Study design and hypotheses
The encouragement of the Chinese central government to local governments to adopt the PPP model through the top-down procedure and build and operate WTI has created favourable external policy circumstances for the development of wastewater treatment PPP projects. However, the acceptance of the PPP model by both local governments and private capital is rooted in the positive effect of it on improving the UWTE. In China there is no completely private-owned WTI before. Compared to the original government monopoly on the construction and operation of the WTI, the introduction of private capital participation is equipped with conditions to improve the UWTE. The participation of private capital can donate sufficient funds, scientific management experience, and advanced technology to the construction and operation of the regional, quasi-natural monopoly, and public welfare WTI21, which are key elements that determine the UWTE. Furthermore, the urban wastewater treatment field was in a state of no market competition before the introduction of private capital, and the government’s early monopoly ensured that private capital could obtain both economic benefits and performance with exclusive agency rights after joining. Meanwhile, the government would conduct a performance assessment of the quality of wastewater treatment during construction and operation, and private capitals whose wastewater treatment efficiency failed to meet the requirements would be barred from obtaining performance benefits40. Therefore, private capital is inherently incentivised to ensure the UWTE and minimise profit loss. Most of the private capital involved in the construction and operation of WTI in China comes from state-owned enterprises, partly due to the remarkable cooperation between the local government and state-owned enterprises at the beginning of the market economy reform41. This is convenient for both sides in reducing the cost of supervising due to information asymmetry in the principal-agent relationship and to facilitate the unique advantages of state-owned capital to undertake social responsibility. Therefore, Hypothesis 1 is proposed:
UWTE is high in prefecture-level cities that have introduced the PPP model compared to prefecture-level cities that have not adopted the PPP model for the construction and operation of WTI.
The return mechanism is related to the risk sharing of the costs of WTI construction and operation. Part of the purpose of introducing private capital is to share the cost risk of the government’s monopoly on the construction and operation of infrastructure by exchanging the government’s appropriate concession of operating revenue42; however, excessive cost and risk sharing reduces the probability of private capital participation in the construction and operation of infrastructure43. In China, the return mechanisms of the public–private WTI include user payment, government payment, and feasibility gap subsidy, of which the cost risks of construction and operation under the first two return mechanisms are primarily borne unilaterally by the private capital and the government, respectively44, whereas the cost risks of construction and operation under the latter are borne by the government to fill the gap of user payment39. Accordingly, Hypothesis 2 is proposed:
The return mechanism of the feasibility gap subsidy has a greater impact on improving the UWTE than the mechanisms of user payment and government payment.
The way to choose private capital to cooperate with the government is related to the efficiency of the construction and operation of the WTI. Private capital selected through competitive procurement usually exhibits sufficient funds, scientific management experience, and innovative technology45. Cooperation between the government and this type of private partner helps obtain the optimal construction and operation plan at the lowest cost. The adoption of competitive procurement can improve efficiency while saving transaction costs, especially for infrastructures with large capital scale and long term and complicated operational systems, such as urban wastewater treatment38. In China, the competitive procurement mechanism of PPPs for WTI includes public bidding, competitive negotiation, invitational bidding, and competitive consultation, whereas the non-competitive procurement mechanism mainly refers to single-source procurement46. In accordance with this, Hypothesis 3 is proposed:
The competitive procurement mechanism has a greater impact on improving the UWTE than the single-source procurement mechanism.
The PPP is ultimately a contract between the principal and the agent that specifies how risks are shared and how benefits are distributed40. Construction and operation of WTI under the PPP model usually require long-term contracts. This means that contracts are often incomplete, and the allocation of remaining control rights has a significant impact on the incentives for private capital parties to participate. Existing research suggests that the greater the remaining control the private capital receives, the stronger their incentive to participate in the construction and operation of infrastructure, and the more they pursue innovation and efficiency47. The remaining control right is related to the manner in which the infrastructure is operated48. In China, PPPs for WTI operate through outsourcing (e.g. Operation and Maintenance [OM], Management Contract [MC], and Build-Transfer [BT]), franchising (e.g. Build-Operate-Transfer [BOT], Build-Own-Operate-Transfer [BOOT], Transfer-Operate-Transfer [TOT], and Rehabilitate-Operate-Transfer [ROT]), and privatisation (e.g. Build-Own-Operate [BOO] and Buy-Build-Operate [BBO]). Therefore, Hypothesis 4 is proposed:
Privatised operations have a greater impact on improving UWTE than outsourcing and franchising.
Promotion after demonstration has long been a feature of public policy formulation and implementation by the Chinese government, and this is also true for the construction and operation of wastewater treatment PPP projects. Selecting a portion of these projects for demonstration can facilitate pre-judgement of the issues encountered in the construction and operation of infrastructure and improve efficiency49. The demonstration of WTI is prioritised for various government policies and funding support and is subject to stringent monitoring by the government50. Therefore, to obtain priority support from the government, WTIs that have not entered the demonstration have greater motivation to perform higher quality wastewater treatment. In this case, Hypothesis 5 is proposed:
Wastewater treatment PPP projects that have not yet entered the demonstration have a UWTE higher than those that have been in the demonstration.
