., 2017; Wittowsky et al., 2020). The recent popularity of electronic vehicles makes charging stations bring additional value to properties in California (Liang et al., 2023). Furthermore, other determinants of housing price, such as the accessibility to parks and other leisure amenities, are hot research topics in the western research context (Carlino & Saiz, 2019; Chen et al., 2022b). In China, there are some follow-up studies (Li et al., 2019; Yuan et al., 2020); however, most are implemented at the intra-urban level.

Socioeconomic factors, including population, income, employment, and economic activities, are suggested to be the key influencing factors of urban land prices (Nicodemo & Raya, 2012). At the intra-urban level, several studies have demonstrated that uneven population distributions, levels of income, and employment opportunities significantly influence urban land prices (Kabba & Li, 2011). Also, there is heterogeneity among different types of urban land regarding influencing factors (Liu & Lin, 2014; Qin et al., 2016). For example, residential and commercial land prices are primarily influenced by demand-side factors. Spatial inequality of industrial land prices is closely related to macroeconomic conditions, such as gross domestic product (GDP) and local market effects (Davis & Palumbo, 2008; Nicodemo & Raya, 2012).

However, this literature has been criticized for the assumption of flexibility in land supply due to heterogeneity among land location and quality, the objectives of market participants, and spillovers of externalities (Yeh & Wu, 1996; Zhu, 2002; Kheir & Portnov, 2016). Major cities worldwide have faced the problem of insufficient land supply and experienced rapid increases in urban land prices in recent decades (Kheir & Portnov, 2016). The urban land supply has been highlighted as a determinant of urban land price and the demand-side factor. Therefore, some studies consider both the supply and demand sides simultaneously, with extra attention to geographical scale and spatial-temporal dynamics (Du & Peiser, 2014; Liu et al., 2016a).

Natural restrictions, regulatory constraints, and inequality in urban land prices

Existing literature has highlighted the importance of land availability in explaining variation in urban land prices, which has triggered intense debates over the determinants of urban land supply elasticity. Existing debates focus on the impacts of natural endowments and government interventions on urban land supply (Ball et al., 2010; Saiz, 2010). One argued that predetermined geographic features such as water, terrain, and wetlands are barriers to urban development and induce a relative scarcity of developable land (Rose, 1989; Saiz, 2010). Thus, earlier studies usually use large water bodies, such as the ocean and lakes, to represent the natural restrictions for land supply (Muth, 1972; Rose, 1989). In follow-up studies, more natural landscape factors, such as wetlands, and slops of terrains, were introduced (Glaeser et al., 2008; Saiz, 2010). Alternatively, the area or proportion of cultivated land is usually taken as a proxy for land supply capacity for urban uses because the government could convert agricultural land into urban land if the city has abundant agricultural resources (He et al., 2016).

Even though an intriguing pattern has been observed as a combination of rising land/housing prices and declining land supply, which does not appear to reflect any lack of land (Glaeser & Ward, 2009). There is a spatial mismatch in that the price-coldspot cities devote a greater share of urban land, while the price-hotspot cities have inadequate urban land supply but are partially constrained by man-made regulations. The existing literature has found that tightening land-use policies of development permits, urban growth boundaries, density limitations, and zoning can lead to an insufficient supply of urban land (Glaeser & Ward, 2009; Ball et al., 2010; Jackson, 2016), while flexible regulation usually contributes to more elastic land supply (Oikarinen et al., 2015).

Further studies highlight the synergies of natural restrictions and regulatory constraints on shaping land supply (Saiz, 2010; Kok et al., 2014), which accounted for 40% of interurban price differentials in Rose (1989) ‘s work and 80% of the elasticity variation in Oikarinen et al. (2015) ‘s work. Huang and Tang (2012) found that restrictive residential land-use regulations and geographic land constraints were associated with the booming housing market. The roles of geographic restrictions and artificial regulatory constraints also vary over spaces and across scales (Aura & Davidoff, 2008; Glaeser & Ward, 2009). These inconsistent conclusions suggest a vague relationship between natural restrictions, regulatory constraints, and urban land prices, which still need further exploration.

