Exploring the Determinants of Residents’ Behavior towards Participating in the Sponge-Style Old Community Renewal of China: Extending the Theory of Planned Behavior

Abstract

In recent years, sponge-style old community renewal (SOCR) has attracted considerable attention as an essential component of urban renewal and sponge city construction in China. As a new method of community governance, the government has increasingly valued residents’ participation in the SOCR. However, as a new concept, the SOCR has not been studied from the perspective of management, and there are few explorations of residents’ participation in the research field. Thus, this study aimed to explore the determinants of residents’ behavior toward participating in the SOCR of China based on the extended theory of planned behavior (TPB), which will further promote residents’ engagement in the SOCR. Data from 1657 respondents were analyzed using the structural equation model (SEM) to verify the direct or indirect relationship between potential variables. The results show that residents’ participation cognition (RPC), attitude (RPA), and intention (RPI) all significantly affect residents’ participation behavior (RPB). The RPC can not only have an indirect impact on the RPB by influencing the RPA but also have an indirect impact on the RPB through the RPA and the RPI. This research not only expands the application boundary of the TPB but also enriches the knowledge system of residents’ participation and the SOCR. Several practical implications for promoting residents’ participation are obtained in the context of the SOCR projects.

1. Introduction

In the process of rapid urban development, stormwater has become a serious problem in China. Water resource depletion and flood events occur from time to time in rapidly developing urban areas [1]. From 1991 to 2020, about two-thirds of the land area was troubled by floods in China, and direct economic losses amounted to an average of 160.4 billion yuan per year [2]. Under the trend of rapid urbanization, to meet the housing needs of the urban population, agricultural land is slowly being occupied by construction land. Thus, the large proportion of impermeable surfaces made of concrete and cement makes built-up land more vulnerable to stormwater than the surrounding environment [3]. The limited capacity to sustainably utilize urban water resources can result in rainwater overflow and bring challenges to urbanization. In addition, low standards of drainage pipes and unsystematic drainage networks make contributions to frequent waterlogging disasters. As a consequence, the driving safety of vehicles is affected, and even the safety of people’s lives and property are threatened [4]. It seems urgent to solve urban waterlogging problems and manage stormwater recyclable in sustainable ways.

Many countries propose various solutions to manage urban stormwater. Ideal practices, such as low impact development (LID) [5], water-sensitive urban design (WSUD) [6], sustainable urban drainage systems (SUDS) [7], and active beautiful clean (ABC) waters program [8] and so forth in different regions of the world, have effectively addressed the problems brought by stormwater on urbanization. The sponge city, proposed by the Chinese government, is a new generation concept of urban stormwater and flood management [9], which means that the city can be like a sponge and has good flexibility in adapting to environmental changes and dealing with natural disasters caused by stormwater. This concept can also be called ‘water elastic cities’ [10]. By 2017, the total investment in all sponge city projects has exceeded 12 billion US dollars [11]. In this context, old communities, which are lacking in public facilities, outdated supporting facilities, poor living conditions, and poor environments, have gradually become of concern to the government and become an important part of the construction of sponge cities [12,13]. However, at the initial stage of the construction of these old communities, the concept of a ‘sponge’ was not integrated, so the community was unable to deal with the problem caused by continuous stormwater. As a result, waterlogging in the community frequently occurs in the rainy season, which severely impacts daily life. To solve this problem, sponge-style old community renewal (SOCR), which combines the sponge city concept with urban renewal, has been formed and is undergoing an upsurge [14]. Consequently, with the promotion of the renewal projects, many communities benefit from it. After the spongy renewal of these communities, the community environment has been significantly improved, and the problems of serious waterlogging, insufficient parking spaces and outdated facilities have been solved [15,16].