Quantifying the UWTE using DEA
In order to measure the efficiency represented by the capacity to increase output at a given input, two methods have been proposed. One is the estimation method based on parameters. The common method is stochastic frontier analysis (SFA). The other is based on the nonparametric estimation method, and the DEA is the most widely used. Although SFA can consider the influence of random factors on output, it needs to determine the specific form of production frontier as the condition when measuring efficiency. This means that if the pre-set production function form is inconsistent with the reality, the efficiency of the measure is not accurate. In contrast, the advantage of DEA is that there is no need to presuppose a specific production function form. It is based on a number of input and output indicators, using the method of linear programming, with the data envelope frontier as the comparison base, the decision making unit (DMU) of the same type of relative evaluation to determine the efficiency. In addition, DEA can also give the improvement space of each DMU in terms of input and output, which is convenient to give optimisation suggestions. Thus, DEA is widely used to assess the efficiency of public services, the environment, and natural resources fields51. With different settings of comparative DMUs, DEA can be divided into the CCR model, which assumes that the comparative DMUs meet the condition of constant returns to scale, and the BCC model, which assumes that the comparative DMUs meet the condition of variable returns to scale, and Shephard distance function introduced to distinguish pure technical efficiency from scale efficiency and determine whether the DMU production is optimal. Most studies have concluded that the BCC model is more consistent with the reality of production52; therefore, it is widely accepted and adopted compared to the CCR model. In this study, DEA based on the BCC model was used to measure the UWTE. The length of the urban wastewater network and the daily treatment capacity of urban wastewater treatment plants are established as input indicators, and the total amount of urban wastewater treatment is established as the output indicator53. The efficiency for each DMU is measured by solving the following linear programming of the BCC model, shown in Eq. (1):
$$begin{array}{l}max theta s.t.mathop {sum }limits_{i = 1}^{283} lambda _i cdot lwn_i le lwn_{i_0} mathop {sum }limits_{i = 1}^{283} lambda _i cdot dtc_i le dtc_{i_0} mathop {sum }limits_{i = 1}^{283} lambda _i cdot tawt_i le theta tawt_{i_0} lambda _i ge 0 mathop {sum }limits_{i = 1}^{283} lambda _i = 1end{array}$$
(1)
where subscript θ denotes the evaluated DMU. lwni and dtci represent the inputs, i.e. length of the wastewater network and the daily treatment capacity of urban wastewater treatment plants in prefecture-level city i, respectively, and the output is tawti, the total amount of wastewater treatment of each prefecture-level city. is a λ vector of intensity variable, and θ represents the efficiency score based on the input-output calculation. This is the UWTE to be calculated in this study.
Causal linking the PPPs to the UWTE using DEA-Tobit regression model
The DEA-Tobit regression model was used to empirically test the causal relationship between the PPPs and the UWTE. It is meaningful to use DEA to measure the UWTE, because the measured relative efficiency can be used to evaluate the capacity of urban wastewater treatment, and make it possible to compare the capacity of urban wastewater treatment between prefecture-level cities, and also creates conditions for finding the factors affecting the UWTE. As the range of UWTE measured by DEA is between 0 and 1, it does not obey the normal distribution and violates the classical assumption of ordinary least squares estimation. Therefore, in order to avoid the bias caused by OLS estimation, the restricted dependent variable model, also known as the Tobit regression model, is usually adopted in previous studies. The regression model which combines DEA and the Tobit regression model is also called the DEA-Tobit regression model. This study employs a DEA-Tobit regression model based on panel data, shown in Eq. (2).
$$uwte_{it} = beta _0 + beta _1 cdot PPP_{it} + X^prime cdot gamma + varepsilon _{it}$$
(2)
where uwte denotes the efficiency of urban wastewater treatment. PPP denotes the degree of development of urban wastewater treatment PPP projects, which is measured in three calibres by determining the presence or absence of wastewater treatment PPP projects, the number of wastewater treatment PPP projects, and the investment amount of wastewater treatment PPP projects. X′ denotes other main control variables that potentially affect UWTE including population density, urbanisation rate, GDP per capita, industrialisation rate, openness, and green innovations. i and t represent prefecture-level city and year, respectively. β0 and εit denote the intercept term and the random disturbance term, respectively. β1 and γ are both parameters to be estimated, and β1 is significantly positive, indicating that the PPP model has a significant positive effect on the UWTE. Because the DEA-Tobit regression model with panel data does not have consistent and unbiased parameter estimates obtained under the fixed effects, the random effects estimation method is used in this study, referring to the parameter estimation recommendations presented by Liu et al.54
Measurements of dependent, explanatory and control variables
The dependent variable in this study is the UWTE. As mentioned above, we use DEA based on the BCC model to measure the UWTE. The closer the value of UWTE is to 1, the higher the efficiency is; the closer it is to 0, the lower the efficiency is.