Strategic land supply, demand-supply mismatch, and inequality in urban land prices in China

In China, the market-oriented economy is gradually replacing central planning controls concerning the leasing and transfer of land-use rights. However, due to the incomplete land market and ambiguity concerning property rights in China (Zhu, 2005), both the central and local governments have the power to intervene in the urban land markets and influence the land-leasing prices through its urban land supply system, monopolization of land supply in the primary land market, and choice of land transaction methods (Tao et al., 2010). Also, the impacts of the transaction methods on land price are highly sensitive to the local context, while there is a significant difference between Beijing and Shanghai concerning Guapai (two-stage auctions) (Hui et al., 2023). The central and local governments usually have mixed effects on the price-determining mechanism because of different land-use types (Zhang, 2000). It can lead to the problems of land resource misallocation among different types of urban land and cities with different administrative hierarchies and sites (Huang & Du, 2016; Huang & Du, 2017; Yuan et al., 2019).

The reliance on land finance contributes to the capitalization of urban land happened when the local governments have powers of expropriation, allocation, and conveyance (Fan et al., 2020; Han et al., 2020; Wu, 2022). Two goals are driving the local governments’ differentiated strategies on land supply and leasing: to increase the supply of industrial land by subsidizing costs to attract more investment (Tao et al., 2010; He, Huang et al., 2014; Xu et al., 2017; Fan et al., 2020), and to maximize their revenues to finance public urban infrastructure projects and compensate for shortfalls in revenue from industrial land leasing (Fan et al., 2020). The local government adopts a discriminatory pricing strategy to maximize its total land revenue (Lu & Wang, 2020). Land available for residential and commercial developments is usually offered at low magnitudes and high prices, while land for industrial development is usually offered at very low prices and high magnitudes in terms of the benefits from the business operations, such as employment growth and (in)direct taxes (Lu & Wang, 2020; Tao et al., 2010). Such strategies contribute to a significant rise in the ratio between housing land price and residential land price after 2015, highlighting the role of political power in land market (Bhatt et al., 2023).

Land supply is also taken as an instrument to mitigate regional inequality, and unequal land supply would contribute to regional inequality. China’s unequal industrial land supply would make uneven urban management efficiency, influencing sustainable urban development (Xu et al., 2020). Governments’ intervention plays a critical role in shaping the land market, and more land supply sometimes contributes to housing destocking and booming real estate (Shen et al., 2018). The uneven residential land supply between city and county regions contributes to housing wealth inequality in China and intensifies the gap.

On the other hand, the Chinese government recently published an ecological redline policy to mitigate eco-environmental problems, while ecological redlines are usually insurmountable boundaries surrounding natural landscapes (Xu et al., 2019). It might contribute to land-use change and different land supply strategies in China, as the primary target is to protect the arable farmland from tackling the food challenge in the future (Bai et al., 2018; Wang et al., 2019). So far, there is limited knowledge of how the natural landscape influences land prices in China, which is worthy of further exploration.

The issues discussed above are all sensitive to local contexts, contributing to the spatial inequality of the urban land market in China. This research primarily aims to reveal the spatial-temporal dynamics of land prices in China, while spatial inequality of the land market is expected to keep rising following rapid economic growth. Also, land supply is suggested as a critical trigger of the land market, and natural restrictions might limit land supply and influence land prices. Therefore, we add the natural restrictions to the land price framework to analyze the determinants of urban land price inequality of demand-and supply-side fundamentals (see Fig. 1). Location significantly influences the volume of urban land demand, including the city rank and the distance to the urban center. Also, urban land demand triggers urban and economic growth, and the social and economic development factors such as population density and GDP are employed as indicators of land demand.

Fig. 1: Analytical framework.
figure 1

In this framework, urban land price is anlyzed from both supply and demand sides. The supply side considers natural conditions and supply strategy and demand side considers economic and social demands. Also, parcel trait factors would be taken as the control variables.

Full size image

We pay particular attention to the supply-side factors, including natural supply/restrictions and the strategic land supply. Although the natural supply of land might influence the allocation of land resources and influence strategic land supply indirectly, there is evidence in 35 Chinese cities suggesting no significant interaction between natural and strategic land supply (Dong, 2016). The basic characteristics of urban land available are the most discussed ones and are also the most critical determinants of urban price. On the other hand, the local governments have their land allocation strategies, mainly focusing on economic profit and influencing the supply of different land-use types. Nature supply/restrictions primarily focus on the natural amenities and the terrain factors which would contribute to spatial heterogeneity. This framework is expected to comprehensively explain China’s large-scale spatial inequality of urban land prices.

