The Higher Land Prices, the Worse Emission Performance? Theory and Evidence

Abstract

This study investigates whether high land prices in China have worsened polluting enterprises’ emission performance. Our model confirms that increases in pollution intensity led by rising land prices are the result of fewer resources being directed to abatement investments, which corresponds to smaller declines in emissions versus output. We use a compiled micropanel dataset to conduct the empirical analyses. It covers Chinese major industrial enterprises’ information about their pollutant emissions and treatment as well as land transfer. Our empirical mechanism exploration corroborates the theoretical findings described earlier. Also, consistent with the results of theoretical comparative statics, the promoting effects of land prices on pollution intensity are found to be more significant among less productive and dirtier enterprises, as well as those facing weaker local environmental regulations. Further heterogeneity tests highlight the roles of land transfer methods and political affiliation. This study complements the literature on the adverse environmental impacts of land finance from the view of rising land prices.

1. Introduction

As industrialization continues to advance around the world, environmental pollution has become one of the most crucial issues threatening global sustainable development. Particularly, since the 1980s, manufacturing centers have gradually shifted from developed countries to China. It makes the manufacturing sector not only an engine for unprecedented economic growth but also a primary source of pollutant emissions in China. The bulletin of the second national survey of pollution sources shows that the number of industrial pollution sources in China is 2,477,400 and accounts for nearly 70% by the end of 2017¹. Among the multiple determining factors of manufacturing production, land acts as a scarce resource and a necessary input for enterprises, and the role it plays in their emission performance, however, has been paid little attention to in either the theoretical or empirical literature.

In recent years, with the steady improvement of China’s urbanization level, a vast number of enterprises and populations have converged in cities, leading to a substantial surge in urban land demand [1]. However, although the reform of China’s land transfer system has rendered the land market increasingly market-oriented, the supply of urban land remains stringently regulated by the central and provincial governments under the dual institutional constraints of strict farmland protection and public land ownership. Consequently, land supply has not expanded significantly along with the processes of urbanization and industrialization, resulting in market demand far exceeding supply and causing a persistent imbalance that has driven land prices to rise continuously [2]. Previous studies have documented the stylized fact of rising industrial land prices in China [1, 2]. Figure 1 illustrates the temporal variation in the average price of industrial land in major Chinese cities from 2003 to 2018, showing an increase from 469 yuan to 834 yuan per square meter.

[IMAGE OMITTED. SEE PDF]

As a vital component of enterprise costs, land investment exerts a direct influence on operational expenses, with rising land prices substantially increasing the cost of business operations [1], thereby posing potential impediments to sustainable and high-quality development. Previous scholars have investigated various negative economic impacts of high land prices. For instance, elevated land costs may drive firms to relocate toward peripheral areas [3–7], limit new entry into the market [1, 8], inhibit exports [2], and reduce incentives for research and development [9]. Nevertheless, the potential environmental losses have been largely ignored. Manufacturing producers are both main users of land and main contributors of industrial wastes. Few studies attempt to explore how a persistent rise in land prices over the past years influences their emission performance from a microperspective. Therefore, this paper conducts a firm-level analysis to assess the impact of increasing industrial land prices on corporate pollution emissions. Against the backdrop of surging land costs and the pursuit of green, low-carbon development, this research holds both theoretical and practical significance.

We first explicate how rising land prices affect a microenterprise’s production and emission strategies through the lens of a seminal monopolistic competition model [10]. Two key elements incorporated into the core structure include one enterprise’s land input with exogenous given costs and its decision to undertake pollution abatement besides setting output prices. Based on the framework discussed earlier, we show that a rise in land prices leads to a reduction in the total amount of pollutants, but to a lesser extent than for total output, which corresponds to increases in pollution intensity, a commonly adopted indicator measuring the pollutants emitted per unit of output. These findings are closely tied to the substitution relationship between land and the other special factor input—pollutants. Along with this substitution effect, a deeper mechanism exploration shows that less resource allocation for abatement investments can account for the disproportionate decline of output level versus emission level. Additionally, our comparative static analyses suggest that the promoting effect of high land prices on pollution intensity is subject to the degree of one enterprise’s dirtiness and its productivity level. Meanwhile, the stringency of local environmental regulations matters.

Combining multiple microenterprise datasets over approximately 12,000 observations during the sample periods from 2000 to 2014, we empirically examine the effects of rising land prices on one enterprise’s emission performance. In particular, we use the emission intensity of either chemical oxygen demand or sulfur dioxide as the outcome of interest and transfer prices of land, the use right of which is remised to one enterprise, as the core explanatory variable. In accordance with the theoretical expectation, the positive relationship between the two is significant. Despite the decline in statistical significance, the estimated coefficients by and large keep validity when we measure pollution intensity based on other subtypes of pollutants or define it at a higher level of categories of industrial wastes and treat land concession funding as an alternative measure for the baseline regressor. The results are also robust after controlling the attributes that may reduce the efficiency of the key estimate, removing the observations with zero pollution, or mitigating the border effects of pollutant emissions.

Empirical mechanism investigation shows that declined industrial value-added output and total emissions of each pollutant are associated with rising land prices; however, their estimated absolute elasticities with respect to land prices are distinct from each other: The former is clearly larger than the latter. Possible channel exploration verifies the negative relationship between land prices and enterprises’ number of pollution treatment facilities. Then, the estimated impacts of high land prices are found to vary by type of enterprise. The estimates are significantly greater in magnitude among high-polluting groups. Conversely, the coefficients of interest on enterprises having higher productivity levels or located in regions with higher environmental protection requirements are relatively small and even statistically insignificant. Further subgroup regression results show that, compared with those participating in public auctions, enterprises privately negotiating with governments are relieved from emission pressure from high land prices. Also notably, the effects of land prices only exist among politically nonaffiliated enterprises but are absent among enterprises that are subordinate to one of the government authorities.

The coexistence of reverse causality and omitted variables potentially biases the benchmark regression estimates. To address this concern and establish causality, we take a regulatory policy implemented in 2010 as a quasiexperiment, which aims at standardization of local governments’ revenue from the sale of land-use rights and its expenditure. Our generalized difference-in-differences strategy introduces the interaction of time variation before and after the policy implementation and cross-sectional variation coming from enterprises’ transferred land areas ex ante. It shows that due to their greater exposure to this exogenous shock, enterprises with a greater level of land scale experience a greater rise in pollution intensity, compared to those with a lower level. In addition, to alleviate the interference of other unobservable time-varying features that correlate with both land prices and pollution intensity, we use the predetermined floor area ratio of each piece of land plot as an instrumental variable for its prices. The instrument estimation results are finally supportive of our baseline results.

The abovementioned empirical findings demonstrate rising land prices are a determinant of worsening emission performance at the enterprise level, but they leave spatial spillover effects of pollutant emissions untouched. Thus, we turn to generalized spatial econometric estimation based on a series of aggregate variables to conduct a tentative analysis. We find that land prices still have a promoting effect on pollution intensity at the province level, but the spatial lag terms (SLTs) of dependent variables have significantly sizable positive impacts as well, suggesting that the indirect effects of pollution externalities from neighboring regions play a relevant role in the assessment of environmental consequences of increases in local land prices.

Compared with existing research, the contributions of this paper are as follows. First, although the current literature on rising land prices predominantly focuses on economic consequences, it pays insufficient attention to environmental impacts. This paper takes the lead in systematically examining, from a microfirm perspective, the effects and underlying mechanisms of persistently increasing land prices on emission performance, thereby providing an important supplement to the existing research. Second, the paper constructs a general theoretical framework to explain how one enterprise’s production and emission decisions are affected under land resource constraints. Employing the classical model of monopolistic competition [10] illustrates how surging land prices affect firms’ production and emission strategies, revealing not only the intrinsic mechanisms but also the channels through which average and heterogeneous effects arise. Third, based on microlevel data of Chinese firms from 2000 to 2014, the paper empirically tests the theoretical propositions. Over a longer period of time, the unavailability of data sources limits the corresponding empirical research. Instead, this paper conducts a rich empirical analysis using the data of Chinese industrial enterprises and microlevel land transactions (LTs), providing empirical evidence on how industrial land prices influence enterprise emissions.

The remainder of our study is organized as follows: In Section 2, we review the related literature while eliciting our contribution. Section 3 introduces the theoretical framework. Section 4 describes the empirical design. Section 5 analyzes the empirical results. Section 6 concludes.

2. Related Literature

Three strands of literature that this study contributes to are discussed here briefly.

2.1. Regional Pollution Effects of Local Fiscal Pressure and Land Finance

First, our work is most directly related to the studies that evaluate the regional pollution effects of local fiscal pressure and land finance. Nearly 30 years since the tax-sharing system was implemented, Chinese fiscal and taxation institution reform accompanies the upward movement of fiscal power and taxation from local government to central government². Under the induced heavy financial stress, local governments are forced to race to the bottom [11] and attract outside investments through the promise of low factor prices. The measures also include lowering environmental protection to introduce enterprises, even if most of them are pollution-intensive, thereby aggravating environmental pollution [12]. Actually, the pollution-haven hypothesis that local governments’ behaviors of earning extra revenue harm the environment has been supported by a lot of empirical evidence [13–15]. As mentioned before, land finance is one of the most important policy tools to generate extra fiscal income, and the role it plays has received much attention in recent years. For example, using Chinese provincial panel data ranging from 1998 to 2006, Wang et al. [16] find that land finance continuously increases carbon emissions along with the economic development after 2003. Yang et al. [17] empirically demonstrate the negative impacts of land finance on haze pollution among 269 Chinese prefectural cities between 2004 and 2017.

From a macroperspective, most existing works connect environmental quality deterioration to the land-based development mode. They undoubtedly accord with the practices of Chinese local governments and provide this study with a realistic background. Instead of focusing on the expansion of pollution scale caused by investment promotion through land finance, the aim of our study is to explore how the rise in land prices that it induced affects polluting enterprises’ emission performance. It reflects a change in the research point of view toward the microlevel.