Although the government and the construction enterprise play vital roles in the SOCR, residents, as the direct stakeholders of the SOCR, are also the key to the smooth implementation of the renewal project. Contributions could be made to the SOCR projects by residents in various ways. For example, residents in the Xingjucun community of Nanjing City provided advice during the early stages of the renewal projects, which refined the scheme of the SOCR projects and provided a successful experience for related projects [17]. Nevertheless, as a result of the shortage of policy support and ways for participation in the SOCR projects, most residents in old communities are not actively involved in such projects [18]. Residents’ participation attitudes and intentions in the SOCR are not motivated [14]. Moreover, it is reported that community residents in some Chinese cities collectively petitioned against the renewal project [19], and they complained that the SOCR could run the risk of wasteful duplication. For instance, the SOCR projects of the Jianger community in Zhenjiang city of Jiangsu Province were hindered by residents [20]. Thus, residents’ participation is crucial to success in the SOCR.

Meanwhile, increasing attention has been paid to the SOCR by the research community. Quite a lot of research on the SOCR has been conducted from a technological perspective, such as evaluating the effects of runoff in residential communities and using different LID practices to control severe urban rain-flood [21,22]. While the evaluation of the SOCR project implementation in China has not been investigated, there is little consensus on residents’ participation in these projects [9]. As a result, this research field faced challenges. First, the SOCR, as a new concept, has not been studied from the perspective of management [23]. In more detail, as an important part of urban renewal, there are relatively few studies on the residents’ participation behavior in the SOCR. Second, studies on the influencing factors of residents’ participation behavior in the SOCR have rarely been reported. Third, up to now, the relationship between residents’ participation behavior and its determinants in the SOCR projects has not been systematically analyzed based on well-established theory.

To fill these gaps, this paper aims to establish an extended conceptual framework based on the theory of planned behavior (TPB) to explore the determinants of residents’ participation in the SOCR. In particular, we seek (1) to explore the level of the RPC, RPA, RPI, and the RPB in the SOCR; (2) to investigate the mechanism of the RPB in the SOCR; (3) to provide practical implications for promoting residents’ participation in the SOCR of China.

2. Literature Review and Hypotheses

2.1. Theory of Planned Behavior (TPB)

The theory of planned behavior (TPB) was first proposed by Ajzen [24]. It has become one of the most influential theories and is widely adopted to study individual behavior and elaborate on the determinants of individual decision-making in sociology and environment studies [25]. For example, Paul [26] and Yadav [27] used the TPB to predict personal green product purchase and consumption behavior. Chen [28] extended the TPB model to study people’s energy savings and carbon reduction behavior. Considering that the main objective of this paper is to explore the influencing factors that affect individual behavior, the TPB was selected as the suitable theoretical framework for this study. Typically, the TPB focuses on relationships among subjective norms, perceived behavior control, attitude, behavior intention, and behavior [29]. Ajzen [24] believes that behavior intention is the intermediary of all factors that may affect behavior, and it is influenced by subjective norms, perceived behavior control, and attitude. In other words, individual attitude, subjective norms, and perceived behavior control affect behavior intention and then affect the specific behavior taken by individuals [29]. Some scholars believe that subjective norms and perceptual behavior control belong to the cognitive level [30,31]. In the field of educational psychology, individual participation includes three dimensions: cognition participation, emotion participation, and behavior participation, while in some literature, the cognition dimension includes residents’ cognition of relevant knowledge, perceived behavior control, and subjective norms [32,33,34]. Therefore, this study takes residents’ participation cognition (RPC), residents’ participation attitude (RPA), and residents’ participation intention (RPI) as potential variables to explore the determinants of residents’ participation behavior (RPB) in the SOCR [35].

2.2. Hypotheses Based on the Extended TPB

(1)

Residents’ participation cognition (RPC)

The cognition towards a particular behavior refers to consciousness and perception, mainly including cognition of relevant knowledge, perceived behavioral control, and subjective norms [29]. Most scholars have made beneficial explorations on whether related knowledge cognition, subjective norms, and perceived behavior control affect individual participation attitude, behavior intention, and behavior. The research results of Gao et al. show that residents’ perception of recycled water positively impacts residents’ attitudes towards recycled water. In addition, residents’ perception of recycled water also significantly affects residents’ behavior intention and final behavior toward adopting recycled water [36]. Sun-Jung Moon found that perceived behavior control was the most significant determinant, followed by attitude and subjective norms, in exploring the cognitive process of patronage of green restaurants in Korea [37]. Pradhananga et al. studied citizens’ participation behavior in storm flood management and found that self-efficacy and environmental perception influenced citizens’ participation behavior [1]. Thus, the following is hypothesized:

Hypothesis 1 (H1).