Data and methodology

Data

Thirty-one provincial administrative units in China are selected as the study area, which can be grouped into three regions based on their locations: eastern, central, and western (Wei et al., 2017). This study investigates spatial inequality in urban land prices at intra-city and inter-city levels. City-level land-leasing data were obtained from the China Land and Resources Statistical Yearbook (2009–2018) and and the data for 2019 and 2020 was obtained from China’s land market website (http://www.landchina.com). City-level sociodemographic and macroeconomic data were obtained from China City Statistical Yearbooks (2009–2021).

Parcel-level land traits were from China’s land market website (http://www.landchina.com), which is the largest nationwide land transaction information platform in China. The primary land market refers to the leasing of land-use rights to buyers for a fixed period through one of four transaction types, namely, negotiated sales (xieyi zhuanrang), and three types of auctions, two-stage auctions (guapai), English auctions (paimai), and sealed bids (zhaobiao). The database contains more than 1.6 million land transaction records by the end of 2020. For more details, the record for each parcel transaction includes values, date, address (location), parcel size, transaction method, and usage.

We focus on the transactions of residential, industrial, and commercial land parcels based in 2008, 2015, and 2020. The global financial crisis happened in 2007, contributing to a global decrease in housing prices. The housing market was used to combat the global financial crisis in China, which contributed to a rapid growth in housing prices after 2008 (Zhang et al., 2012). Another significant increase in housing prices in China was found around 2015, which has been reported in many large metropolitans because of the loosened housing purchase and mortgage loan policies (Duan et al., 2021; Zhao et al., 2020). Therefore, we examine the land-use change between 2008 and 2015 when the land market was created, and between 2015 and 2020 when the land market was booming. Negotiated sales is taken as the non-market transaction and three kinds of market transactions have been identified: two-stage auctions, English auctions, and sealed bids. The incomplete recorded high and low transaction values have been removed (bottom and top 1%), which is suggested in a previous study by Qin et al. (2016).

Methodology

The spatial inequality of land supply is examined using the Gini coefficient, Theil’s T statistic, and the coefficient of variation (CV). These three indices are all used in this research as they are sensitive to different dimensions. In this study, the spatial inequality of industrial, residential, and commercial land in different regions is separately examined regarding the heterogeneity (Wei et al., 2017). Then, we applied hot-spot analysis (Getis–Ord Gi*) to investigate the changing spatial patterns in urban land prices between 2008, 2015, and 2020. We also used ordinary least squares (OLS) regression and the spatial regime model to examine the driving mechanism behind urban land values.

Based on existing studies, urban land prices can be calculated using the OLS approach as follows:

$$\ln P_i = F\left( {S,\,E,\,G,\,L} \right) = \beta _o + \mathop {\sum}\nolimits_j {\beta _jS_j} + \mathop {\sum}\nolimits_k {\beta _kE_k} + \mathop {\sum}\nolimits_l {\beta _lG_l} + \mathop {\sum}\nolimits_t {\beta _tL_t} + \varepsilon ,$$
(1)

where Pi represents the average land price of observed land parcel i, S represents the urban land supply in the city where the parcel is located, E represents the level of economic and social development in the city, G represents the effect of local government multi-tasking in the city, L represents the land parcel characteristics, βo, βj, βk, βl, and βt are coefficients to be estimated, and ε is an error term.

Regarding the heterogeneity between different regions, a spatial regime model is employed. This technique enables us to derive the specific parameters of cities in particular spatial regimes. In our model, the effect of the explanatory variables on land prices is assumed to be sensitive to location. Thus, we allow the estimations of the coefficients to vary across the eastern, central, and western regions of China. The basic formula for the spatial regime model used in this study is as follows:

$$\left( {\begin{array}{*{20}{c}} {P\left( {{\mathrm{East}}} \right)} \\ {P\left( {{\mathrm{Central}}} \right)} \\ {P\left( {{\mathrm{West}}} \right)} \end{array}} \right) = \left( {\begin{array}{*{20}{c}} X & 0 & 0 \\ 0 & X & 0 \\ 0 & 0 & X \end{array}} \right)\left( {\begin{array}{*{20}{c}} {\beta \left( 0 \right)} \\ {\beta \left( 0 \right)} \\ {\beta \left( 0 \right)} \end{array}} \right) + \left( {\begin{array}{*{20}{c}} {C\left( {{\mathrm{East}}} \right)} \\ {C\left( {{\mathrm{Central}}} \right)} \\ {C\left( {{\mathrm{West}}} \right)} \end{array}} \right) + \left( {\begin{array}{*{20}{c}} {e\left( {{\mathrm{East}}} \right)} \\ {e\left( {{\mathrm{Central}}} \right)} \\ {e\left( {{\mathrm{West}}} \right)} \end{array}} \right),$$
(2)

where P presents the average price for each land parcel, which is an N × 1 vector, X is a matrix that represents the explanatory variables, β contains the coefficients for the explanatory variables in X, C is a constant term, and e is the error term.