2.2. The Consequences of Rising Land Prices on Industries and Enterprises

The second large body of literature to which this study adds is the consequences of rising land prices. Despite the fact that Chinese local governments promote local economic growth at the expense of lowering transfer prices of industrial land, many scholars set their sights on various negative economic impacts of the continuous rise in land prices. The new economic geography (NEG) theory indicates that local high land prices, as one of the typical centrifugal forces, cause industrial enterprises to spatially relocate from center to periphery to save rent expenses [3–5]. In the case of a shortage of land supply, China currently sees the phenomenon of industry outward movement too soon from city center to suburb [6] or from coast to inland [7], not conducive to taking advantage of agglomeration economies to enhance industrial competitiveness [18]. Beyond manufacturing decentralization, rising land prices also pose a threat to the economic outcomes of enterprises, such as limiting new entry into the market [8], constraining exports [2] but accelerating their steps of outward foreign direct investment [19, 20], and reducing R&D incentives [9]. Much worse, under the positive feedback between high housing prices and high land prices [21, 22], the ill impacts above can be further amplified [23–25].

Although scholars expect the significantly negative role high land prices play on the development of industry and enterprises, current literature concentrates largely on the economic consequences instead of the environmental impacts. We still lack mechanism analyses and empirical evidence on the effects of increases in land prices on enterprises’ emission performance, which will be discussed later.

2.3. The Determinants of Polluting Enterprises’ Emission Performance

Third, this study also complements much of the foregoing work exploring what factors determine one polluting enterprise’s emission performance. From a microperspective, some theorists incorporate abatement investments and pollutant emissions into the Melitz-type trade framework [26–29]. Broadly speaking, they all argue that exporters are more environment-friendly than nonexporters. The induced abatement investments matter for the environmental implication of trade. The critical roles of productivity level and stringency of pollution control are emphasized. Both of them are supposed to be negatively correlated with pollution intensity. These theoretical inferences resonate with empirical evidence linking emission reduction with trade openness [30, 31], environmental policy and regulation [32, 33]. Apart from these factors, other empirical research investigates the effects of neutral technology innovation [34] and industry intellectualization [35, 36], which can also bring about environmental improvements to polluting enterprises.

However, most existing research puts more attention on the determinants of environmental benefits, rather than those of losses. As pollutant emission is a type of self-selection behavior, output and emission performance are endogenously subject to one enterprise’s budget constraints. Hence, two questions that come up still receive little response: whether the absence of land resources has negative impacts on both outcomes and, if so, how does it work? The orientation of effect of high land prices on enterprises’ pollution intensity remains ambiguous. This study attempts to answer the questions theoretically and empirically, while helping to clarify them.

3. Theoretical Framework

To investigate the impact of rising land prices on enterprises’ pollution intensity promotion, we follow Shapiro and Walker [28] to establish a simple theoretical framework. We describe how one enterprise makes trade-offs in the allocation of land resources in terms of production and abatement.

3.1. Demand

There exists a differentiated sector in a market environment of monopolistic competition, where the constant elasticity of substitution (CES) utility function of one representative consumer can be expressed as 1 $\begin{array}{}U={\left[{\int }_{\omega \in \Omega }{q\left(\omega \right)}^{\left(\left(\sigma -1\right)/\sigma \right)}d\omega \right]}^{\left(\sigma /\left(\sigma -1\right)\right)},\end{array}$ where Ω is a product set, including all kinds of varieties available in the market; σ > 1 represents the constant elasticity of substitution between any two kinds of varieties. Given the representative consumer’s total income E, the optimal demand for variety ω is q(ω) = (p(ω)^−σ/P^1−σ)E, where p(ω) is the price of variety ω and $P={\left[{\int }_{\omega \in \Omega }{p\left(\omega \right)}^{1-\sigma }d\omega \right]}^{\left(11/-\sigma \right)}$ is the aggregate price index that is assumed to be exogenous.

3.2. Production and Emission Technology

Each enterprise labeled as its productivity φ chooses to produce a horizontally differentiated variety. We assume that z(φ) units of pollutants would be emitted if enterprise φ produces q(φ) units of variety requiring l(φ) units of land input³. For every unit of land input, a fraction a(φ) ∈ (0, 1) is used to abate pollution, whereas the remaining fraction is used for variety production. Following Shapiro and Walker [28], enterprise φ produces q(φ) = [1 − a(φ)]φl(φ) units of output and z(φ) = {d[1 − a(φ)]}^(1/β)φl(φ) units of pollutants; meanwhile, β ∈ (0, 1) represents the pollution elasticity and d is positively related to one enterprise’s degree of dirtiness⁴. Hence, enterprise φ’s pollution intensity (its pollutant emissions per unit of output) can be expressed as 2 $\begin{array}{}i\left(\phi \right)=\frac{z\left(\phi \right)}{q\left(\phi \right)}=d^{\left(1/\beta \right)}{\left[1-a\left(\phi \right)\right]}^{\left(\left(1-\beta \right)/\beta \right)}.\end{array}$

Equation (2) shows that pollution intensity, which is positively correlated with enterprise-level degree of dirtiness, decreases with an increase in an enterprise’s abatement investments.

3.3. Abatement and Output Prices

Before a variety is put into production, enterprise φ using technology with increasing returns to scale must pay a fixed overhead cost f > 0. Then, it selects abatement investments a and output prices p to maximize its profit while covering the land prices per unit r and exogenous pollution taxes t per unit of its pollutants, both of which are regarded as exogenously given. Based on the output production function q(φ), pollutant emission function z(φ), and profit function π(φ) = p(φ)q(φ) − rl(φ) − tz(φ) − f, the profit optimization problem can be given by 3 $\begin{array}{}\underset{a\left(\phi \right),p\left(\phi \right)}{\max }\pi \left(\phi \right)=\left[p\left(\phi \right)-c\left(\phi \right)\right]q\left(\phi \right)-f,\end{array}$ 4 $\begin{array}{}\text{s}.\text{t}.c\left(\phi \right)=\frac{r}{\left[1-a\left(\phi \right)\right]\phi }+t{d}^{\left(1/\beta \right)}{\left[1-a\left(\phi \right)\right]}^{\left(\left(1-\beta \right)/\beta \right)},\end{array}$ where c(φ) denotes the enterprise φ’s variable costs of production. Enterprise φ first makes a decision on the optimal pollution control. The basic trade-off here is that more abatement investments a(φ) increase the variable land costs r/[1 − a(φ)]φ but decrease the variable pollution costs td^(1/β)[1 − a(φ)]^{(1 − β/β)}. To minimize the total variable cost, the optimal share a(φ) of inputs employed to abate pollution is given by the following first-order condition: 5 $\begin{array}{}a\left(\phi \right)=1-\frac{1}{d}{\left[\frac{\beta r}{\left(1-\beta \right)\phi t}\right]}^{\beta }.\end{array}$

From Equation (5), increases in land prices prompt enterprise φ to allocate fewer resources toward pollution control as ∂a(φ)/(∂r) < 0. We then substitute Equation (5) into Equation (4) to obtain the Cobb–Douglas-type variable cost c(φ) = (dr^1−βt^βφ^β−1)/κ, where κ ≡ β^β(1 − β)^1−β. According to the pricing rule of monopolistic competition market, the optimal price set by enterprise φ is equal to a constant markup over the marginal cost [10], which can be written as 6 $\begin{array}{}p\left(\phi \right)=\frac{\sigma }{\sigma -1}c\left(\phi \right)=\frac{\sigma d{r}^{1-\beta }{t}^{\beta }{\phi }^{\beta -1}}{\left(\sigma -1\right)\kappa }.\end{array}$

3.4. Land Prices and Pollution Intensity

More formally, combined with Equation (5), Equation (2) can be rewritten as follows: 7 $\begin{array}{}i\left(\phi \right)=d{\left[\frac{\beta r}{\left(1-\beta \right)\phi t}\right]}^{1-\beta }.\end{array}$

Clearly, a direct connection is established between enterprise φ’s pollution intensity and land prices by Equation (7). We can calculate the first-order derivative of i(φ) with respect to r: 8 $\begin{array}{}\frac{\partial i\left(\phi \right)}{\partial r}={\left(1-\beta \right)}^{\beta }{\beta }^{1-\beta }d{\left(\phi t\right)}^{\beta -1}{r}^{-\beta }>0.\end{array}$

Equation (8) demonstrates that rising land prices will lead to increases in pollution intensity.

Furthermore, to study what determines the magnitude of its promoting effect on pollution intensity, we discuss how the elasticity of pollution intensity with respect to land prices changes with one enterprise’s productivity φ, the extent of dirtiness d, and strictness of local environmental regulation t. The corresponding second-order derivatives are calculated as follows: 9 $\begin{array}{}\frac{{\partial }^{2}i\left(\phi \right)}{\partial r\partial \phi }=-{\left(1-\beta \right)}^{\beta +1}{\beta }^{1-\beta }d{\phi }^{\beta -2}{t}^{\beta -1}{r}^{-\beta }<0,\end{array}$ 10 $\begin{array}{}\frac{{\partial }^{2}i\left(\phi \right)}{\partial r\partial d}={\left(1-\beta \right)}^{\beta }{\beta }^{1-\beta }{\left(\phi t\right)}^{\beta -1}{r}^{-\beta }>0,\end{array}$ 11 $\begin{array}{}\frac{{\partial }^{2}i\left(\phi \right)}{\partial r\partial t}=-{\left(1-\beta \right)}^{\beta +1}{\beta }^{1-\beta }d{\phi }^{\beta -1}{t}^{\beta -2}{r}^{-\beta }<0.\end{array}$

From Equations (8)–(10), we have verified the following proposition.

$$

Statement

Proposition 1.

Rising land prices can lead to an increase in one enterprise’s pollution intensity, with a stronger effect on a dirtier enterprise, but the promoting effect is weaker for enterprises with higher productivity and those faced with tougher enforcement of environmental regulation.

3.5. Interpretation

To explicate the intuition underlying Proposition 1, we attempt to compare the changes in one enterprise’s total output and pollutant emissions led by an exogenous increase in land prices.

