Exploring Public Preference and Willingness to Pay for the Ecosystem Benefits of Urban Green Infrastructure: Evidence from a Discrete Choice Experiment of Pilot Sponge Cities in China

Abstract

Global extreme weather events such as rainstorms, floods and droughts have become increasingly frequent, posing significant threats to human beings worldwide. Green infrastructure has been implemented for decades to mitigate these issues. However, its widespread adoption in cities is still limited. A lack of sustainable funds was viewed as a great challenge for its widespread implementation. Many developed countries have practiced public participation and stormwater fee systems to mitigate these challenges. To assess the value that citizens place on the ecosystem benefits provided by urban green infrastructure is of great importance for increasing public participation and their willingness to pay. Thus, this paper aims to explore public preferences and their willingness to pay for the benefits of green infrastructure to support the construction and maintenance costs. A discrete choice experiment method was employed and five attributes were selected: reduction in run-off pollutant, degrees of ponding, plant type, planting aesthetics and the amount of payment. The Ngene1.2 software was used to generate a questionnaire, and data collected from the case cities, Xianyang and Xi’xian New Area, China, were analyzed using the mixed logit model. The results revealed that in Xianyang, the willingness to pay was not significantly associated with cognition, while in Xi’xian New Area, willingness to pay was positively related to cognition. Respondents exhibited a significant willingness to pay for green infrastructure to achieve reduced degrees of ponding in both cities, as well as improving planting aesthetics in Xi’xian New Area. Based on these findings, we conclude that government decision-makers should prioritize disseminating knowledge about green infrastructure to residents before implementing such projects in communities. The findings provide valuable insights into the potential economic value of urban green infrastructure and can guide policymakers and urban planners in making decisions regarding the design, implementation, and management of green infrastructure. This study contributes to the understanding of citizen perspectives and the economic evaluation of ecosystem services provided by urban green infrastructure.

1. Introduction

The increasing frequency of extreme weather events around the world has led to more heavy rains, floods and droughts [1,2,3,4,5]. These events seriously threaten the safety of human lives and property [6,7,8,9]. Rapid urbanization has resulted in an increase in impervious surfaces, dramatically altering hydrological patterns and flow regimes, and increasing the pressure on urban drainage systems [10,11,12]. Moreover, stormwater runoff from these impervious surfaces conveys a variety of pollutants into nearby streams and rivers, causing not only the degradation of water quality, but also the deterioration of habitats and other significant ecosystems as well as water scarcity [13,14].

To alleviate the pressure of urban drainage systems, save water resources and maintain ecological balance, the concept of green infrastructure (GI) has been introduced and implemented for urban sustainable stormwater management [15,16,17]. In general, GI is defined as the integration of natural and engineered systems at different scales to ensure clean water, preserve ecosystem values and functions, and provide multiple benefits to humans and wildlife [18]. In our study, GI refers to environmental stormwater management facilities at the local scale, which include raingardens, bio-swales, detention and retention ponds, and particularly emphasizes the decentralized microscale control measures at the source. These measures employ techniques such as infiltration, retention, storage and purification to restore the deteriorated hydrological conditions to a nearly pre-development state [19,20,21,22]. Apart from dealing with rainwater, GI can provide humans with a wide range of ecosystem services (e.g., regulating climate, purifying the air, educational and aesthetic value, and entertainment benefits, etc.) [15,23,24,25].

However, it has been reported that the construction and maintenance costs of GI are very high [24]. The lack of long-term construction and maintenance funds has become the main challenge that prevents the long-term functionality and widespread implementation of GI [26,27,28]. In developed countries (e.g., the United States, Germany, Australia and France), governments have established fees charged to residents for accessing these services, which help supplement the high construction and maintenance costs of GI [26,29,30,31]. However, it is worth noting that research conclusions on the standard of fee collection vary among developed countries. For instance, an Australian study found that the public would be willing to pay $5.60/month to prevent flash flooding [32], while another American study found that the public would like to pay $0.50/month to reduce flooding and would like to pay $0.47/month to improve water quality [33]. Interestingly, in certain cities, consumers have shown a willingness to pay higher fees for increased levels of GI. However, it is important to mention that their willingness to pay more is not always accompanied by a clear understanding of the specific benefits they are receiving from GI [18]. In developing countries, the process of urbanization started late, and urban ecological problems emerged late. The awareness of rainwater ecological management was inspired by the experiences of developed countries. For example, China has undergone rapid urbanization since the 1980s [2]. Due to the rapid urbanization and extreme weather, urban flooding has become a major phenomenon in many parts of China [34]. The Chinese central government initiated the sponge city initiative in 2013, which is a holistic water management strategy implemented by using GI [6,27,35]. Moreover, in 2015, 16 cities were selected as the first batch of pilot sponge cities [36].