The RPC will positively affect the RPA in the SOCR.

Hypothesis 2 (H2).

The RPC will positively affect the RPI in the SOCR.

Hypothesis 3 (H3).

The RPC will positively affect the RPB in the SOCR.

(2)

Residents’ participation attitude (RPA)

The attitude towards a particular behavior refers to how a person makes an optimistic or pessimistic evaluation of a certain behavior [24]. This means that those assessed as having a positive attitude are more likely to participate in the decision-making and construction process of the SOCR than those assessed as having a negative attitude. Ajzen pointed out that attitude will directly affect participation intention and behavior [29]. Flores et al. found that farmers’ attitude towards water resources protection affects farmers’ intention to protect water resources. The attitude towards water resources protection plays an intermediary role between farmers’ awareness of water resources protection and their behavioral intention to protect water resources [38]. The research result of Scherer et al. shows that teachers’ positive attitudes toward digital technology will positively affect their behavior toward using digital technology in teaching [39]. Kim et al. found that the intermediary effect between perception and behavior of public attitude towards nuclear energy is particularly significant [40]. Thus, the following is hypothesized:

Hypothesis 4 (H4).

The RPA will positively affect the RPI in the SOCR.

Hypothesis 5 (H5).

The RPA will positively affect the RPB in the SOCR.

Hypothesis 4a (H4a).

The RPA will play a positive mediating role between the RPC and the RPI.

Hypothesis 5a (H5a).

The RPA will play a positive mediating role between the RPC and the RPB.

(3)

Residents’ participation intention (RPI)

The intention towards a particular behavior refers to an individual’s judgment of the subjective probability of taking a particular behavior, reflecting an individual’s intention to take a particular behavior. Ajzen [29] pointed out that an individual’s intention is a central and motivational factor influencing behavior. In this study, behavior intention indicates how hard residents are willing to participate in the SOCR and how much effort or time they plan to exert. As a general rule, the stronger the intention to engage in a behavior, the more likely should be its performance. Similarly, the more substantially residents intend to participate in the SOCR, the more likely they are to participate in it. For example, Habibah et al. verified the mediating role of psychological budget intention between psychological budget attitude and psychological budget behavior in the research field of psychological budget behavior [41]. Gao et al. found that positive attitude and behavior intention toward recycled water serially mediate the effect of cognition of recycled water reuse on recycled water reuse behavior [36]. Thus, the following is hypothesized:

Hypothesis 6 (H6).

The RPI will positively affect the RPB in the SOCR.

Hypothesis 6a (H6a).

The RPI plays a positive mediating role in the RPC and the RPB.

Hypothesis 6b (H6b).

The RPI plays a positive mediating role in the RPA and the RPB.

Hypothesis 6c (H6c).

The RPA and RPI will serially mediate the effect of the RPC on the RPB.

The conceptual framework of the determinants of the RPB in the SOCR can be established, as shown in Figure 1.

3. Methods

3.1. Questionnaire Design

Considering the characteristics of the SOCR projects, a questionnaire was designed based on the above conceptual framework to gain a better understanding of residents’ participation behavior and its determinants. Specifically, this questionnaire consisted of a brief introduction to the aim of this study, respondents’ demographic and socioeconomic information, indicators of the independent variables (i.e., the RPC, PRA, and RPI), and indicators of the dependent variable (i.e., the RPB). A total of 8 items related to individual information and 19 items corresponding to the indicators of all variables were determined by the translation/back-translation method. Each item was extracted from the mature scales in the literature and slightly modified to adapt to the current research background. The final version of the questionnaire included the following parts (the literature sources of measurement items are given in

Table A1. Indicators and related items.