Based on the optimal price of its variety given by Equation (6), enterprise φ’s total output, which exactly equals the representative consumer’s demand, is calculated as follows: 12 $\begin{array}{}q\left(\phi \right)=\frac{{p\left(\phi \right)}^{-\sigma }}{P^{1-\sigma }}E={\left[\frac{\sigma }{\left(\sigma -1\right)\kappa }\right]}^{-\sigma }{d}^{-\sigma }{r}^{\left(\beta -1\right)\sigma }{t}^{-\beta \sigma }{\phi }^{\left(1-\beta \right)\sigma }{P}^{\sigma -1}E.\end{array}$

Then, according to technologies for production and emission as well as optimal abatement investments from Equation (5), enterprise φ’s total pollutant emissions can be calculated as follows: 13 $\begin{array}{}z\left(\phi \right)=d^{\left(1/\beta \right)}q\left(\phi \right){\left[1-a\left(\phi \right)\right]}^{\left(\left(1-\beta \right)/\beta \right)}={\left(\frac{\beta }{1-\beta }\right)}^{1-\beta }{\left[\frac{\sigma }{\left(\sigma -1\right)\kappa }\right]}^{-\sigma }{d}^{1-\sigma }{r}^{\left(\beta -1\right)\left(\sigma -1\right)}{t}^{\beta \left(1-\sigma \right)-1}{\phi }^{\left(1-\beta \right)\left(\sigma -1\right)}{P}^{\sigma -1}E.\end{array}$

Low pollution elasticity (β ∈ (0, 1)) and high elasticity of substitution among varieties (σ > 1) make it clear to find that ∂q(φ)/(∂r) < 0 and ∂z(φ)/(∂r) < 0 from Equations (12) and (13). Clearly, increases in land prices force enterprise φ to cut down its production scale and total amount of pollutant emissions as well, but it must be noted that, as shown in Equation (8), increases in pollution intensity imply that rising land prices actually lead to a disproportionate decrease in one enterprise’s total output and total pollutant emissions. Essentially, the rationale behind the result is closely related to two different effects of land price shock.

The first one is similar to what we call the income effect: As one enterprise is forced to reduce its total land input due to increases in land prices, the levels of both its total output and pollutant emissions drop, but for one enterprise, the income effects on the two outcomes are the same, which therefore do not directly lead to disproportionate decreases in them. In other words, the income effect on pollution intensity is zero. Second, rising land prices also create what we call the substitution effect: Given a fixed level of pollution taxes, for each unit of output, higher land prices imply higher costs of land input relative to pollutants emitted, which can be regarded as the other special factor input, thus decreasing land input per unit of output while increasing pollutants emitted per unit of output. Essentially, it happens because rising land prices stimulate one enterprise to reduce the share of each unit of land input used for abating, that is, abatement intensity (∂a(φ)/(∂r) < 0). Ultimately, the combination of negative income effect and positive substitution effect causes a slight decrease in total pollutant emissions compared to total output, equivalent to increases in pollution intensity.

Furthermore, we can offer an explanation of the heterogeneous effects from Equations (9)–(11). Note that the “income effect” of higher land prices on pollution intensity remains unchanged with variations in one enterprise’s productivity and the strictness of local environmental regulations, as we can see from Equation (2), but both higher productivity and stricter environmental regulation weaken the positive “substitution effect” (∂²a(φ)/(∂r∂φ) > 0; ∂²a(φ)/(∂r∂t) > 0). Therefore, the overall promoting effect on pollution intensity becomes weak with an increase in φ and t. However, a dirtier enterprise, on which the “substitution effect” of rising land prices is albeit muter (∂²a(φ)/(∂r∂d) > 0), originally emits more pollutants per unit of output than a cleaner one, as shown in Equation (2). Taken together, the overall promoting effect on pollution intensity becomes more significant with an increase in d.

Combining all the analysis results above, we further give the following proposition.

$$

Statement

Proposition 2.

High land prices will result in the shrinkage of enterprises’ production scale and total amount of pollutant emissions. However, rising land prices force enterprises to invest less in abatements, leading to higher pollutants emitted per unit of output, thus increasing their pollution intensity.

The following Figure 2 visually illustrates the mechanism details of the theoretical model: The model’s derivation shows that an increase in land prices leads to an increase in pollution intensity, caused by a smaller decrease in pollutant emissions and a larger decrease in production scale. Why do pollutant emissions and production scale decrease by different magnitudes? The key reason lies in the simultaneous occurrence of income and substitution effects due to the rise in land prices. The income effect implies that higher costs for firms lead to a reduction in both output and abatement investments. The substitution effect suggests that, to maximize profits, firms will opt to increase output while reducing abatement investments. The combination of these two effects ultimately leads to the differing magnitudes of the decrease in pollutant emissions and production scale.

[IMAGE OMITTED. SEE PDF]

4. Empirical Design

4.1. Econometric Specification

To empirically examine whether rising land prices have a promoting effect on enterprises’ pollution intensity, our econometric regression specification based on a two-way fixed effects (TWFE) model is as follows: 14 $\begin{array}{}{\left\{\text{CODI},\text{SDI}\right\}}_{it}=\alpha +\beta {\text{Ave}\_\text{Price}}_{it}+X_{it}^{\prime }{\theta }_i+X_{pt}^{\prime }{\theta }_p+{\lambda }_t+{\delta }_i+{\epsilon }_{it},\end{array}$ where the dependent variables, CODI_it and SOI_it, respectively, represent chemical oxygen demand emission (CODE) intensity and sulfur dioxide emission (SOE) intensity of enterprise i in year t; the key explained variable, Ave_Price_it, denotes the average prices of lands acquired by enterprise i in year t; and the control variables, X_it and X_pt, are a vector of time-varying observable characteristics at the firm level and province level in year t, respectively (to be discussed later). We also include enterprise fixed effect δ_i and year fixed effect λ_t to account for time trend and time-invariant unobservable factors that may interfere with the relationship between land prices and pollution intensity; ε_it is an error term.

4.2. Data Sources

We build a unique unbalanced panel dataset with a sample period spanning from 2000 to 2014 and more than 12,000 observations, by merging the following datasets mainly from three sources: First, the Annual Survey of Industrial Firms (ASIF) database, released by the China National Bureau of Statistics, covers all state-owned enterprises and private enterprises with more than five million yuan in annual sales. Second, the Industrial Firms’ Pollutant Emissions (IFPE) database, released by the China Ministry of Ecology and Environment, is the most comprehensive database including energy consumption and pollutant emissions at the firm level in China. Third, the LT database, collected from China’s land market website, contains the detailed high-frequency LT information for different regions in China, especially including names of owners with land-use rights, which allows us to match these data with the microenterprise databases above. The remaining one used for the mechanism test is the enterprise patent (EP) database, issued by the China National Intellectual Property Administration, which records detailed patent information of enterprises such as names, types, applications, and citations.

4.3. Main Variables

Chemical oxygen demand and sulfur dioxide—the representative indices of industrial wastewater and industrial waste gas, respectively—are first listed as two types of pollutants under total quantity control in the layout of the Eleventh Five-Year Plan for National Economic and Social Development of China. According to Wang and Wheeler [37] and He et al. [32], one enterprise’s pollution intensity could be defined as the ratio of its pollutant emissions to gross industrial output value. Hence, we choose two measures as dependent variables: chemical oxygen demand intensity (CODI) and sulfur dioxide intensity (SDI), using data from IFPE and ASIF. Based on the individual-level land transfer data from LT, we select the transfer prices per unit of land area supplied (Price) as the key independent variable. Obviously, a larger value of the measure means higher land transfer costs. To avoid the missing values, we add one to the actual value of each variable above before taking the natural logarithm of it. With reference to the practices of existing literature [38–42], we also add the individual and regional control variables to the regressions, aiming to eliminate their confounding effects on an enterprise’s pollution intensity. In particular, we control for the firm-level features including the enterprise’s total assets (Asset), asset–liability ratio (Leverage), profit rate (Profit), and its type of ownership (ownership), which come from ASIF. To control for a vector of time-varying provincial indicators, we further introduce the degree of market competition (HHI), industrial structure (Structure), ratio of government expenditure to revenue (Fiscal), GDP growth rate (GDP_growth), and GDP per capita (GDP_per_capita) of the province where the enterprise is located. Table 1 shows detailed categories, names, and definitions of main variables used in the following empirical research. Table 2 reports their summary statistics. In addition, to reduce the impact of outliers, we winsorize the data at the 1st and 99th percentiles.

Table 1 Type, name, and definition of main variables.

Variables	Name	Definition
Dependent variables
CODI	Chemical oxygen demand intensity	Natural logarithm of one plus an enterprise’s ratio of chemical oxygen demand discharges to gross industrial output value
SOI	Sulfur dioxide intensity	Natural logarithm of one plus an enterprise’s ratio of sulfur dioxide emissions to gross industrial output value
Independent variables
Ave_Price	Average land transfer price	Natural logarithm of one plus transfer prices per unit of land area supplied to an enterprise
Enterprise-level control variables
Asset	Enterprise size	Natural logarithm of one plus an enterprise’s total assets
Leverage	Asset–liability ratio	Total liabilities/total assets
Profit	Profit rate	Total profit/total assets
Ownership	Ownership type	Assigned as 1 if an enterprise is state-owned and 0 otherwise
Province-level control variables
HHI	Degree of market competition	Herfindahl–Hirschman index based on enterprise scale
Structure	Industrial structure	Value-added of secondary industry/value-added of tertiary industry
Fiscal	Ratio of government expenditure to revenue	General budget expenditure/general budget revenue
GDP_growth	GDP growth rate	Change in GDP over the 2-year period
GDP_per_capita	GDP per capita	GDP/total population

Table 2 Descriptive statistics of main variables.