Similarly, the pilot sponge cities also faced funding problems to implement GI. Learning from developed countries, stormwater fees charged to residents for accessing GI’s services can be a viable funding source to supplement sponge city programs [8,21,24]. Actually, studies have been carried on the aspect of willingness to pay (WTP) for GI construction or implementation in pilot sponge cities. For example, Wang, et al. (2017) examined public perceptions of urban flooding and sponge city construction, as well as the public’s WTP for sponge cities and the influence of demographic and socio-economic factors [2]. Qiao and Randrup (2022) found that two-thirds of all respondents were willing to pay for the GI maintenance in six pilot sponge cities [26]. Similarly, researchers reported that 76% of the respondents agreed to pay for life-cycle maintenance of GI in Shenzhen, Zhenjiang, and Xi’xian New Area [8].

Meanwhile, factors influencing the public WTP were analyzed [24]. It was reported that respondents’ educational level was a significant influencing factor for their WTP in all six cities, but age, gender and family monthly income correlated differently with respondents’ WTP in different cities. Furthermore, some argued that education level affected residents’ WTP by influencing their cognition [37]. Residents with high cognition of green buildings had significantly higher WTP than those who did not live in green buildings [38]. Since cognitive bias may decline with more information disclosure, it is suggested that more disclosure and publicity of green power are necessary to eliminate the cognitive bias, leading the WTP back to the real situation [39]. Thus, the cognitive but not just education should be considered during the estimation of WTP [40]. Similarly, researchers suggested that stormwater treatment did not generate a significant WTP, but other benefits affecting people’s quality of life showed a significant impact on the public WTP for GI [21].

Thus, in this study, we formed two hypotheses: (i) Residents’ cognitive factors, such as the presence of completed GI facilities and increased awareness through disclosure, can influence their WTP for GI; (ii) Residents’ WTP for GI vary between the constructed areas and areas without such construction. As far as we know, there has been no research focusing on the differences of residents’ WTP between the construction areas of GI and the areas without pilot construction of GI. Accordingly, we selected Xianyang without GI constructed and Xi’xian New Area with well-constructed GI as case cities. We hope that the paper can provide valuable insights into the potential economic value of urban GI and can guide policymakers and urban planners in making decisions regarding the design, implementation, and management of GI.

2. Materials and Methods

2.1. Study Area

Xianyang and Xi’xian New Area, China, were selected as the case cities for several reasons (Figure 1). Firstly, they are geographically situated in the southwest of Shaanxi province and bordered on each other, making them suitable for comparative analysis. Secondly, the areas experience a similar average annual precipitation, which is 537 mm. Lastly, GI have been conducted in Xi’xian New Area but not in Xianyang. Fengxi New City, one of the five clusters in Xi’xian New Area, as a pilot nuclear area for sponge city construction, has taken shape after construction. Fengxi New City not only eliminated more than 10 waterlogging prone points, but also the continuous expansion of green area made the average temperature of this area about one degree lower than that of adjacent Xi’an and Xianyang in the same period.

2.2. Methodology and Data Collection

2.2.1. Attributes and Levels

Based on a literature review and expert interviews, we defined five attributes and their corresponding levels of ecosystem benefits. The attributes include reduction in run-off pollutant, degrees of ponding, plant type, planting aesthetics and their cost. Urban pluvial flooding, resulting from localized intense rainfall, leads to ponding and overland flow, causing various socio-environmental impacts. Meanwhile, stormwater carries pollutants that pose a significant environmental threat. The most important function of the GI is to alleviate water accumulation, thereby reducing the likelihood of flooding, while at the same time mitigating stormwater runoff. Thus, the reduction in run-off pollutants and degrees of ponding were selected as two main attributes. GI can also enhance the biodiversity of the community environment to some extent by enriching plant types. This, in turn, contributes to the aesthetic appeal and recreational value of urban spaces. Therefore, plant type and planting aesthetics were selected as the other two main attributes. Lastly, cost is a crucial attribute to investigate residents’ WTP. The specific attribute levels were determined as follows: the reduction in run-off pollutant were set as the low, medium and high levels by 0%, 40%~80%, 80%~100%, respectively [21]. We set level three, level two and level one as the levels of degrees of ponding, and set * to *** as three levels of plant type and planting aesthetics. The cost attribute referred to the additional water bill each family needs to pay per year. Based on the data of the Water Resources Bulletin of Shaanxi province in 2021, we calculated the average annual personal water fee, and in Xianyang, it was 649.98 yuan/year, and in Xi’xian New Area, it was 706.56 yuan/year. Therefore, we set 0%, 5%, 10%, 15%, 20% and 25% of the average annual personal water fee in the two regions as the levels of the cost attribute (

Table 1. Attributes and levels.