Variables	Indicators	Items	References
RPC	Understanding techniques used in the SOCR (RPC1)	I understand the low-impact development technology commonly used in the SOCR.	[33,60,61,62,63]
	Knowing the benefits of the SOCR (RPC2)	I think the SOCR brings many benefits, such as improving the quality of life.
	Providing channels for participation (RPC3)	Community or government departments will provide relevant channels for me to participate in the SOCR.
	Knowing ways of participation (RPC4)	I know how to participate in governance in the SOCR.
	Understanding the importance of participation (RPC5)	My participation is essential for the SOCR, such as reducing conflict in the implementation of projects.
	Mastering related knowledge (RPC6)	If I participate in the decision-making, implementation, and maintenance of the SOCR, I can master related knowledge.
	Believing others will do (RPC7)	Family, friends, and neighbors who are important to me will participate in the SOCR.
	Believing others hope we will do (RPC8)	Family, friends, and neighbors who are important to me hope I can participate in the SOCR.
RPA	Supporting the SOCR (RPA1)	I support the SOCR in the community.	[33,64,65]
	Concern regarding the progress (RPA2)	I am concerned about the progress of the SOCR in our community.
	Knowing the importance of the SOCR (RPA3)	I think it is of great significance to implement the SOCR.
	Suggesting increasing channels (RPA4)	I think the government should provide channels for the residents to actively participate in the SOCR.
RPI	Willing to participate in the SOCR (RPI1)	If relevant government departments provide opportunities, I am willing to participate in the whole process of the SOCR, such as participating in the preliminary consultation meeting.	[36,38,61,65]
	Willing to promote the SOCR (RPI2)	I am willing to publicize the benefits of the SOCR to others and encourage others to participate.
	Willing to pay extra fees (RPI3)	To reduce the problem of waterlogging in the community, I am willing to pay extra fees for the SOCR.
	Willing to devote more time and effort (RPI4)	In order to reduce the problem of waterlogging in the community, I am willing to devote more time and effort to participating in issues related to the SOCR.
RPB	Participating in the decision-making stage (RPB1)	I participate in the decision-making stage of the SOCR, such as proposing the SOCR plan to related policymakers.	[18,36]
	Participating in the implementation stage (RPB2)	I participate in the implementation stage of the SOCR, such as collaborating with the construction crew.
	Participating in the maintenance stage (RPB3)	I participate in the maintenance stage of the SOCR, such as regularly cleaning the garbage in the rainwater garden and protecting our rainwater bucket.

in Appendix A):

(a)

A succinct explanation of the SOCR and the purpose of this investigation;

(b)

Respondents’ demographic and socioeconomic information, including gender, age, education, length of residence, living status, employment, tenant, and income;

(c)

The determinants of respondents’ participation behavior, including the RPC (‘understanding techniques used in the SOCR’—RPC1, ‘knowing benefits of the SOCR’—RPC2, ‘providing channels for participation’—RPC3, ‘knowing ways of participation’—RPC4, ‘understanding the importance of participation’—RPC5, ‘mastering related knowledge’—RPC6, ‘believing others will do’—RPC7, ‘believing others hope we will do’—RPC8; 8 questions measured with a 5-point scale), the RPA (‘supporting the SOCR’—RPA1, ‘concerning about the progress’—RPA2, ‘knowing the importance of the SOCR’—RPA3, ‘suggesting to increase channels’—RPA4; 4 questions measured with a 5-point scale), and the RPI (‘willing to participate in decision-making’—RPI1, ‘willing to promote the SOCR’—RPI2, ‘willing to pay extra fees’—RPI3, ‘willing to devote more time and effort’—RPI4; 4 questions measured with a 5-point scale) in the SOCR. Answers were given on 5-point scales from 1  =  low to 5  =  high.

(d)

Respondents’ participation behavior, including respondents’ participation behavior in the decision-making phase, construction phase, and maintenance phase of the SOCR projects (the three items were chosen according to the research of Gu et al. [18], and three questions measured with a 5-point scale were adopted in the questionnaire).