Furthermore, it is challenging to determine the importance of the variables using traditional regression models. Thus, a machine learning-based model, the regression tree model, is employed for further exploration, which can help identify the non-linear relationships between the variables and reveal the relative importance (Xiao & Wei, 2023). The model results are expected to help answer the question about the importance of natural restrictions and government interventions. The regression tree models are implemented using the R package “rpart” and python library “sklearn”. In the part of regression analysis, the determinants of all these three land uses will be examined separately.

Variables

The dependent variable is the average land price (LP) for each industrial, residential, and commercial land. Drawing upon the existing literature, we attributed an average land price to four variables shown in Table 1: nature restrict, strategic land supply, economic and social development, and land parcel characteristics (control variables). To distinguish the different mechanisms between the 2008–2015 and 2015–2020 periods, we employ two models using the land price in 2015 and the land price in 2020 as the dependent variables.

Table 1 The definitions of dependent and independent variables.
Full size table

Since the end of the planned economy era, China has constructed a dual land system based on unique property rights relating to either state-owned land in the cities or collective-owned land in the countryside (Ding, 2003). First, land supply depends on the availability of land resources. The developable urban land in the plain regions is usually more than in the mountain regions, while it is costly to utilize the urban land with high slopes. Also, the high coverage of green and water space would lead to urban land fragmentation and decrease the value of urban land. Thus, the slope (SLOPE) and elevation (ELEVATION) describe the terrain and the percentages of water (WATER) and green space (GREEN) to represent the accessibility to nature amenities (Glaeser et al., 2008; Saiz, 2010). Furthermore, we use the per capita area of cultivated land (POPCUL) to represent the developable land endowment, which is expected to have a negative effect on land prices (Wen & Goodman, 2013; He et al., 2014). In practice, land supply is much more complicated because of governments’ land-use regulations and land hoarding (Du & Peiser, 2014). Thus, we also introduced the per capita area of leased land for industrial (POPCIND), residential (POPCESL), and commercial (POPCCOM) uses from 2008 to 2020, to qualify the real land supply, which is respectively added to the model. We also employ the changing rates of different land-use between 2009, 2015, and 2020 (RESCHANGE for residential land, INDCHANGE for residential land, and COMCHANGE for commercial land) to estimate the changing focus of land supply. The dummy variable NEGS represents the influence of a local government’s choice of land transaction methods on land values (1 for negotiated sale and 0 for two-stage auction, English auction or a sealed bid).

Second, land demand is determined by several economic and sociodemographic factors. GDP, foreign direct investment (FDI), and fiscal revenue per capita are all good indicators for the level of economic development of a city (Du & Peiser, 2014; Liu et al., 2016a). Fiscal revenue per capita (POPER) was employed as the economic indicator to measure the fiscal pressure on local governments and thus the incentive for them to place upward pressure on urban land prices (Liu et al., 2016a). However, these factors are highly interrelated. Thus we combine GPD, FDI, and fiscal revenue factors using principal component analysis, represented by the variable “ECONOMIC”, to avoid multicollinearity problems. The first principle is taken as an economic indicator for regression analysis, containing over 90% of the total variance of these three variables. We used population density (POPDEN) to examine the influence of land demand (He, Huang et al., 2014; Huang et al., 2015; Wu, et al., 2023) and the proportion of total output accounted for by tertiary industry (THR) to capture the influence of industrial restructuring on urban land values (Liu & Lin, 2014; Liu et al., 2016a).

In addition, three variables are used to capture the parcel-level market power, including the size of land parcels (SIZE), the location attributes of the land parcel (GRADE) (Qin et al., 2016), and a variable regarding whether the land was a new parcel (SOURCE). Furthermore, China has several levels of administration, and cities at a higher administrative level have more power in relation to the administration of the urban land market and land supply (Li et al., 2015). An administrative hierarchy variable for city rank (CR) helps identify the differences between the centrally administered municipality, sub-provincial city, or provincial capital, and the variable URBAN reveals the differences between cities and counties. More details are available in Table 1.