Attributes	Levels	Explanation
Reduction in run-off pollutant	0%	Status quo
	40–80%	Reduce leaves and impurities brought by part of the rainwater
	80–100%	Reduce the leaves and impurities brought by most of the rainwater, making the rainwater meet the standard of reclaimed water, and reduce the sewage treatment fee
Degrees of ponding	Level three	Status quo
	Level two	Shoes get wet but there are no splashes when stepping on it
	Level one	Shoes do not get wet while the ground gets wet
Plant type	***	Status quo
	****	One level higher than the status quo
	*****	Two levels higher than the status quo
Planting aesthetics	***	Status quo
	****	One level higher than the status quo
	*****	Two levels higher than the status quo
Cost	0 yuan	Status quo
	5% of water fee	5% of personal annual average water bill
	10% of water fee	10% of personal annual average water bill
	15% of water fee	15% of personal annual average water bill
	20% of water fee	20% of personal annual average water bill
	25% of water fee	25% of personal annual average water bill

Note: *** means status quo, **** means one grade better than the status quo, ***** means two grades better than the status quo.

).

2.2.2. Experimental Design

Excluding the status quo, a total of 485 choices were generated. Each choice set contained the status quo and other two choices (

Table 2. An example of choice set.

Attribute	Status Quo	Option 1	Option 2
Reduction in run-off pollutant	0%	80–100%	40–80%
Degrees of ponding	Level three	Level one	Level three
Plant type	***	***	*****
Planting aesthetics	***	*****	***
Cost	0 yuan	141.31 yuan	141.31 yuan
Your choice:	□	□	□

Note: *** means status quo, ***** means two grades better than the status quo.

). The status quo was assigned the lowest-level utility for all attributes. This was an “opt-out” option that simulated the real-life decision-making of respondents [41]. In total, 234,740 choice sets were generated. We conducted an orthogonal experimental design in Ngene1.2 on the attributes and levels of GI and generated 36 groups of choice sets. Since the orthogonal experimental design generated more choice sets than feasible choice sets, we manually screened all the sets and made the number of occurrences of each level in the remaining choice sets approximately the same to ensure that the probability of each level being selected approximately equal. Following the screening process, 12 groups of choice sets were obtained and they were evenly distributed into four questionnaires. Each questionnaire contained three groups of choice sets.

2.2.3. Questionnaire Design

The questionnaire included three main parts. In the first part, information regarding respondents’ socioeconomic characteristics, e.g., gender, age, education, monthly family income and housing type, were collected. In the second part, the questions about respondents’ knowledge and perceptions of GI were asked. There were three option alternatives: “Familiar”, “Have heard of”, “Don’t know at all”. Respondents would make a multiple choice from the set of alternative photos containing two GI photos (Figure 2) when they chose the “Familiar” option and “Have heard of” option. Pictures c and e in the Figure 2 are the correct options and the questionnaire with other options selected were deemed as invalid. In the third part, surveyors explained attributes and the attribute levels of GI benefits and their descriptions to the respondents (Table 1) and sample selection sets (Table 2), to ensure they understood the characteristics and options available for each attribute. In our study, some GI benefits (e.g., planting aesthetics and plant type) were considered to be cultural services within the ecosystem services. Meanwhile, reduction in run-off pollutant and degrees of ponding were considered to be regulating services. Then, respondents are informed that if they are willing to pay, the fee will be charged in the form of an additional property fee. After that they were asked to choose an option in the three selection sets, namely, “Status quo”, “Option 1” and “Option 2”. After the questionnaire design was completed, volunteers were asked to fill out the questionnaire as a questionnaire pre-test to assess its rationality and effectiveness.

2.2.4. Data Collection

By 2022, the total population of Xianyang is 4.17 million, and the total population of Xi’xian New Area is 1.30 million. We used Scheaffer equation to determine the sample size [2]. The sampling error was set as 0.05 in our study. Therefore, the statistical theoretical sample size in Xianyang was about 205 and Xi’xian New Area was about 98. Combined with the administrative divisions, the effective sample size of Xianyang should be 150, and that of Xi’xian New Area should be 100. The questionnaire survey was conducted on April 2023. During this period, the temperature is relatively comfortable and residents are more willing to engage in outdoor activities, which is beneficial for questionnaire surveys. We selected the residential areas on both sides of the road section which were prone to ponding in Xianyang and Xi’xian New Area to carry out the questionnaire survey. According to news reports, a total of eight roads including Leyu North Road, Xinxing North Road, Dongfeng Road, Renmin East Road, Changling Road, Pioneer Street, Yingbin Avenue and Tianzhang Avenue were selected. Volunteers distributed questionnaires through random sampling by intercepting the adult interviewees, explaining the intention and the purpose of the study to obtain verbal consent of the interviewees.