Variables	Mean	Std. dev.	Min	Max	Obs.
CODI	0.098	0.282	0	1.828	12,739
SOI	0.269	0.464	0	2.348	12,739
Ave_Price	5.148	0.932	0.267	8.116	12,739
Asset	12.352	1.768	8.543	16.975	12,739
Leverage	0.566	0.254	0.024	1.215	12,739
Profit	0.117	0.182	−0.126	0.959	12,739
Ownership	0.151	0.358	0	1	12,739
HHI	0.153	0.15	0.014	0.748	12,739
Structure	1.199	0.217	0.697	1.783	12,739
Fiscal	1.913	0.6	1.118	3.958	12,739
GDP_growth	0.142	0.057	0.026	0.259	12,739
GDP_per_capita	10.436	0.437	9.354	11.268	12,739

5. Results and Discussion

5.1. Baseline Results and Robustness Checks

Table 3 reports the regression results of Equation (14) using a stepwise regression approach. With only the fixed effects included, results in Columns (1) and (3) suggest a 1% increase in average transfer prices per piece of land acquired by one enterprise leads to a 0.393% increase in its CODI and a 0.0416% increase in SDI, respectively. By further introducing control variables and clustering standard errors at the province level, the coefficient estimates in even columns are statistically significant at the 1% level, whereas each of them experienced a rise in economic significance. To examine the robustness of the baseline results, which show that increases in land prices lead to higher pollution intensity, we conduct several checks as follows.

Table 3 The baseline regression results.

Variables	CODI	SOI
(1)	(2)	(3)	(4)
Ave_Price	0.0393∗∗∗ (0.0062)	0.0404∗∗∗ (0.0065)	0.0416∗∗∗ (0.0073)	0.0449∗∗∗ (0.0091)
FEs	Y	Y	Y	Y
Controls	N	Y	N	Y
Cluster Province	N	Y	N	Y
Observations	12,739	12,739	12,739	12,739
R-squared	0.119	0.142	0.119	0.191

First, note that both outcome variables CODI and SOI in the benchmark equation are defined at the pollutant-specific category. According to information available from the IFPE database, ammonia nitrogen is the other pollutant in industrial wastewater discharged by manufacturing enterprises in addition to chemical oxygen demand. Instead, along with sulfur dioxide, nitrogen oxide and dust mainly come from industrial waste gas emissions. Hence, we select ammonia nitrogen intensity (ANI), nitrogen oxide intensity (NOI), and dust intensity (DI) as three alternative measures of dependent variables. As before, each of them is defined as taking the natural logarithm of one plus the actual intensity of the corresponding pollutant. The results reported in Columns (1)–(3) of Table 4 indicate that, in contrast to the estimates of two representative pollutants in Table 3, those of NOI and DI remain qualitatively similar but are smaller in magnitude. Moreover, the promoting effect of land price on ANI is rather weak and even statistically insignificant. Second, to comprehensively take various pollutant subtypes into consideration, we then redefine an enterprise’s pollution intensity based on two large categories of industrial wastes. We proceed to test whether the estimation results are robust to using wastewater discharge intensity (WDI) and waste gas emission intensity (WGEI), calculated in the same method as CODI and SOI. As exhibited in Columns (4) and (5) of Table 4, the estimated coefficients of WDI and WGEI are 0.0312 and 0.0097, with statistical significance at 1% and 10%, respectively. Third, we substitute the total land transferring fees (Total_fees)—calculated as the natural logarithm of one plus the transferring fees of the total land area an enterprise pays—for the average land prices as an alternative proxy of the key explanatory variable. The estimates in the same direction as previous results can be obtained from the last two columns in Table 4. Combined, it turns out that the relationship between high land prices and high pollution intensity is positive and robust.

Table 4 The estimation results of alternative measures of dependent variables.

Variables	ANI	NOI	DI	WDI	WGEI	CODI	SOI
(1)	(2)	(3)	(4)	(5)	(6)	(7)
Ave_Price	0.0040 (0.0024)	0.0204∗∗ (0.0081)	0.0304∗∗ (0.0123)	0.0312∗∗∗ (0.0092)	0.0097∗(0.0049)
Total_fees						0.0141∗∗∗ (0.0041)	0.0108∗∗ (0.0043)
Observations	5857	9389	1063	10,527	2622	12,740	12,740
R-squared	0.214	0.209	0.185	0.184	0.125	0.131	0.175

Beyond the substitution of dependent and independent variables, we further conduct several other robustness checks, including the use of DK standard errors, removing zero observations, and excluding boundary firms. The detailed results can be found in Table A1 of Appendix A.

5.2. Mechanism Tests

Proposition 2 uncovers the nature that one enterprise’s emission performance deteriorates under the negative impact of a land price shock. The increase in land prices results in a shrinkage of production scale and total amount of pollutants emitted. At the same time, these changes coincide with the decline in abatement investments, which leads to a slight decrease in pollutant emissions compared to output. We then examine the validity of the underlying mechanism from an empirical perspective.

The disproportionate decrease in output and emissions: Based on the benchmark regression Equation (14), we use the levels of value-added industrial output (VIO), CODEs, and SOEs as three other independent variables. In particular, we, respectively, regress the outcome variables on our regressor of interest, Ave_Price. The corresponding coefficient estimators are listed in Columns (1)–(3) of Table 5. Although they are of great statistical significance at the 5% level, these estimates suggest that a 1% increase in land prices contributes to a higher percent reduction in one polluting enterprise’s VIO, corroborating our aforementioned theoretical interpretation.

Table 5 The estimation results of mechanism tests.

Variables	VIO	CODE	SOE	WWTF	WGTF
(1)	(2)	(3)	(4)	(5)
Ave_Price	−0.0978∗∗∗ (0.0014)	−0.0208∗∗ (0.0084)	−0.0279∗∗ (0.0106)	−0.0362∗∗ (0.0135)	−0.0658∗ (0.0332)
Observations	12,739	12,739	12,739	8599	2152
R-squared	0.992	0.133	0.213	0.070	0.111

The role of abatement investments: To understand what causes the disproportionate decrease in output and emissions, we further explore the response of one polluting enterprise’s abatement investment decision to a land price shock, which plays a nondeductible role in the rise of pollution intensity as illustrated in the theoretical analyses. The availability of the IFPE database enables us to conduct two exercises by selecting the number of wastewater treatment facilities (WWTF) and the number of waste gas treatment facilities (WGTF) as proxies, each corresponding to a specific type of pollutant. Regression results using them as independent variables are reported in Columns (4)–(5) of Table 5. We find that Ave_Price has significantly negative effects on both outcome variables. It implies that the increase in land prices indeed reduces one enterprise’s incentive to abate pollution.

5.3. Heterogeneity Tests

As Proposition 1 shows, the land prices exert a differential influence on polluting enterprises’ emission performance with different characteristics, located in different areas. To test the possible heterogeneous effects of the impact, we further conduct several subgroup regressions based on firm variations in the degree of dirtiness and productivity level, as well as location variations in the stringency of environmental regulations. The estimation results for the two dependent variables—CODI or SDI—are presented in Panels A and B of Table 6.

Table 6 The subgroup estimation results of the main heterogeneity tests.

	Degree of dirtiness	Productivity level	Environmental regulations
High	Low	High	Low	High	Low
(1)	(2)	(3)	(4)	(5)	(6)
Panel A (CODI):
Ave_Price	0.0668∗∗∗ (0.0148)	0.0001∗(0.0000)	0.0077∗∗ (0.0037)	0.0696∗∗∗ (0.0132)	0.0284∗∗∗ (0.0102)	0.0374∗∗∗ (0.0112)
Observations	5880	6859	3394	9345	5573	7166
R-squared	0.178	0.261	0.142	0.169	0.106	0.277
Panel B (SOI):
Ave_Price	0.0531∗∗∗ (0.0124)	0.0002 (0.0003)	0.0075 (0.0081)	0.0759∗∗∗ (0.0140)	0.0185 (0.0149)	0.0326∗∗ (0.0125)
Observations	6397	6342	3394	9345	5573	7166
R-squared	0.337	0.088	0.110	0.243	0.284	0.224

5.3.1. High Polluting Versus Low Polluting

As land prices rise, enterprises with a higher degree of dirtiness are supposed to experience a greater magnitude of increase in pollution intensity. To confirm this theoretical expectation, we categorize the enterprises into two subgroups based on whether their chemical oxygen demand discharges or SOEs are above or below the median value in the sample. The estimation results for the heterogeneity in the impact of land prices on the two groups are listed in Columns (1) and (2) of Table 6. Indeed, the promoting effects on either CODI or SOI are stronger for high-polluting enterprises. On the contrary, the estimates are small and even statistically insignificant for the other group with a low degree of dirtiness. Therefore, the results corroborate our theoretical inference.

5.3.2. High Productivity Versus Low Productivity

As analyzed before, under the circumstance of the same factor prices, an enterprise with a higher productivity level suffers from a weaker substitution effect. Hence, with the rise in land prices, less productive resources are diverted from abatement investments. In other words, compared with less productive ones, enterprises close to technological frontiers are capable of offsetting the increase in pollution intensity to a certain extent. According to whether their number of patent applications from the EP database was above or below the sample median, we are able to sort all enterprises with different frontier levels of production technology into two groups. Columns (3) and (4) of Table 6 present the subgroup regression results from the high and low groups, respectively. In terms of the economic magnitude of the coefficient estimates in either Panel A or Panel B, the latter is more than nine times greater than the former, which means the promoting effects of land price shock are mainly observed in the low-productivity enterprises, consistent with one of our theoretical findings.

5.3.3. High Environmental Protection Requirements Versus Low Environmental Protection Requirements

To investigate how the stringency of environmental regulations works on the increase in pollution intensity, we divide China’s provinces into two subgroups based on whether their total discharge of the “Three Wastes” is above or below the sample median. The rationale behind this is that a province with stricter pollution controls tends to create less soil and water pollution, as well as greenhouse gas. The differential impacts between provinces with higher environmental protection requirements and lower ones are reported in Columns (5) and (6) of Table 6. Consistent with theoretical expectations, the promoting effect is greater for enterprises located in provinces with less emission limitation. In those areas, local governments often tend to pursue economic development at the sacrifice of the ecological environment. It seems that lowering land prices reduces an enterprise’s pollution intensity, which acts as a complementary measure for environmental regulations. On the flip side, implementing administrative penalties for pollution could also reduce the negative impact of high land prices on local enterprises’ emission performance.

In addition, we have conducted further heterogeneity analyses regarding the methods to acquire land-use rights and the presence of political affiliation. The detailed results are presented in Table A2 of Appendix B.