Uneven development of the urban land market in China

The spatial inequality in urban land prices is closely linked to China’s urban land market reform and China’s economic reform (Li & Wei, 2010). In the centrally planned economy prior to 1978, land transactions were banned, and all urban land was administratively allocated free of charge to socioeconomic units (Danwei). The demand for deepened reform and a desire to attract foreign direct investment contributed to China’s urban land market (Xie et al., 2002). A gradual and partial approach was applied to the development of the urban land market, similar to the approach used concerning other aspects of economic reform (Naughton, 1996; Wu et al., 2006). Initially, the old land allocation system was challenged by foreign enterprises’ increasing demand for land (Ding, 2003). Thus, in the mid-1980s, the reform of the urban land market commenced in a few coastal cities, such as Guangzhou, Shenzhen, and Foshan, whereby foreign enterprises were allowed to pay for land-use, and land transactions were constitutionally banned at that time (Yeh, 2005).

The spatially uneven economic reform in China makes the coastal cities become pioneers of urbanization and industrialization, leading to the uneven development of the urban land market. In 2015, the total area of and revenue from leased state-owned construction land in the eastern region of China was 0.1 million ha and two trillion yuan, respectively, accounting for 45% and 62.8%, respectively, of all leased land in China (see Table 2). There was a decline in the area of leased land but an increase in revenue from leased land in the eastern region as a share of the total area of leased land and total revenue, respectively, during the period 2008–2015 because of limited land supply and relatively higher urban land prices. Although the total value of leased land has increased from 2015 to 2020, the ratio for each region is highly stable, suggesting that the inequality was not intensified.

Table 2 The changes of urban land prices in China, 2008, 2015 and 2020.
Full size table

In general, more market-oriented transaction methods were adopted in 2015. However, the land area leased by two-stage auctions, English auctions, and sealed bids in the eastern region dropped from 55.4% in 2008 to 44.2% in 2015. This decline could be because governments in the western and central region also tried to use market tool for land transaction during 2008 and 2015. The ratios did not change much after 2015 because the land markets in China in highly developed and all the regions, including eastern, western and central regions have utilized the land market. Therefore, a significant decrease in the proportion of revenue was found in eastern China in 2020, while there was no significant decrease in the area of leased land using the negotiated sale. Thus, the discriminatory price strategy on industrial land sale was even strengthened from 2015 to 2020.

Scale and geography also serve essential roles in urban land expansion in China (Wei et al., 2017) and are critical to the urban land market. We measured the regional inequality and spatial-temporal dynamics of urban land prices during the periods 2008–2015 and 2015–2020 using three Gini coefficients, the CV, and Theil’s T statistic (see Table 3). As shown in Table 3, the increasing spatial inequalities of industrial and commercial urban land prices in eastern regions were mainly found between 2008 and 2015 because there was a “race to the bottom” in the eastern region to attract foreign investment (Huang & Du, 2016). The spatial inequalities decreased after 2015 as the discriminatory price strategy on industrial land was strengthened in the eastern region, as we mentioned above (Lu & Wang, 2020). In the central region, the rising spatial inequalities of industrial urban land prices were mainly found between 2015 and 2020, following the rise of the Chengdu-Chongqing and Wuhan metropolitan regions (Xie et al., 2021; Zheng & He, 2022). Regarding residential land, the general patterns were similar to the ones for industrial urban land prices. The difference here is that the spatial inequality in the eastern region kept rising after 2015. This phenomenon corresponds with the rising housing inequality in China, and the housing prices in the core cities were skyrocketing (Wang et al., 2021). In the western region, the spatial inequalities are stable, and there was no significant rise or drop during our study period. The analysis of the inequality in urban land price suggests the heterogeneity regarding regime and land-use type in different time periods, which would be fully considered in further analysis

Table 3 The spatial inequality of urban land prices in 2008, 2015 and 2020.
Full size table

More specifically, the changes in land prices from 2008 to 2020 are presented in Figs. 2–4 for industrial, residential, and commercial land-use. The increase in land price from 2008 to 2015 was mainly found in the eastern coastal regions, including the delta regions and Beijing. In central China, the increases in land prices were mainly found in the Chengdu-Chongqin metropolitan region, suggesting the effects of the Great Western Development Strategy. Comparing the changes in the industrial, commercial and residential land from 2008 to 2015, we do not detect a significant difference, as the land market was developing rapidly during that period. After 2015, the industrial land market got weak, and the increasing clusters were only found in the Pearl River Delta. Oppositely, the commercial and residential land market grew rapidly in the eastern coastal region as the growth in central and western China slowed down. Therefore, distinguished spatial-temporal patterns could be found for industrial, residential, and commercial land prices from these three figures.