2.3. Econometric Model

The theoretical basis for DCE was from Lancaster’s Theory of Value [42]. It could be applied to situations where multiple attributes change simultaneously and to calculate the relative value between each attributes, ranking the relative importance of each attribute more scientifically. In our study, we define the utility that respondent n chooses plan j that maximizes its benefit from other plan j as:

$$U_{jn}=Z_{jn}({\beta }_n)+{\epsilon }_{jn}$$

(1)

where Z_jn is the preference and individual socioeconomic factors of respondents that influence their choice, ε_jn is an unobserved error term. We defined the probability density function of ε_jn as f(ε). β_n is the parameter of the degree of preference of individual n for each indicator in plan j, the utility model function is shown as follows:

$$\begin{array}{l}{U}_j=U_j^{\prime }+{\epsilon }_j\hfill \\ =AS{C}_j+{\beta }_1{x}_1+\cdots +{\beta }_k{x}_k+r_1{r}_1+\cdots +r_k{r}_k+{\epsilon }_j\hfill \end{array}$$

(2)

where ASC_j is the constant term corresponding to plan j. The explanatory variables include k key attribute variables x_j and n individual characteristic variables of respondents s, ε_jn is the random disturbance term of zero mean. The cross-term composed of constant term ASC_j and personal characteristic variable s is used as the control variable to analyze the reasons of preference differences among different individuals, and to reduce the estimation error to a certain extent, as:

$$U_j=AS{C}_j+{\beta }_1{x}_1+\cdots +{\beta }_k{x}_k+r_1(AS{C}_j\times s_1)\cdots +r_m(AS{C}_j\times s_m)+{\epsilon }_j$$

(3)

When f(ε) follows the Gumbel extremum distribution, the probability of respondents choosing combination j is expressed as:

$$P_{jn}=P[(Z_{jn}+{\epsilon }_{jn})>(Z_{in}+{\epsilon }_{in}),i,j\in C]$$

(4)

We use a mixed logit model to explain the heterogeneity of respondents’ preferences, respondent n chooses the probability distribution function of plan j that maximizes its utility as follows:

$$P_{jn}={\displaystyle \int \frac{\exp Z_{jn}}{{\displaystyle \sum \exp Z_{in}}}}f(\beta )d\beta$$

(5)

where f(β) is the probability density function of β_jn. For each sample simulation, the 500 Halton extraction method was used for likelihood estimation.

Finally, we measure the marginal willingness to pay (MWTP) of respondents. It is shown in the following equation:

$$MWTP=-\frac{{\beta }_n}{{\beta }_{wtp}}$$

(6)

where β_n is regression parameter for each attribute, β_wtp is payment amount parameter.

3. Results

3.1. Sample Characteristics

A total of 287 respondents completed the survey (

Table 3. Sample characteristics.

Variable	Xianyang	Xi’xian New Area	Total	Percentage
Sample population	146	121	267
Gender (ratio of male to female)	65.91%	89.06%
Age
18–25	42	35	77	28.84%
26–30	10	14	24	8.99%
31–40	23	25	48	17.98%
41–50	24	16	40	14.98%
51–60	26	15	41	15.36%
Over 60	21	16	37	13.86%
Monthly family income
1000 yuan	5	10	15	5.62%
1000–1999 yuan	5	12	17	6.37%
2000–4999 yuan	39	22	61	22.85%
5000–9999 yuan	61	55	116	43.45%
Over 10,000 yuan	36	22	58	21.72%
Education
Primary school and below	11	10	21	7.87%
Junior high school	24	16	40	14.98%
Senior middle school	39	38	77	28.84%
Bachelor	69	50	119	44.57%
Postgraduate	3	7	10	3.75%
Housing type
Owner-occupied	122	96	218	81.65%
Rental	24	25	49	18.35%
Understanding level
Familiar	10	10	20	7.49%
Have heard of	42	33	75	28.09%
Don’t know at all	94	78	172	64.42%

), including 267 valid questionnaires. After data cleaning, a total of 2403 valid choice observations were retained for further analysis. About 29% of respondents (N = 77) were young adults (18–25 years old). The most common monthly family income of residents was “5000–9999 yuan” (N = 116). About 45% of respondents claimed a bachelor’s degree was their highest qualification. Most respondents (64%) were completely unaware of GI’s ecosystem services.