5.4. Endogeneity Analyses

It should be noted that the TWFE estimation method can still not address the endogeneity problem caused by other external or internal attributes that are hard to control for. First, there is no denying that land is regarded as a relevant resource of production factor, and its transfer prices will affect one enterprise’s various decisions on production, abatement investment, and pollutant emissions, but we cannot ignore the fact that over recent years in China, local governments are increasingly motivated to strengthen environmental protection. Therefore, the pressure of pollution control is likely to inhibit their motivation for land transfer [43]. Particularly, polluting enterprises are likely to be obliged to withdraw from cities’ central districts and then move out to the sparsely populated suburbs. It implies that enterprises varying with emission performance could be spatially sorted into different areas, thereby passively accepting different land prices. Therefore, there exists an issue of reverse causality in the benchmark regression. Second, our independent variables—two measurement indices of pollution intensity at the firm level, to a great extent, depend on one enterprise’s own production scale and business performance. These potential factors, highly correlated with one enterprise’s budget constraint, directly determine its affordable land prices, which typically lead to omitted variable biases. Two techniques are applied separately to solve the two endogeneity issues above.

5.4.1. Generalized Difference-in-Differences Methods

To address the concern that increased pollution intensity, caused by high land prices, may conversely make local governments impose strict environmental penalties toward highly polluting enterprises through restricting land transfer to them, we adopt generalized difference-in-differences strategies to make use of a nationwide policy targeting land transferring fees for local governments, jointly enacted by four central government authorities in 2010⁵. Under this policy, all the revenue and expenditures from land sales must be timely included in the local authority exchequer and attributed to local fund budgets. To avoid loss of revenue from land transfer, local governments are tightly restricted from allowing enterprises to delay or cut down the payment, thus to a large extent exacerbating enterprises’ budget constraints, equivalent to increases in land prices.

In particular, with reference to the practices of Nunn and Qian [44] as well as Lu and Yu [45], our generalized DID estimation equation is as follows: 15 $\begin{array}{}{\left\{\text{CODI},\text{SDI}\right\}}_{it}=\alpha +\beta \left\{{\text{Land}}_{i,20002009-},{\text{Land}\_\text{dummy}}_i\right\}\times {\text{Post}}_t+X_{it}^{\prime }{\theta }_i+X_{pt}^{\prime }{\theta }_p+{\lambda }_t+{\delta }_i+{\epsilon }_{it},\end{array}$ where we select two different types of treatment variables to construct key independent variables {Land_i,2000-2009,  Land_dummy_i} × Post_t. The first one is a continuous variable Land_t,2000-2009, which denotes the average annual land area acquired by enterprise i during the pretreatment period from 2000 to 2009. Its variation among polluting enterprises characterizes the differential treatment intensity of the policy shock. Hence, we can compare the changes in pollution intensity in the postimplementation period relative to the preimplementation period between enterprises occupying different sizes of land areas before 2010. Second, for robustness, we also employ a dummy variable Land_dummy_i as an alternative measure, which equals one (treatment group) if the average annual land area acquired by enterprise i exceeded the mean value in the sample before 2010, the time of policy implementation, and 0 otherwise (control group). Under this setting, the standard DID estimation compares one enterprise with a large size of transferred land scale before and after the implementation of the land tightened policy (the first difference) with changes among the counterparts with transferred land scale below the sample median (the second difference). Besides, Post_t is an indicator variable that equals 1 if t is the year 2010 and onward and 0 otherwise. All other variables have been defined as in Equation (14).

As presented in Table 7, the first and last two columns report the regression results obtained by using Land_i,2000-2009 × Post_t and Land_dummy_i × Post_t, respectively. The outcomes for CODI and SOI are listed in the odd and even columns, respectively. All four estimates are positive and of great significance at the 1% level. Perhaps a continuous measure of the regressor of interest captures more variation, and it yields two estimates slightly smaller in magnitude. These results imply that either CODI or SDI of enterprises with larger land scale before 2010 experiences a more substantial increase in pollution intensity after the land tightened policy took effect than that of enterprises with a smaller land scale. Altogether, the positive coefficients suggest that rising land prices due to a decline in land supply worsen enterprises’ emission performance.

Table 7 The generalized difference-in-differences estimation results.

Variables	CODI	SOI	CODI	SOI
(1)	(2)	(3)	(4)
Landi,2000-2009 × Postt	0.114∗∗∗ (0.0229)	0.109∗∗∗ (0.0171)
Land_dummyi × Postt			0.158∗∗∗ (0.0443)	0.140∗∗∗ (0.0371)
Observations	4427	4427	4427	4427
R-squared	0.569	0.537	0.556	0.525

Convincing casual identification using the generalized DID method based on Equation (15) relies on check on the pretrend assumption. To confirm whether or not similar pretreatment parallel trends have occurred among enterprises having use rights of different sizes of land, we select 2010 as the baseline year, given the year in which the land tightened policy was issued by the government. Following Naun and Qian [44], instead of interacting Land_i,2000-2009 with a post–ante indicator variable as in Equation (15), we interact the treatment intensity measure with each of the year fixed effects spanning from 2007 to 2014 in the following specification: 16 $\begin{array}{}{\left\{\text{CODI},\text{SDI}\right\}}_{it}=\alpha +{\displaystyle \sum }_{j=-3}^{+4}{\beta }_j{\text{Land}}_{i,20002009-}\times I_t^j+X_{it}^{\prime }{\theta }_i+X_{pt}^{\prime }{\theta }_p+{\lambda }_t+{\delta }_i+{\epsilon }_{it},\end{array}$ where $I_t^j$ is an indicator variable which equals 1 if t − 2010 = j, with j = −3, −2, −1, 0, +1, +2, +3, +4. The parameters of interest β_j values capture the treatment effect of the land tightened policy in each year relative to the chosen baseline time period, which is taken to be 2010 (i.e., β₀ = 0). Therefore, we are concerned with its evolutionary trend before and after the year when the policy was issued, rather than the absolute level of coefficient estimators in each period. Based on the estimation from Equation (16), Figures 3(a) and 3(b) separately plot the point estimates of the interaction term and their 95% confidence intervals from two dependent variables, CODI and SOI. We observe that before the policy implementation event, there exist no significant time trends in both estimated coefficients. However, immediately after 2010, we witness a gradual increase in the magnitude of the treatment effect on enterprises’ CODI, whereas the coefficient for the correlation between ${\text{Land}}_{i,20002009-}\times I_t^j$ and SDI almost remains at a fixed positive level after experiencing a significant increase 1 year following the policy’s occurrence.

[IMAGE OMITTED. SEE PDF]

5.4.2. Instrumental Variable Method

We resort to the instrumental variable method to tackle the identification challenge in the benchmark estimation from the unobservable omitted variables in Equation (14). With reference to the existing practices, for each piece of land being transferred, its floor area ratio, listed in land transfer data from the LT database as well, can be used as a valid instrument for its transfer prices [9]. In reality, the upper and lower limits of its permitted plot ratio are already set by the government department before land is transferred in the local land development market. Therefore, this index is supposed to be independent of one enterprise’s confounding features, which may bias the estimated effect of Ave_Price on CODI or SOI. As floor area ratios are one of the most relevant indices to measure the value of a piece of land itself, variation in land prices is thought of as the sole channel through which it links to changes in one enterprise’s emission performance. The analyses described earlier suggest that this instrument simultaneously meets the requirements of relevance and exclusion constraints. Overall, to rule out the interference of time-varying unobservable omitted variables might exist in Equation (14), we set the following first-stage regression equation: 17 $\begin{array}{}{\text{Ave}\_\text{Price}}_{it}=\alpha +\beta {\text{floor}\_\text{area}\_\text{ratio}}_{it}+X_{it}^{\prime }{\theta }_i+X_{pt}^{\prime }{\theta }_p+{\lambda }_t+{\delta }_i+{\epsilon }_{it},\end{array}$ where floor_area_ratio_it is calculated as taking the natural logarithm of one plus the floor area ratio per unit of land plots supplied to enterprise i in year  t. It denotes average floor area ratios of all land acquired by enterprise i in year t Table 8 reports the 2SLS regression results. As shown, the first-stage IV estimate in Column (1) is significantly positive, suggesting a strong correlation between land’s floor area ratios and its transfer prices. Besides, according to the rule of thumb, its Cragg–Donald Wald F-statistic exceeds the empirical threshold (note that 882.224 > 10), succeeding in passing the weak instrument test. In contrast to the baseline counterparts in Columns (2) and (4) of Table 3, both the second-stage IV estimates drop (note that 0.0131 < 0.0404 and 0.0113 < 0.0449). These results imply that potential confounding factors not controlled for in Equation (14) lead to overestimating the promoting effect of land prices on one enterprise’s pollution intensity. Therefore, our instrument regressions provide more valid estimates.

Table 8 The two-stage least squares estimation results.

Variables	First stage	Second stage
Ave_Price	CODI	SOI
(1)	(2)	(3)
Ave_Price		0.0131∗∗∗ (0.004)	0.0113∗∗ (0.005)
floor_area_ratio	5.900∗∗∗ (0.318)
Kleibergen-Paap rk LM statistic	17.433
Cragg-Donald Wald F statistic	882.224
Observations	8958	1475	1475

5.5. Spatial Econometric Analyses

Given the fact that spillover effects of waste gas or wastewater often take place between different regions, accurate evaluation of the promotion effect on pollution intensity of land prices may not be realized using ordinary least squares (OLS) estimation. The traditional DID method, a plausible causal identification of which builds upon the stable treatment unique variable (STUVA) assumption [46], also finds it hard to cope with this problem. As mentioned before, for example, under local strict environmental regulations, polluting enterprises located in the upper river area may directly discharge wastewater downstream into adjacent areas, which deteriorates water quality there with impunity. In the spirit of the negative spatial externalities of pollutant emissions, we follow Jia et al. [47] and then concentrate our analyses on the regional level rather than the individual level. We employ the following generalized spatial econometric specification to alleviate the interference of spatial correlations of variables between different provinces in China: 18 $\begin{array}{}{\left\{\text{CODI},\text{SDI}\right\}}_{pt}=\alpha +\rho {\displaystyle \sum }_{p^{\prime }}^N{w}_{p{p}^{\prime }}{\left\{\text{CODI},\text{SDI}\right\}}_{p^{\prime }t}+\beta {\text{Ave}\_\text{Price}}_{pt}+\tau {\displaystyle \sum }_{p^{\prime }}^N{w}_{p{p}^{\prime }}{\text{Ave}\_\text{Price}}_{p^{\prime }t}+X_{pt}^{\prime }{\theta }_p+\left({\displaystyle \sum }_{p^{\prime }}^N{w}_{p{p}^{\prime }}{X}_{p^{\prime }t}^{\prime }\right)\theta +{\lambda }_t+{\delta }_p+{\epsilon }_{pt},{\epsilon }_{pt}=\gamma {\displaystyle \sum }_{p^{\prime }}^N{w}_{p{p}^{\prime }}{\epsilon }_{p^{\prime }t}+{\mu }_{pt}.\end{array}$

In Equation (18), the terms of dependent variables {CODI,  SDI}_pt, independent variable Ave_Price_pt, and control variables $X_{pt}^{\prime }$, individual fixed effect δ_p, and error term ε_pt are all defined at the province level. Not only that their own SLTs are, respectively, introduced, where $w_{p{p}^{\prime }}$—the pth row and p^′th column elements of the N × N spatial weight matrix—is an indicator variable, which equals 1 if province p and province p^′ are adjacent and 0 otherwise (if p = p^′ or the two provinces are not adjacent). N = 30 denotes the number of provinces.