Fig. 2: Change of industrial land prices from 2008, 2015 and 2020.
figure 2

These two figures present the changes of industrial land prices during two periods, 2008–2015 and 2015–2020 at city level. The cities with red color have higher industrial land price while the blue ones have lower industrial land price.

Full size image
Fig. 3: Change of residential land prices from 2008, 2015 and 2020.
figure 3

These two figures present the changes of residential land prices during two periods, 2008–2015 and 2015–2020 at city level. The cities with red color have higher residential land price while the blue ones have lower residential land price.

Full size image
Fig. 4: Change of commercial land prices from 2008, 2015 and 2020.
figure 4

These two figures present the changes of commercial land prices during two periods, 2008–2015 and 2015–2020 at city level. The cities with red color have higher commercial land price while the blue ones have lower residential land price.

Full size image

The analysis above provides an overview of the spatial-temporal dynamics of the land market in China. The spatial inequality of commercial and industrial land decreased after 2015, while spatial inequality of housing prices kept rising during the study period. We also identified the impacts of discriminatory price strategy on land price, which was getting popular in the eastern coastal region after 2015.

Determinants of uneven urban land prices in China

Results of the spatial regime model

We utilized the spatial regime model to investigate the driving forces behind urban land prices for industrial, residential, and commercial use, respectively, in mainland China. The variance inflation factors suggest that there were no significant collinearity problems among the independent variables. The model results are presented in Table 4, and the R-square values of these three models are all above 0.2, which shows the explanatory power of our models for the urban land price.

Table 4 The result of the spatial regime model for years 2015 and 2020.
Full size table

The results of the models for the 2008–2015 period suggest that urban land prices are clearly highly correlated with the available land resources. The negative coefficients of POPCUL for all types of land-use indicate that a scarcity of land resources has become a restrictive factor to land supply, which may lead to prices soaring (Huang et al., 2015). From the perspective of strategic land supply, industrial, residential, and commercial land per capita have different impacts on different types of land prices. In contrast to the government’s direct intervention during the planned economy era and the early days of economic reform, land supply is a tool for local governments to promote economic development (Wei & Leung, 2005; Yuan et al., 2014).

According to Table 4, residential land supply would contribute to higher residential land prices. The general mechanism of the supply-demand relationship in land market suggests that increasing residential land supply contributes to lower land prices. The model in 2008–2015 for residential land shows the opposite result, highlighting the government intervention’s significance. It is possible that the local government would increase residential land supply while residential land prices increase to maximize economic profit. Also, increasing land supply would make a prosperous land market that helps destock both land and houses and increase the land price. On the other hand, industrial and commercial urban land prices manifest different mechanisms, while the price would decrease following the increase in land supply. The discriminatory pricing strategy of the local government (Lu & Wang, 2020) works, as the local government provides industrial and commercial land at low prices to attract more investment.

Economic and social conditions are also critical determinants of industrial, residential, and commercial land-use. As expected, economic indicators significantly contribute to higher land prices, while well-developed cities usually have better economic development. CR is negatively associated with land price, indicating that the higher-level cities such as directly controlled municipalities, sub-provincial cities, and provincial capital cities, have better infrastructure systems and more job opportunities and are beneficial to the appreciation of land assets because of the hierarchical system in China (Li et al., 2015). The land parcels located in urban districts have price premiums compared with those found in counties or county-level cities, while the coefficients of URBAN were positive in both Table 4. The coefficients of NEGOS are significantly negative for the model on industrial land price, indicating that negotiated sales can reduce industrial land prices. In other words, the more market-oriented transaction methods can increase industrial land prices. On the other hand, the residential and commercial land for negotiated sales is worth more than the ones for the open land market in China. Thus, bidding or auction cannot contribute to higher residential and commercial land prices.

Generally, the results related to the land parcel-level variables are the same as our expectations. A “plattage” effect is found in land transactions, whereby smaller land parcels attract a price premium (Bai et al., 2014). Not surprisingly, the expropriation of land in counties is suggested as a strategy to save the cost of developing new urban land. The coefficients of GRADE are significantly negative, consistent with the findings of previous studies (Qin et al., 2016).