The degree of PPP development is the key explanatory variable of this study. It can be measured in various ways. The most common approach is determining the presence or absence of PPP projects, the number of PPP projects, and the investment amount of PPP projects31,33. To assess the impact of PPP on the UWTE in a comprehensive and reliable manner, this study uses all three metrics simultaneously.
The endogeneity of mutual causation must be addressed when investigating the causal relationship between PPPs and the UWTE. This is because prefecture-level cities that use PPP models to build and operate WTI may consider wastewater treatment to be important, for example, the promotion of local government officials is closely related to the quality of public services in their jurisdictions during their tenure. To obtain a higher promotion probability, these prefecture-level cities focus on the efficiency of urban public services, including wastewater treatment, and the higher UWTE determines their willingness to adopt PPPs. Therefore, this study uses instrumental variables to eliminate the endogeneity problem in the regression analysis.
Exogenous and correlation conditions are required for suitable instrumental variables. Waste treatment PPP development measured by determining the presence or absence and the number of waste treatment PPP projects is an instrumental variable for the degree of wastewater treatment PPP projects. This is because waste treatment and wastewater treatment are both urban environmental protection infrastructures. Furthermore, prefecture-level cities that consider wastewater treatment are highly likely to consider waste treatment, which are highly correlated. The PPP development for waste treatment does not directly affect the UWTE. Furthermore, the mean number of wastewater PPP projects in neighbouring prefecture-level cities in the prefecture-level city’s province was an instrumental variable for wastewater treatment PPP projects there. This is because, on the one hand, local government officials proactively follow the practices of other neighbouring prefecture-level cities in the province55. Assuming that other neighbouring prefecture-level cities in the province are inclined to promote wastewater treatment PPP projects, the prefecture-level city is highly likely to adopt a PPP model for the construction and operation of WTI. However, the mean number of wastewater treatment PPP projects in other neighbouring prefecture-level cities in the province will not directly affect UWTE in the prefecture-level city.
Control variables: based on IPAT theory56, population density, urbanisation rate, GDP per capita, industrialisation rate, openness, and green innovations were selected in this study to measure the influence of three dimensions of population, wealth, and technology on the UWTE. The population density is measured as the urban population divided by the urban area. The higher the population density, the greater the need for an urban wastewater treatment capacity. The urbanisation rate is calculated as the share of urban population in the total population of the prefecture-level city. The higher the urbanisation rate, the higher the population in urban areas and the higher the demand for urban wastewater treatment capacity. Meanwhile, the urban population produces relatively more wastewater.
GDP per capita is measured as GDP divided by population. The higher the GDP per capita, the higher the level of economic development of the prefecture-level city, and the more the government can regulate urban wastewater35, thus affecting the UWTE. The industrialisation rate is obtained by calculating the ratio of the output value of the secondary industry to GDP. The higher the industrialisation rate, the greater the demand for urban water resources, and more wastewater discharges are generated2, which affects the UWTE. Openness is measured by the proportion of imports and exports to GDP. The higher the openness, the more likely it is to attract companies with advanced environmental technologies57, reducing the amount of wastewater discharged from the prefecture-level city’s production sector. The ‘pollution heaven’ hypothesis may attract additional pollution discharge enterprises to the prefecture-level city58, affecting the prefecture-level city’s UWTE. Green innovations are measured using the number of green patents for wastewater treatment. Green patents for wastewater treatment are obtained from the Green List of International Patent Classification provided by the World Intellectual Property Organization (WIPO). If there are green patents for wastewater treatment, the reduction of wastewater discharge from enterprises is more likely, and thus the UWTE is improved59. This study considers the logarithm of the number of green patents for wastewater treatment to avoid the influence of data heteroscedasticity on the regression estimation results.
Data
The research sample in this study comprised 1303 wastewater treatment PPP projects in 283 prefecture-level cities in China from 2014 to 2019, excluding Hong Kong, Macao, and Taiwan. To estimate the impact of PPPs on the UWTE, we needed data on the length of urban wastewater network, daily treatment capacity of urban wastewater treatment plants, total amount of urban wastewater treatment, wastewater treatment PPP projects, population density, urbanisation rate, GDP per capita, industrialisation rate, openness, and green innovations. Data on the length of the urban wastewater network, the daily treatment capacity of urban wastewater treatment plants, and the total amount of urban wastewater treatment were obtained from the China Urban Construction Statistical Yearbook 2014–201960. The PPP data were obtained from the Ministry of Finance’s Public–Private Partnerships Center61 and were captured by python technology. Data on population density, urbanisation rate, GDP per capita, industrialisation rate, and openness were obtained from China City Statistical Yearbook 2015–202062, and data on green patents were obtained from China National Intellectual Property Administration63. Supplementary Table 1 presents the descriptive statistics of the main variables, and Supplementary Fig. 1 reports the UWTE of 283 prefecture-level cities in China from 2014 to 2019.
Source: Resources - nature.com