In practice, we aggregate the variables at the enterprise level to the province level. Limited to data availability, we finally obtain a balanced panel dataset of 30 provinces in China from 2007 to 2012. It is necessary to conduct spatial correlation tests for CODI and SOI before we decide whether to use the spatial econometric method. In particular, we draw the scatterplots and the linear fitted lines of the global Moran’s indices in Chinese provinces during the sample period, capturing the correlation between two outcomes of interest and their SLTs, as presented in Figures 4(a) and 4(b). The horizontal axis and the vertical axis, respectively, denote the standardized levels of CODI or SOI and its SLT. Clearly, we find that most points are concentrated in the first and third quadrants, that is, high or low values of outcome variables are separately clustered together, implying that there indeed exists positive spatial autocorrelation of pollution intensity among different provinces.

[IMAGE OMITTED. SEE PDF]

Actually, Equation (18) can be transformed into different spatial econometric models according to whether or not the corresponding estimated parameters are assigned zero values ex ante [48]: (1) a “spatial lag model (SLM)” if ρ ≠ 0, τ = 0, θ = 0, and γ = 0; (2) a “spatial Durbin model (SDM)” if ρ ≠ 0, τ ≠ 0, θ ≠ 0, and γ = 0; and (3) a “spatial error model (SEM)” if ρ = 0, τ = 0, θ = 0, and γ ≠ 0. The results of spatial correlation tests shown in Figure 2 indicate that, because of the spatial spillover effects of pollutant emissions, the estimates obtained from the OLS method in Equation (14) could not be consistent. Therefore, we select an SLM based on Equation (18) to reexamine the validity of the benchmark results.

With a series of provincial covariates and two fixed effects introduced, the estimated coefficients in terms of CODI and SOI are, respectively, presented in Columns (1) and (4) of Table 9. Resonating with the baseline estimates at the enterprise level, the coefficients of Ave_Price are both significantly positive. Moreover, the same goes for the SLT of CODI or SOI in the SLM, indicating that the estimated impacts of pollution externality are sizable. For robustness, we further introduce the SLT of the independent variable—Ave_Price—into the SLM, and the estimates using an SDM are listed in Columns (2) and (5) of Table 9. Significant and positive effects of the SLT on the two dependent variables remain. However, unlike pollutant emissions, we find no evidence that there exist significant spatial spillover effects of land prices themselves. Using an SEM, we also report the estimation results with only spatial weight matrices of the error term in Columns (3) and (6) of Table 9, and the effects of this term are both negligible and insignificant. Overall, these exercises show that local emission performance is not only directly determined by land prices in the local province but also indirectly affected by pollution spillovers from adjacent provinces. The negative effects of Ave_Price on regional pollution intensity might be underestimated unless we take the indirect effect into consideration.

Table 9 The spatial econometric regression results.

Variables	CODI	SOI
Models	SLM	SDM	SEM	SLM	SDM	SEM
(1)	(2)	(3)	(4)	(5)	(6)
Ave_Price	0.0423∗∗ (0.0168)	0.0424∗∗ (0.0168)	0.0419∗∗ (0.0171)	0.0828∗∗∗ (0.0220)	0.0827∗∗∗ (0.0219)	0.0823∗∗∗ (0.0227)
Spatial lag terms of	0.0652∗∗∗	0.0658∗∗∗		0.0511∗∗	0.0506∗∗
Dependent variables	(0.0173)	(0.0175)		(0.0243)	(0.0244)
Spatial lag terms of		0.0298			−0.0265
Independent variables		(0.0203)			(0.0324)
Spatial lag terms of			−0.0002			0.0194
Error term			(0.0170)			(0.0199)
Observations	210	210	210	210	210	210
R-squared	0.559	0.567	0.520	0.360	0.364	0.345

6. Conclusion

The Chinese development mode, which relies on land finance, has driven up land transfer prices, with land-use rights leased to developers by the government. However, the impact of rising land prices on the emission performance of polluting enterprises remains underexplored. This study investigates the role of land prices in emission performance from a microeconomic perspective. Using a DS-type monopolistic competition model, we show that rising land prices, coupled with a fixed pollution tax, lead to reduced investment in pollution abatement, resulting in higher pollution intensity due to a disproportionate decline in emissions compared to output. Our empirical analysis, based on a comprehensive dataset of Chinese industrial enterprises, confirms these findings and highlights the role of political connections in moderating the impact of land prices on pollution performance. This research contributes to understanding the environmental consequences of land finance, emphasizing the need for reforms to mitigate its harmful effects on both economic and environmental outcomes.

Based on the findings of this study, several policy recommendations are proposed to address the negative environmental impacts of rising land prices. First, land pricing reforms should be implemented to stabilize land prices and reduce their volatility, ensuring that increases in land costs do not disproportionately burden polluting enterprises. This can be achieved by reducing local governments’ reliance on land sales for fiscal revenue and promoting more sustainable land-use policies. Second, green finance mechanisms should be strengthened to incentivize enterprises to invest in pollution abatement technologies. Providing targeted financial support, such as green bonds or subsidies for pollution control investments, would help alleviate the financial pressure caused by rising land prices, encouraging businesses to prioritize environmental sustainability. Lastly, differentiated emission standards should be introduced, where stricter regulations are applied to more polluting enterprises while providing flexibility for cleaner, more efficient firms. This approach would ensure that environmental regulations are more equitable and responsive to the varying capacities of firms to reduce emissions. These policy measures would not only improve environmental equity but also foster a transition to greener and more sustainable industrial practices.

We acknowledge several limitations in this study that warrant further exploration. Firstly, the dataset used in our analysis ends in 2014, whereas China’s land and environmental policies have undergone significant changes since then. Notably, the implementation of the “Two Mountains” policy and the expansion of the emission trading system (ETS) pilot programs post-2015 could have had substantial impacts on enterprises’ emission performance. Future research should incorporate more recent data to examine the effects of these policy developments on pollution intensity. Additionally, our theoretical model can be extended to explore how other factors, such as financial credit constraints, rising labor costs, import competition, tax reforms, and technological advancements, influence microlevel pollution intensity. Investigating how these external shocks affect the emission performance of polluting enterprises will be an important avenue for future research.

Appendix

appendix

Appendix A

First, we use the DK standard errors proposed by Driscoll and Kraay [49] to estimate Equation (14) once more. The results are reported in Columns (1) and (2) of Table 5. We observe that the coefficient estimator on either CODI or SOI is statistically significant at 1%, suggesting that the promoting effect of high average land prices on enterprises’ pollution intensity remains robust after alleviating the concern that heteroscedasticity, autocorrelation, and cross-sectional correlation might exist in the error term. Second, note that the two proxies of pollution intensity, CODI and SOI, might be left-censored to zero during the data-generating process, as shown in their summary statistics in Table 2. Hence, to address the concern that our results are affected by the extreme value of the dependent variables, we repeat our regression in a subsample after removing the enterprises with “zero pollution.” The results in Columns (3) and (4) of Table 5 indicate that, with zero observations excluded, rising land prices yield a more pronounced positive effect (note that 0.0470 > 0.0404; 0.0478 > 0.0449). Third, pollutant emission is well-known as a typical activity with negative externalities, especially for the agents located on the borders between neighboring provinces [50, 51]. As emphasized by Wang and Wang [52], in the context of decentralization for governance, the polluting enterprises located near jurisdictional boundaries are likely to be imposed weaker environmental regulations, thus having an incentive to pollute their neighbor, while the host jurisdiction can dispense with internalizing the border pollution spillovers. To mitigate the concern that our results could be misestimated due to environmental externalities, we restrict the sample to enterprises located in the center (not on the border) of provinces. We then conduct the test and report the results in Columns (5) and (6) of Table 5. As expected, the promoting effect of average land prices on an enterprise’s pollution intensity weakens after eliminating the border effect.

Table A1 Robustness checks: alternative standard errors and subsample restrictions.

Variables	Using the DK standard errors	Excluding the enterprises with “zero pollution”	Excluding the enterprises located on the provincial borders
CODI	SOI	CODI	SOI	CODI	SOI
(1)	(2)	(3)	(4)	(5)	(6)
Ave_Price	0.0415∗∗∗ (0.0051)	0.0429∗∗∗ (0.0017)	0.0470∗∗∗ (0.0107)	0.0478∗∗∗ (0.0095)	0.0359∗∗∗ (0.0070)	0.0422∗∗∗ (0.0097)
Observations	12,742	12,742	7276	9253	10,789	10,789
R-squared	0.101	0.163	0.128	0.254	0.126	0.165

Appendix B

Our theoretical mechanism analyses indicate that severe cost burdens will force one enterprise to direct less productive resources to abatement, which makes its pollution intensity rise. The underlying meaning behind this is that the negative effect of rising land prices on a land user’s emission performance is highly dependent on the extent to which its budget constraint is affected. Land transfer in China essentially connects governments, which control the release of new land and land developers. For one enterprise, whether it builds certain political connections with government authorities matters. It is likely that enterprises with stronger political connections find it easier to gain access to land-use rights through a nonmarket-oriented way. We expect that enterprises that obtain preferential treatments from governments typically experience relatively limited negative effects of rising land prices on their emission performances. Two exercises are conducted to manifest our expectation as follows.