3.2. Results of Basic Model

In the software Stata17, the method of 500 Halton extraction was used to estimate the likelihood of each sample, and then we obtained the basic model regression results and the cross regression results with control variables and constant items.

3.2.1. Regression Results of the Basic Model

In the regression results of the basic model of Xianyang, residents’ WTP for reduction in run-off pollutant was insignificant, however, the degree of ponding showed a significantly positive effect at the 1% level (

Table 4. Regression results of Xianyang.

Attributes	Model I			Model II
	Coefficient		SD	Coefficient		SD
	(SE)	(95% CI)	(SE)	(SE)	(95% CI)	(SE)
ASC	3.7432 ***			3.5325
	(1.2194)	(1.3532, 6.1332)		(5.9240)	(−8.0783, 15.1432)
Cost	−0.0044			−0.0052
	(0.0047)	(−0.0136, 0.0048)		(0.0050)	(−0.1495, 0.0045)
Reduction in run-off pollutant	−0.2753		0.7260 **	−0.3085		0.8336 ***
	(0.1876)	(−0.6431, 0.0924)	(0.3399)	(0.1968)	(−0.6942, 0.0772)	(0.3180)
Degrees of ponding	0.6185 **		0.5225 **	0.6547 ***		0.5902 **
	(0.1482)	(0.3280, 0.9090)	(0.2629)	(0.1585)	(0.3441, 0.9654)	(0.2585)
Plant type	−0.3878 **		0.9251 ***	−0.4193 **		1.0112 ***
	(0.1808)	(−0.7422, −0.0335)	(0.3071)	(0.1904)	(−0.7926, −0.0460)	(0.3107)
Planting aesthetics	−0.0050		0.0676	−0.0007		0.0595
	(0.1376)	(−0.2746, 0.2646)	(0.4238)	(0.1438)	(−0.2826, 0.2812)	(0.3539)
ASC × Gender				5.1656 *		−9.9350 ***
				(2.6704)	(−0.0683, 10.3996)	(2.8239)
ASC × Age				−1.0310 **		1.1117 ***
				(0.4911)	(−1.994, −0.0684)	(0.3429)
ASC × Monthly family income				0.1850		0.2625
				(0.7066)	(−1.2000, 1.5700)	(0.3277)
ASC × Education				1.0995		0.0202
				(1.0231)	(−0.9058, 3.1048)	(0.2773)
ASC × Housing type				1.7843		0.6936
				(2.0058)	(−2.1470, 5.7157)	(0.9188)
ASC × Understanding level				−1.5961		1.0909 **
				(1.1123)	(−3.7762, 0.5840)	(0.4715)
Number of observations	1314			1314
Chi2	195.33			177.90
Log likelihood	−360.02 ***			−344.56 ***

Note: * means significant at 10% level, ** means significant at 5% level, *** means significant at 1% level.

). It indicated that for every improvement of one level in the degree of ponding, the probability that residents WTP for GI’s ecosystem services would increase by 61.85%. The plant type was significantly negative at the 5% level, indicating that with each increase of one level in the plant type, the probability that residents WTP for GI’s ecosystem services would decrease by 38.78%.

In the regression results of the basic model of Xi’xian New Area, residents’ WTP for reduction in run-off pollutant was also insignificant and the degree of ponding was significantly positive at the 1% level (

Table 5. Regression results of Xi’xian New Area.