Comparing the models in Table 4, we find some notable differences. First, the cultivation land supply and green space are not significant variables in the model in 2015–2020 for industrial land presented in Table 4. Therefore, the negative impacts of natural restrictions on industrial land prices have been reduced, while natural restrictions still influence residential and commercial land prices. Also, negotiated sales significantly reduced the industrial land price after 2015, as the coefficient NEGOS became positive in Table 4. This result corresponds with the popularity of the discriminatory price strategy on industrial land using negotiated sales.

Results of regression tree model

The primary target of the regression tree model is to identify the relative importance of the variables to urban land prices. The results are presented in Fig. 5, which only includes seven important variables in the models for the 2008–2015 and 2015–2020 periods. The cultivation of land supply is the dominating determinant of the industrial land price before 2015, and socioeconomic development factors are next to cultivation. Thus, industrial land price is primarily influenced by natural restrictions such as the cultivation of land supply rather than the land market during 2008–2015. The industrial urban land market is dominated by the government, while the targets of industrial land transactions are to promote investment and increase the tax base (Ding, 2003; Tao et al., 2010; He, Huang et al., 2014; Xu et al., 2017).

Fig. 5: Results of the regression tree models for 2015 and 2020.
figure 5

It presents the results of the regression tree models for different land use types. The bars represent the relative importance of each variable and the blue one is for the data in 2015 and orange one is for the data in 2020. a The model results taking industrial land price as the dependent variable; b the model results taking residential land price as the dependent variable; c the model results taking commercial land price as the dependent variable.

Full size image

After 2015, there was a decline in the importance of cultivated land supply, suggesting that the negative impacts of natural restrictions on industrial land prices have been reduced. Economic and population density are the variables with rising relative importance because socioeconomic factors are becoming more and more important in industrial investment. Also, there is a sharp decline in the importance of the urban variable because the land in county regions has been fully utilized after 2015, and the gaps between cities and counties were closing.

Residential land price is more associated with government intervention such as biding method and size than the natural restrictions factors. Selling residential land is a critical source of land finance for local government, and government interventions make the mechanism of the residential land market complex which is usually sensitive to time (Huang & Tang, 2012; Wu, 2022). In the long run, the residential land price is negatively associated with supply, while the governors would try to make quick profits from selling residential land during the residential land price rise. On the other hand, natural restrictions such as water and green space are less likely to be the real “restrictions” for residential land. The literature suggests the positive effects of natural amenities on residential land (Yuan et al., 2019), making natural factors less likely to limit residential land prices.

After 2015, the impacts of natural factors on residential land prices have been further reduced, and the differences regarding land sources have become a critical determinant. According to Table 4, the new residential parcel’s land price is usually lower than the existing ones, and such gaps have been widened. The primary reason is that the early residential land market would target the land with more potential for residential development. Thus, the existing residential lands have more advantages in location than the new parcels. Even though the dominating determinants of residential land are still the size and the transaction type because the local government still relies on land finance and has to control the residential land transaction.

For the commercial land price, the cultivation land supply is also an essential barrier to commercial land sale, which even grows after 2015. In recent years, the development of commercial centers in China is getting more professional, requiring the support of urban amenities and visiting people (Cheng, 2021). It is difficult for cultivated land to benefit from commercial activities and reduce the potential of commercial urban land. Thus, the contradiction between cultivated land supply and commercial activities has been intensified.

On the other hand, commercial urban land within the urban district is one of the dominating determinants of commercial urban land prices. However, with the great development in rural and county regions of China (Liu, 2018), the gap between city and county is closing, and urban is not the only choice for commercial land sale after 2015. Furthermore, government intervention is less mentioned here because the local government prefers to provide a low price to attract more investment. Thus, they pay more attention to the profits made by the commercial infrastructure rather than the commercial land price. Such a trend continues after 2015, while the determinants of commercial land prices do not change much.

Synthesizing the results from the regression tree model and the reality in China, we can draw some conclusions regarding the relationship between natural restrictions and land prices. First, industrial land price is primarily determined by natural restrictions. After 2015, such negative impacts from natural restrictions on industrial land prices have been reduced as the market force represented by socioeconomic factors is rising. Residential land is dominated by institutional power as land finance is still an important source of the government budget. The negative impacts of natural restrictions have been significantly reduced as environmental factors are more and more emphasized in the housing market after 2015. Commercial land price is highly sensitive to its location and the natural environment, such as cultivated land supply. It is also notable that the gaps in the land sale between city and county regions are closing after 2015, which benefits from the great development in rural and county regions of China.