Methods to acquire land-use rights: First, compared with land transferring through bidding, auction, and listing, agreement-based assignment of land-use rights belongs to nonmarket resource acquisition strategies among enterprises in China. We therefore investigate whether the effect of increases in land prices on one enterprise’s pollution intensity varies depending on its specific method to acquire land-use rights. Information from the LT database enables us to sort the enterprises into two subsamples distinguished by their methods. Columns (1) and (2) of Table 8, respectively, present the estimated coefficients in terms of CODI in Panel A and SOI in Panel B. As shown, the results of differential effects signify that the promoting effects for those privately negotiating with government authorities are either small or statistically insignificant. On the contrary, given that land prices rise, enterprises using a market-oriented approach are more likely to experience a deterioration of environmental performance. This implies that, in contrast to land transfer by a private agreement with government agents, land transfer by public auctions exacerbates and worsens emission performance led by high land prices.

Whether to obtain political affiliation: Second, the other approach that one enterprise can use to seek access to land-use rights in China is through subordinating local, provincial, and central governments. By transferring part of the power to governments to regulate their business activities, enterprises typically have priority to enjoy favorable treatment in taxation and subsidies [53, 54]. Hence, we attempt to empirically examine whether those who have obtained an affiliation are less impacted by a shock from land prices than those without such connections. With reference to the practices of the existing literature [55, 56], we divide the samples into two groups according to whether or not one enterprise is affiliated with a certain level of government⁶. The coefficients of interest are listed in Columns (3) and (4) of Table 8. The promoting effects on pollution intensity are mainly observed in nonaffiliated groups, whereas the estimates for both dependent variables among politically affiliated enterprises are statistically insignificant. The subgroup estimation results discussed earlier corroborate our inference.

Table A2 Heterogeneity analysis: land transfer methods and political affiliation.

	Methods of land transfer	Political affiliation
Negotiation	Auctions	Affiliated	Nonaffiliated
(1)	(2)	(3)	(4)
Panel A (CODI)
Ave_Price	0.0239∗ (0.0128)	0.0460∗∗∗ (0.0106)	0.0046 (0.0043)	0.0584∗∗∗ (0.0097)
Observations	2488	10,251	1075	11,664
R-squared	0.155	0.157	0.189	0.157
Panel B (SOI)
Ave_Price	−0.0036 (0.0150)	0.0580∗∗∗ (0.0119)	0.0044 (0.0134)	0.0629∗∗∗ (0.0092)
Observations	2488	10,251	1075	11,664
R-squared	0.301	0.212	0.174	0.220

Data Availability Statement

The data that support the findings of this study are available from the corresponding author upon reasonable request.

Ethics Statement

No ethical approval is required for this study.

Disclosure

All authors read and approved the final manuscript.

Conflicts of Interest

The authors declare no conflicts of interest.

Author Contributions

Lin Guo conceived and wrote the original manuscript and provided grant support. Ying Liu and Huzhou Zhu were responsible for manuscript revision and validation. Chunyuan Zhang and Shanna Li were responsible for project management, data collection, and proofreading of the manuscript. Lin Guo, Huzhou Zhu, and Shanna Li contributed equally to this work and shared the first authorship.

Funding

The research was funded by the research project of Zhejiang Chinese Medical University (Grant No. 2024RCZXSK06) and the project of Weifang Science and Technology Bureau “Enterprise Digital Transformation Smooth Green Low-Carbon Big Cycle: Effect and Mechanism” (Grant Number 2023RKX133).

Acknowledgments

The authors thank all the authors for their efforts and the foundation’s support, as well as the editors and reviewers for their dedication.

Endnotes

¹.

²It is widely accepted that the tax-contracting system (cai zheng bao gan zhi in Chinese) initiated in the 1980s is the prelude to fiscal decentralization in China, but some scholars argue that the tax-sharing system in 1994 should be seen as a reform of centralization rather than decentralization according to the flow direction of fiscal revenue [57–59]. With the income tax reform in 2002 and the reform of replacing business tax with VAT in 2016, the central government’s share of tax revenue further increases [12].

³For simplicity, we assume that land is the only productive factor for all enterprises in the model.

⁴Shapiro and Walker [28] note that an equivalent way is to combine output production technology and pollution emission technology above via a Cobb–Douglas production function. That is, we can express output as a Cobb–Douglas function of two factors—pollutant and land q(φ) = (z(φ)^β[φl(φ)]^1−β)/d, where β ∈ (0, 1) represents the share of pollutant as a factor of production.

⁵In 2010, the Ministry of Finance, the Ministry of Land and Resources, the People’s Bank of China, the Supervision Department, and the Auditing Administration jointly issued “Notice on Further Strengthening the Management of Land Transfer Revenue and Expenditure.”

⁶Enterprises in the ASIF database can be classified by their subordinative relationship. The affiliation level from the highest to the lowest includes (1) central authorities; (2) province; (3) city and prefecture; (4) county; (5) subdistrict, town, and village; (6) residents’ and villagers’ committee; and (7) no affiliation [52].

References

1 Xi Q. M. and Mei L., Industrial Land Price, Selection Effect and Industrial Efficiency, Economic Research Journal. (2019) 54, no. 02, 102–118.

2 Huang J. L. and Feng Z. Y., The Effect of Land Cost on Firm’s Exporting Behaviour and Its Mechanism, China Industrial Economics. (2017) no. 09, 100–118, https://doi.org/10.19581/j.cnki.ciejournal.2017.09.006.

3 Helpman E., The Size of Regions, 1998, Topics Pub Econ Theo App Ana, https://books.google.com.hk/books?hl=zh-CN%26lr=%26id=otnyrzLXJfoC%26oi=fnd%26pg=PA33%26dq=The+size+of+regions%26ots=VX5OjV4iDs%26sig=6ankyC9mcu7ub3K6hYH6wmnIJvE%26redir_esc=y#v=onepage%26q=The%20size%20of%20regions%26f=false.

4 Tabuchi T., Urban Agglomeration and Dispersion: A Synthesis of Alonso and Krugman, Journal of Urban Economics. (1998) 44, no. 3, 333–351, https://doi.org/10.1006/juec.1998.2087, 2-s2.0-0012234294.

5 Murata Y. and Thisse J. F., A Simple Model of Economic Geography à la Helpman–Tabuchi, Journal of Urban Economics. (2005) 58, no. 1, 137–155, https://doi.org/10.1016/j.jue.2005.01.002, 2-s2.0-20444490398.

6 Gao B. Y., Liu W. D., and Dunford M., State Land Policy, Land Markets and Geographies of Manufacturing: The Case of Beijing, China, Land Use Policy. (2014) 36, 1–12, https://doi.org/10.1016/j.landusepol.2013.06.007, 2-s2.0-84881228560.

7 Li Y. and He C. F., Characteristics and Mechanism of Manufacturing Industry Shift in the Pearl River Delta During 1998-2009, Progress in Geography. (2013) 32, no. 05, 777–787.

8 Feng Z. Y. and Huang J. L., Does Industrial Land Price Affect Firm Entry: Micro-Evidence From China’s City, South China Journal of Economics. (2018) no. 04, 73–94, https://doi.org/10.19592/j.cnki.scje.351076.

9 Yan H. S. and Sun J. W., Land Price and Corporate Innovation: Evidence From Micro Data, Economic Theory Business Management. (2020) no. 04, 26–38.

10 Dixit A. K. and Stiglitz J. E., Monopolistic Competition and Optimum Product Diversity, The American Economic Review. (1977) 67, no. 3, 297–308, http://www.jstor.org/stable/2117514.

11 Tao R., Lu X., Su F. B. et al., China’s Transition and Development Model Under Evolving Regional Competition Patterns, Economic Research Journal. (2009) 44, no. 07, 21–33.

12 Xi P. H., Liang R. B., and Xie Z. F., Tax Sharing Adjustments, Fiscal Pressure and Industrial Pollution, Japan and the World Economy. (2017) 40, no. 10, 170–192.

13 Zhang K. Z., Wang J., and Cui X. Y., Fiscal Decentralization and Environmental Pollution: From the Perspective of Carbon Emission, China Industrial Economics. (2011) 10, 65–75, https://doi.org/10.19581/j.cnki.ciejournal.2011.10.007.

14 Bai J., Lu J., and Li S., Fiscal Pressure, Tax Competition and Environmental Pollution, Environmental and Resource Economics. (2019) 73, no. 2, 431–447, https://doi.org/10.1007/s10640-018-0269-1, 2-s2.0-85049065075.

15 Cao J. and Mao J., Fiscal Decentralization and Environmental Pollution: A Re-Examination Based on the Dual Perspectives of Internal and External Budget, Chinese Journal of Population, Resources and Environment. (2022) 32, no. 04, 80–90.

16 Wang L. O., Wu H., and Hao Y., How Does China’s Land Finance Affect Its Carbon Emissions?, Structural Change and Economic Dynamics. (2020) 54, 267–281, https://doi.org/10.1016/j.strueco.2020.05.006.

17 Yang X., Wang W., Su X. et al., Analysis of the Influence of Land Finance on Haze Pollution: An Empirical Study Based on 269 Prefecture‐Level Cities in China, Growth and Change. (2023) 54, no. 1, 101–134, https://doi.org/10.1111/grow.12638.

18 Wei H. K. and Wang S. J., Phenomenon Analysis and Theoretical Reflection of China’s Over De-Industrialization, China Industrial Economics. (2019) no. 01, 5–22, https://doi.org/10.19581/j.cnki.ciejournal.2019.01.001.

19 Chen C., Tian W., and Yu M., Outward FDI and Domestic Input Distortions: Evidence From Chinese Firms, Economic Journal. (2019) 129, no. 624, 3025–3057, https://doi.org/10.1093/ej/uez034.

20 Feng Z. Y. and Jia H. Y., Land Cost and Outward FDI From China, World Economics Journal. (2022) no. 06, 35–49+135, https://doi.org/10.13516/j.cnki.wes.2022.06.003.