Attributes	Model I			Model II
	Coefficient		SD	Coefficient		SD
	(SE)	(95% CI)	(SE)	(SE)	(95% CI)	(SE)
ASC	1.0451			10.5644
	(1.0029)	(−0.9206, 3.0107)		(4.9308)	(0.9003, 20.2286)
Cost	−0.0022			−0.0025
	(0.0044)	(−0.0108, 0.0064)		(0.0044)	(−0.0111, 0.0062)
Reduction in run-off pollutant	0.1803		0.0207	0.1669		−0.0483
	(0.1763)	(−0.1653, 0.5258)	(0.4247)	(0.1722)	(−0.1675, 0.5073)	(0.5668)
Degrees of ponding	0.4920 ***		0.5732 **	0.4863 ***		0.4828 **
	(0.1424)	(0.2129, 0.7710)	(0.2395)	(0.1365)	(0.2187, 0.7538)	(0.2407)
Plant type	−0.1123		0.9086 ***	−0.0956		0.8537 ***
	(0.1927)	(−0.4900, 0.2654)	(0.3221)	(0.1849)	(−0.4581, 0.2669)	(0.3023)
Planting aesthetics	0.2692 **		−0.0727	0.2557 *		0.0597
	(0.1619)	(−0.0482, 0.5866)	(0.4429)	(0.1577)	(−0.0534, 0.5649)	(0.3021)
ASC × Gender				−2.3637 *		0.3663
				(1.3495)	(−5.0086, 0.2812)	(1.4813)
ASC × Age				0.4277		2.5550 ***
				(0, 4713)	(−0.4961, 1.3515)	(0.7107)
ASC × Monthly family income				0.1534		0.0207
				(0.5096)	(−0.8454, 1.1522)	(0.2104)
ASC × Education				0.3172		−0.0256
				(1.2523)	(−2.1373, 2.7716)	(0.2297)
ASC × Housing type				−3.9394 **		−0.0530
				(1.7776)	(−7.4234, −0.4554)	(0.7155)
ASC × Understanding level				−2.4101 **		0.8684 **
				(1.0495)	(−4.4671, −0.3530)	(0.3526)
Number of observations	1089			1089
Chi2	171.13			150.18
Log likelihood	−301.97 ***			−284.17 ***

Note: * means significant at 10% level, ** means significant at 5% level, *** means significant at 1% level.

). This suggested that for every improvement of one level in the degree of ponding, the probability that residents WTP for GI’s ecosystem services would increase by 49.20%. The plant type was significantly positive at the 5% level. It indicated that for each level of plant type increase, the probability that residents WTP for GI’s ecosystem services would increase by 26.92%.

3.2.2. Results of Cross-Regression

Taking the basic characteristics of the respondents as control variables can better explain the heterogeneity of their preferences and WTP. Based on the cross-analysis results of Xianyang (

Table 6. Cross regression results of Xianyang.

Variable	Coefficient Mean	Z-Value	Standard Deviation
Additional annual fee required	−0.0052 (0.0050)	−1.05
Reduction in run-off pollutant	−0.3085 (0.1968)	−1.57	0.8336 *** (0.3180)
Degrees of ponding	0.6547 *** (0.1585)	4.13	0.5902 ** (0.2585)
Plant type	−0.4193 ** (0.1904)	−2.20	1.0112 *** (0.3107)
Planting aesthetics	−0.0007 (0.1438)	0.00	0.0595 (0.3539)
Constant term	3.5325 (5.9240)	0.60	1.4683 *** (0.8913)
Constant term × Gender	5.1656 * (2.6704)	1.93	−9.9350 *** (2.8239)
Constant term × Age	−1.0310 ** (0.4911)	−2.10	1.1117 (0.3429)
Constant term × Monthly family income	0.1850 (0.7066)	0.26	0.2625 (0.3277)
Constant term × Education	1.0995 (1.0231)	1.07	0.0202 (0.2773)
Constant term × Housing type	1.7843 (2.0058)	0.89	0.6936 (0.9188)
Constant term × Understanding level	−1.5961 (1.1123)	−1.43	1.0909 *** (0.4715)
Number of observations	1314
Chi2	177.90
Log likelihood	−344.56 ***

Note: * means significant at 10% level, ** means significant at 5% level, *** means significant at 1% level.

), gender was significantly positive at the 10% level, and it means that male residents in Xianyang were more willing to pay for the improvement of GI’s ecosystem services. Age was significantly negative at the 5% level, indicating that the older residents in Xianyang had lower WTP.

In Xi’xian New Area, gender was significantly negative at the 10% level, indicating that female residents of Xi’xian New Area were more willing to pay for the improvement of GI’s ecosystem services (

Table 7. Cross regression results of Xi’xian New Area.

Variable	Coefficient Mean	Z-Value	Standard Deviation
Additional annual fee required	−0.0025 (0.0044)	−0.56
Reduction in run-off pollutant	0.1699 (0.1722)	0.99	−0.0483 (0.5668)
Degrees of ponding	0.4863 *** (0.1365)	3.56	0.4828 ** (0.2407)
Plant type	−0.0956 (0.1849)	−0.52	0.8537 *** (0.3023)
Planting aesthetics	0.2557 * (0.1577)	1.62	0.0597 (0.3021)
Constant term	10.5644 (4.9308)	2.14	−2.6374 ** (1.2067)
Constant term × Gender	−2.3637 * (1.3495)	−1.75	0.3663 (1.4813)
Constant term × Age	0.4277 (0.4713)	0.91	2.5550 (0.7107)
Constant term × Monthly family income	0.1534 (0.5096)	0.30	0.0207 *** (0.2104)
Constant term × Education	0.3172 (1.2523)	0.25	−0.0256 (0.2297)
Constant term × Housing type	−3.9394 ** (1.7776)	−2.22	−0.0530 (0.7155)
Constant term × Understanding level	−2.4101 ** (1.0495)	−2.30	0.8684 ** (0.3526)
Number of observations	1089
Chi2	150.18
Log likelihood	−284.17 ***

Note: * means significant at 10% level, ** means significant at 5% level, *** means significant at 1% level.