Conclusions

Land supply is considered a key to decoding the urban land price with unique urban land reform. As a major input for production and an important source of wealth, land plays a vital role in urban development, economic growth, and regional sustainability. The urban land market in China is highly complex and fragmented under the dual land-use system, which has already been introduced and has been well-researched to understand the driving factors behind the economic growth, rapid urbanization, and inequality in regional development in China (Wu et al., 2006; Lin, 2007; He, Huang et al., 2014; Wei et al., 2017).

This study investigates the spatial inequality and dynamics of urban land prices in China from 2008 to 2020. Our findings suggest that the industrial land and commercial land price inequality at the city-level rose rapidly from 2008 to 2015 and decreased from 2015 to 2020, particularly in the eastern regions. We assume that the popularity of discriminatory price strategies for better investment attraction contributes to the change of inequality. The inequality keeps rising for residential land from 2008 to 2020 because the reliance on land finance contributes to extremely high residential land prices in the core cities. In the western region, the inequality does not change much from 2008 to 2020, indicating that the land market there has not be fully utilized.

The further analysis focuses on the underlying mechanism in the context of uneven regional development and heterogeneity in land-use types, with particular attention to natural restrictions and land supply. We find that natural factors such as cultivated land supply are the critical negative determinant of land price, while the regression tree model reveals that the cultivated land supply is usually an important determinant of land price. The land supply of the specific land-use does not play an important role in determining urban land price as we expected, and the market forces represented by socioeconomic factors are more important after 2015. Furthermore, we find that the discriminatory pricing strategy on industrial land is getting more and more popular after 2015, which significantly reduces the industrial land price. These outcomes enrich the literature on the determinants of urban land price in China, considering natural restrictions other than institutional power and land features (He et al., 2016; Lu & Wang, 2020; Lin, 2007; Song et al., 2021; Yuan et al., 2019).

These findings also could provide implications for some recent social-ecological coordinating planning strategies. The “ecological redline” is the primary one that aims to protect the environment regarding urban land development. It is suggested to potentially help reduce the trade-off between ecological protection and development (Bai et al., 2018). However, limited research was available on the relationship between “ecological redline” related issues, such as green space and biodiversity, and the urban economic system. Our research reveals that green and water negatively impact land prices, and the “ecological redline” policy is expected to profoundly impact land supply. These results call for attention to urban and economic systems, such as the land market, while delineating the “ecological redline”, which is an ongoing work in China.

The territorial space planning strategy is another national planning strategy in China that limits urban built-up areas (Liu & Zhou, 2021). It makes the increasing supply of new land parcels difficult in urban regions. Our research outcomes show that the existing parcels are usually more expensive than the new ones, particularly for the residential parcels after 2015. Also, the residential land market is unequal in China, and land supply strategies are getting more important in determining land prices than natural restrictions. The current territorial space planning strategy and “ecological redline” both limit land supply and would challenge the development of the housing market, which is one of China’s economic mainstays. These facts make barriers to real estate developers as residential land storage is a driving determinant of their competitiveness, which would influence many industries in China (Li et al., 2021). Optimizing the residential land supply strategies might be critical to maintaining the vitality of the housing market in China.

Moreover, we find that the gaps between city and county regions in urban land prices are closing, as the city-county difference becomes a weak variable to explain the variance in the land price. Development in rural and county regions has gained currency in China, and the value of rural and county region have risen a lot. So far, there are insufficient empirical studies about the changes in city and county land prices caused by the development in recent years. Since the focus on rural and county regions will continue in China, more efforts are expected to be devoted to land research.

These research outcomes contribute to a deeper understanding of the spatial-temporal dynamics of the national land market in China and could bring some implications for national or regional land-use policies. First, the temporal variance in the spatial inequalities in western China manifests a less developed land market compared to the eastern cities. Therefore, the land markets in the western regions should draw more attention to marketed-oriented strategies for land-use planning in China. For the eastern cities, making better use of existing urban land should be the key concern, while current national planning strategies like “ecological redline” and territorial space planning strategy would limit land supply. Currently, the discriminating pricing strategies on industrial land might contribute to low-efficient industrial land-use in eastern cities of China (Tu et al., 2014). Future land-use planning policies should balance the relationship between land supply, pricing, and sustainable growth, particularly in eastern regions whose economic development is at a high level. Finally, the homogenization of city and county land markets provides initiatives to promote land market development policies in county and county-level city. Regarding the fact that there is intensive land-use in the urban regions of China, the land market in county region is promising.