21 Zhou B. and Du L. S., Land Finance and Rising Real Estate Prices: Theoretical Analysis and Empirical Research, Finance and Trade Economics. (2010) no. 08, 109–116.

22 Kuang W. D. and Li T., Land Sale Pattern, Land Price and Real Estate Price, Journal of Financial Research. (2012) no. 08, 56–69.

23 Wang W. C. and Rong S., Housing Boom and Firm Innovation: Evidence From Industrial Firms in China, China Economic Quarterly. (2014) 13, no. 02, 465–490, https://doi.org/10.13821/j.cnki.ceq.2014.02.009.

24 Lei W. N. and Gong L. T., Housing Price Fluctuation and Social Welfare: A Research Based on Endogenous Firm Entry, Journal of Financial Research. (2016) no. 08, 51–67.

25 Guo J. J., Xian G. M., and Tian S., Can Rising House Prices Boost OFDI by China’s Manufacturing Enterprises?, Japan and the World Economy. (2020) 43, no. 12, 126–150, https://doi.org/10.19985/j.cnki.cassjwe.2020.12.007.

26 Cui J., Lapan H., and Moschini G. C., Are Exporters More Environmentally Friendly Than Non-Exporters? Theory and Evidence, Economics Working Papers (2002-2016). (2012) 12022, https://dr.lib.iastate.edu/handle/20.500.12876/22546.

27 Forslid R., Okubo T., and Ulltveit-Moe K. H., Why Are Firms That Export Cleaner? International Trade, Abatement and Environmental Emissions, Journal of Environmental Economics and Management. (2018) 91, 166–183, https://doi.org/10.1016/j.jeem.2018.07.006, 2-s2.0-85052319929.

28 Shapiro J. S. and Walker R., Why is Pollution From US Manufacturing Declining? The Roles of Environmental Regulation, Productivity, and Trade, The American Economic Review. (2018) 108, no. 12, 3814–3854, https://doi.org/10.1257/aer.20151272, 2-s2.0-85057777208.

29 Lin G., Zhu H., Zhang C., Liu Y., and Yang C., Does Improved Credit Access Promote Pollution Reduction? The Role of Bank Proximity, Economic Analysis and Policy. (2024) 84, no. 84, 78–97, https://doi.org/10.1016/j.eap.2024.08.019.

30 Cui J., Lapan H., and Moschini G. C., Productivity, Export, and Environmental Performance: Air Pollutants in the United States, American Journal of Agricultural Economics. (2016) 98, no. 2, 447–467, https://doi.org/10.1093/ajae/aav066, 2-s2.0-84966348750.

31 Cherniwchan J., Trade Liberalization and the Environment: Evidence From NAFTA and US Manufacturing, Journal of International Economics. (2017) 105, 130–149, https://doi.org/10.1016/j.jinteco.2017.01.005, 2-s2.0-85010452271.

32 He G., Wang S., and Zhang B., Watering Down Environmental Regulation in China, Quarterly Journal of Economics. (2020) 135, no. 4, 2135–2185, https://doi.org/10.1093/qje/qjaa024.

33 Zhao Y., Shen H. T., and Liu Q., Border Pollution Governance in China: Evidence From the Pilot Project of Environmental Protection Supervision Center and Establishment Emission Data, Economic Research Journal. (2021) 56, no. 07, 113–126.

34 Chen F., Wang M., and Pu Z., The Impact of Technological Innovation on Air Pollution: Firm-Level Evidence From China, Technological Forecasting and Social Change. (2022) 177, https://doi.org/10.1016/j.techfore.2022.121521.

35 Lin X., Liu Q. R., and Feng G. M., Intelligent Manufacturing and Green Development: From the Perspective of Chinese Industrial Firms Importing Robots, Japan and the World Economy. (2023) 46, no. 08, 3–31, https://doi.org/10.19985/j.cnki.cassjwe.2023.08.001.

36 Zhu H. Z., Sang B., Zhang C. Y., and Guo L., Have Industrial Robots Improved Pollution Reduction? A Theoretical Approach and Empirical Analysis, China and World Economy. (2023) 31, no. 4, 153–172, https://doi.org/10.1111/cwe.12495.

37 Wang H. and Wheeler D., Equilibrium Pollution and Economic Development in China, Environment and Development Economics. (2003) 8, no. 3, 451–466, https://doi.org/10.1017/S1355770X030024X, 2-s2.0-2542430939.

38 Brock W. A. and Taylor M. S., Economic Growth and the Environment: A Review of Theory and Empirics, Handbook of Economic Growth. (2005) 1, 1749–1821, https://doi.org/10.1016/S1574-0684(05)01028-2, 2-s2.0-66049161681.

39 Kahn M. E., Li P., and Zhao D., Water Pollution Progress at Borders: The Role of Changes in China’s Political Promotion Incentives, American Economic Journal: Economic Policy. (2015) 7, no. 4, 223–242, https://doi.org/10.1257/pol.20130367, 2-s2.0-84946734982.

40 Long W. B. and Hu J., Energy-Saving, and Emission-Reduction Plan, Environmental Assessment and Boundary Pollution, Finance and Trade Economics. (2018) 39, no. 12, 126–141, https://doi.org/10.19795/j.cnki.cn11-1166/f.2018.12.010.

41 Hu G. and Zong J., Has Fiscal Pressure Exacerbated Air Pollution at Boundaries: Empirical Evidence From the Micro Panel Data of Prefecture Level Cities, Journal of Shanxi University of Finance and Economics. (2021) 43, 15–28, https://doi.org/10.13781/j.cnki.1007-9556.2021.10.002.

42 Kong D., Ma G., and Qin N., The Political Economy of Firm Emissions: Evidence From a Quasi-Natural Experiment in China, European Journal of Political Economy. (2022) 75, https://doi.org/10.1016/j.ejpoleco.2022.102181.

43 Feng Z. Y., Huang J. L., and Yan H. R., Pollution Control Pressure, Environmental Incentives and Industrial Land Leasing: Empirical Evidence From Micro Land Transactions, China Economic Quarterly. (2022) 22, no. 06, 2085–2106, https://doi.org/10.13821/j.cnki.ceq.2022.06.13.

44 Nunn N. and Qian N., The Potato’s Contribution to Population and Urbanization: Evidence From a Historical Experiment, Quarterly Journal of Economics. (2011) 126, no. 2, 593–650, https://doi.org/10.1093/qje/qjr009, 2-s2.0-80052922814.

45 Lu Y. and Yu L., Trade Liberalization and Markup Dispersion: Evidence From China’s WTO Accession, American Economic Journal: Applied Economics. (2015) 7, no. 4, 221–253, https://doi.org/10.1257/app.20140350, 2-s2.0-84945191565.

46 Huang W., Zhang Z. Y., and Liu A. R., From Difference-In-Differences to Event Study, Indian Economic Review. (2022) no. 02, 17–36, https://doi.org/10.19313/j.cnki.cn10-1223/f.20211227.002.

47 Jia R., Shao S., and Yang L., High-Speed Rail and CO2 Emissions in Urban China: A Spatial Difference-In-Differences Approach, Energy Economics. (2021) 99, https://doi.org/10.1016/j.eneco.2021.105271.

48 Elhorst J. P., Spatial Econometrics: From Cross-Sectional Data to Spatial Panels, 2014, Springer.

49 Driscoll J. C. and Kraay A. C., Consistent Covariance Matrix Estimation With Spatially Dependent Panel Data, The Review of Economics and Statistics. (1998) 80, no. 4, 549–560, https://doi.org/10.1162/003465398557825.

50 Duvivier C. and Xiong H., Transboundary Pollution in China: A Study of Polluting Firms’ Location Choices in Hebei Province, Environment and Development Economics. (2013) 18, no. 4, 459–483, https://doi.org/10.1017/S1355770X13000168, 2-s2.0-84879803155.

51 Lipscomb M. and Mobarak A. M., Decentralization and Pollution Spillovers: Evidence From the Re-Drawing of County Borders in Brazil, The Review of Economic Studies. (2016) 84, no. 1, 464–502, https://doi.org/10.1093/restud/rdw041, 2-s2.0-85014271614.

52 Wang S. and Wang Z., The Environmental and Economic Consequences of Internalizing Border Spillovers, 2020, The University of Chicago, http://www.sdwang.org/uploads/4/4/8/5/44856715/draft_ws.pdf, Technical Report.

53 Li H., Meng L., Wang Q., and Zhou L. A., Political Connections, Financing and Firm Performance: Evidence From Chinese Private Firms, Journal of Development Economics. (2008) 87, no. 2, 283–299, https://doi.org/10.1016/j.jdeveco.2007.03.001, 2-s2.0-48049119272.

54 Du J., Guariglia A., and Newman A., Do Social Capital Building Strategies Influence the Financing Behavior of Chinese Private Small and Medium–Sized Enterprises?, Entrepreneurship Theory and Practice. (2015) 39, no. 3, 601–631, https://doi.org/10.1111/etap.12051, 2-s2.0-84927913228.

55 Li S., The Puzzle of Firm Performance in China: An Institutional Explanation, Economics of Planning. (2004) 37, no. 1, 47–68, https://doi.org/10.1007/s10644-004-1583-x, 2-s2.0-25844518519.

56 Xia J., Li S., and Long C., The Transformation of Collectively Owned Enterprises and Its Outcomes in China, 2001–05, World Development. (2009) 37, no. 10, 1651–1662, https://doi.org/10.1016/j.worlddev.2009.03.007, 2-s2.0-69349092668.

57 Chen K., Hillman A. L., and Gu Q. Y., Fiscal Re-Centralization and Behavioral Change of Local Governments: From the Helping Hand to the Grabbing Hand, China Economic Quarterly. (2002) no. 04, 111–130, https://doi.org/10.13821/j.cnki.ceq.2002.04.003.

58 Zhou F. Z., A Decade of Tax-Sharing: The System and Its Evolution, Social Sciences in China. (2006) no. 06, 100–115+205.

59 Sun X. L. and Zhou F. Z., Land Finance and Tax-Sharing System: An Empirical Explanation, Social Sciences in China. (2013) no. 04, 40–59+205.