). The housing type was significantly negative at the 5% level, indicating that residents renting houses in the community were unwilling to pay for the improvement of GI’s ecosystem services. The understanding level was significantly negative at the 5% level, and it suggests that residents who were less aware of GI’s ecosystems were less willing to pay.

Residents in Xianyang have little contact with green infrastructure, and there is no correlation between their WTP and their degree of awareness. While residents of Xi’xian New Area can learn GI through government propaganda and other channels, their WTP is significantly related to their understanding level.

3.2.3. Marginal Willingness to Pay

Based on the significant attributes in the regression results of the basic model, the marginal WTP of residents in Xianyang and Xi’xian New Area was obtained, as shown in

Table 8. Marginal willingness to pay of residents.

Area	Attributes	MWTP [Yuan/(Family·Year)]	Confidence Interval
Xianyang	Degrees of ponding	139.8469	74.1564	205.5374
	Plant type	−87.6946	−167.8135	−7.5757
Xi’xian New Area	Degrees of ponding	197.8169	88.9740	306.6597
	Planting aesthetics	104.0322	−21.7299	229.7944

. Residents in Xianyang were willing to pay for the improvement of degrees of ponding, but not for the plant type. Moreover, the additional annual fee they were willing to pay for was 139.8469 yuan per family per year. For residents in Xi’xian New Area, they had stronger WTP for the improvement of degrees of ponding than that in Xianyang, with the additional annual fee of 197.8169 yuan per family per year. Meanwhile, residents in Xi’xian New Area were willing to pay for the plant type improvement, which was quite different from the residents in Xianyang.

4. Discussion

4.1. Public Preferences on GI Facilities’ Functions

4.1.1. Reduction in Run-Off Pollutant

Residents’ WTP for reduction in run-off pollutant was insignificant in both case cities, and this was inconsistent with some pervious findings below. A study showed that citizens valued the improvement of water quality and were willing to pay $40 per year to avoid further degradation of the stream [43]. In Italy, evidence was given to confirm that the public was willing to pay for environmental protection and the importance of pollutant industries in the regional economy [44]. It is worth noting that peoples’ choices about pollution may depend more on their subjective perceptions of pollution levels [45]. In our study, the public might not realize that GI can reduce the pollutant of runoff. Furthermore, they did not understand what benefits that this function can bring to them. It has been reported that the public was reluctant to pay for those functions that they were not familiar with [46]. Since most respondents may not have sufficient educational background on relevant environmental issues, residents’ awareness of GI ecosystem benefits was neither objective nor scientific. The majority of people even perceived cultural services (e.g., planting aesthetics, educational value, and entertainment benefits, etc.) as the most important services [47].

4.1.2. Degrees of Ponding

Regarding the regulation function provided by GI, residents of Xianyang and Xi’xian New Area preferred to change the current degree of ponding and were willing to pay about 139.8 yuan/family/year and 197.8 yuan/family/year, respectively. The findings align with previous studies. For example, a case study in Japan indicated that respondents had a high interest and were willing to pay for improvements in rainwater harvesting and rainwater infiltration system, which could be attributed to the intuitive and easily comprehensible benefits of installing rainwater tanks [48]. In our study, residents’ intuitive perception of the degree of ponding was influenced by walking through on water splashes on the road surface, and which had a noticeable and intuitive impact. In addition, ref. [49] found that respondents were still willing to pay more for flood protections than for any secondary benefits in some areas where flooding was common. Ref. [21] found that residents were willing to contribute to environmental protection through different involvement activities. These findings were consistent with studies that assessed residents’ WTP for stormwater management. It is essential to enhance public awareness that GI facilities are an important local flood control management strategy to adapt to continuous climate change. Reinforcing the promotion of GI facilities’ rainwater drainage capabilities and building public confidence in GI can prove to be effective approaches in increasing the acceptance and popularity of GI among residents.

4.1.3. Plant Type and Planting Aesthetics

Residents of Xianyang believed that it was not necessary to pay for the increasing plant species in the community. Similarly, existing studies have found that ecological services were often overlooked by some residents [50,51]. People were more likely to perceive their immediate needs, such as aesthetic pleasure and entertainment. Residents of Xianyang may be more concerned about the practicability of GI.

Results showed that planting aesthetics affected the WTP of residents in Xi’xian New Area, which aligns with previous studies. For example, ref. [52] found that residents of Hangzhou in China were willing to pay an average of $14.28 per person for the improvement of planting aesthetics, while tourists of Hangzhou were willing to pay an average of $16.18 per person for the improvement of planting aesthetics. Additionally, ref. [53] discovered that in Iowa, farm operators were willing to pay $3.09 for the improvement of aesthetics associated with field windbreaks and non-farmers were willing to pay $5.90 for the improvement of aesthetics associated with field windbreaks. People were more likely to perceive their immediate needs, such as aesthetic pleasure and entertainment [48]; however, there was insignificant WTP of planting aesthetics for residents of Xianyang. This may be because the government had failed to propagate the benefits of planting aesthetics brought by GI in Xianyang.

4.2. Influencing Factors of WTP

4.2.1. Cognitive Level

As a national sponge city construction demonstration site, residents of Xi’xian New Area have more opportunities to deepen their cognition of GI through government publicity or self-understanding, so the cognitive differences are strongly related to WTP. However, Xianyang did not carry out the construction and publicity of sponge city, so the residents’ cognition of GI was relatively shallow, resulting in an insignificant correlation between cognitive differences and WTP. The results showed that there was no correlation between the WTP and cognition of residents of Xianyang. However, in Xi’xian New Area, the higher the cognitive level, the higher the residents’ WTP for GI facilities. Interestingly, the education level of residents in both case areas did not affect their WTP. The findings align with a previous study that the education level affected residents’ WTP indirectly through influencing their cognition [37]. The findings are further supported by [54], where it was found that people who had heard about self-driving vehicles had a higher WTP for the technology, placed more trust in the technology, and perceived higher benefits. Similarly, ref. [55] found that respondents who were familiar with recycling status, recycling laws and policies would have higher WTP for waste mobile phone recycling. Overall, residents’ WTP was more influenced by their cognitive level rather than education level. Thus, popularizing GI knowledge to residents is of great importance for achieving public support for GI implementation.

4.2.2. Gender

In our study, female residents in Xi’xian New Area had a higher WTP for improving benefits that GI provided, while male residents in Xianyang had a higher WTP. This was similar to previous studies where there were significant differences in the impact of gender on WTP in various contexts. Ref. [56] implied that females had higher WTP for environmentally-friendly products and they were more willing to participate in the design of environmental policies compared to males [57]. However, ref. [58] found that, in typical households, males have a higher WTP for clean heating. In addition, a contingent valuation study on sustainable drainage systems in Hong Kong showed that gender had no significant difference on WTP [59]. Moreover, we speculated that the reason for this discrepancy was that we did not collect the questionnaire with a 1:1 ratio of male to female, and female respondents of Xi’xian New Area accounted for a larger proportion. Thus, our findings are still consistent, or somewhat limited, with long-standing findings on gender differences.

5. Conclusions

In this paper, residents’ preferences and their WTP for functions that GI provided were studied by using a DCE method. It was found that the residents of Xianyang and Xi’xian New Area, China had positive attitudes towards the degree of ponding and were willing to pay for it. It is worth noting that there was a heterogeneity preference and WTP regarding functions of plant types and planting aesthetics. In Xianyang, residents were willing to pay approximately 139 yuan per family per year for the improvement of degrees of ponding, but they did not show the same willingness to pay for plant types. On the other hand, in Xi’xian New Area, residents were willing to pay about 197 yuan per family per year to improve degrees of ponding and 104 yuan for planting aesthetics. These findings suggest that the degree of support for various GI projects varied among the two regions and was influenced by factors such as the extent to which the government promoted GI facilities and residents’ subjective perceptions of the benefits. Although improving degree of ponding has always been the top priority of government officials, improving planting aesthetics in the community and other GI facilities projects may gain higher levels of public support. Additionally, the differences in residents’ willingness to pay between the two areas underscore the importance of conducting further research in this field.

The support of residents is crucial for making decisions regarding the design, implementation, and management of GI. Meanwhile, due to residents’ WTP being commonly influenced by their cognitive level rather than the education level, further efforts are needed to raise residents’ cognition of the function of GI. Therefore, government decision-makers should firstly popularize GI knowledge to residents before building GI in the community in order to obtain paid support from them. Moreover, the knowledge of rainwater utilization can also be incorporated into school education to raise cognition of the younger generations. This study contributes to the understanding of citizen perspectives and the economic evaluation of ecosystem services provided by urban green infrastructure.