3.2. Sample and Procedure

Five cities in the Yangtze River Delta of China, namely Shanghai, Ningbo, Jiaxing, Zhenjiang, and Chizhou, were selected as investigation cities (Figure 2) for the following reasons. First, these five cities are characterized by rapid urbanization, a shortage of freshwater resources, and severe waterlogging in old communities. Thus, it is crucial to implement the SOCR to address waterlogging issues in such cities. Second, these five cities have been selected for the pilot list of sponge cities, which helps residents of these cities better understand the concept of sponge cities. Third, differences in economic and social levels exist in these cities, which is conducive to comparative analysis. On this basis, 17 communities in these 5 cities were determined as sample collection sites through interviews with 38 experts in the SOCR, including 5 old communities of Shanghai (i.e., the Haishangmingyueyuan community, Hailuhuimingyuan community, Xinluyuan A community, Xinluyuan F community, and the Hailuyuehuayuan community), 2 old communities of Ningbo (i.e., the Yaojianghuayuan community and the Sanhejiayuan community), 3 old communities of Jiaxing (i.e., the Yanyu community, Zhenhe community, and the Lingxiangfang community), 3 old communities of Zhenjiang (i.e., the Sanmaogong community, Jiangbin community, and the Huarunxincun community), and 4 old communities of Chizhou (i.e., the Yijingyuan community, Qingxinjiayuan community, Huijing community, and the Pijiuchang community). The interviewed expert information is shown in

Table A2. Basic information of interviewed experts.

Variables	Items	Percentage (%)
Gender	Male	56.25
	Female	43.75
Education level	Doctor	37.50
	Master	46.88
	Others	15.63
Working experiences	More than 5 years	25.00
	3 to 5 years	28.13
	1 to 3 years	37.50
	Less than 1 year	12.50
Professional title grade	High professional title	21.88
	Associate professional title	34.38
	Intermediate professional title	25.00
	Junior professional title	12.50
	Others	6.24
Profession	College teachers	37.50
	Government staff	25.00
	Enterprise managers	28.13
	Others	9.38

in Appendix A.

In this survey, random sampling was used to ensure that the number of observations from each community was proportional. To determine the sample size necessary for the research, the calculation formula of the sample size of Krejcie and Morgan was applied [42]. It should be emphasized that as the population grows, so does the sample size, which remains relatively stable at slightly more than 380 instances [43]. Hence, taking into account the possibility of multiple deviations in the survey, the statistical sample size of each city was about 400 cases, and the total statistical sample size was about 2000 cases. To ensure the reliability and validity of the sample, face-to-face interviews were performed with respondents in selected communities. From June to August 2021, our research team traveled to all investigation sites and guided the respondents to fill in the questionnaire in the presence of our research team.

A total of 2000 questionnaires were distributed in these old communities, and 1657 valid questionnaires were collected with a valid return rate of 82.85%. The descriptive statistics of the samples are shown in

Table 1. Descriptive statistics of individual characteristics.

Variables	Items	Percentage (%)	Variables	Items	Percentage (%)
Gender	Male	47.37	Living alone or not	Yes	58.06
	Female	52.63		No	41.94
Age	under 20	7.97	Rent or not	Yes	71.27
	20–34	44.42		No	28.73
	35–49	25.29	Working condition	Unemployment	12.37
	50–64	16.11		In employment	80.99
	65 and over	6.22		Retire	6.64
Education level	Primary school or below	5.07	Monthly income	Less than ¥2000	10.26
	Junior high school	29.87		¥2000–¥3999	20.40
	High school	27.76		¥4000–¥5999	25.05
	Junior college	19.67		¥6000–¥8000	25.47
	Bachelor’s degree or above	17.62		More than ¥7999	18.83
Length of residence	Less than or equal to 1 year	31.93	Location	Shanghai	23.60
	2 to 5 years	43.21		Ningbo	18.65
	6 to 10 years	20.76		Jiaxing	18.47
	More than 10 years	4.10		Zhenjiang	18.83
				Chizhou	20.46

. In terms of demographic and socioeconomic characteristics, the gender, age, educational background, length of residence, living status, renting, working condition, monthly income, and location of the respondents were in line with the actual situation of the SOCR, which means that the survey data can be further analyzed.

3.3. Structural Equation Modeling (SEM)

Structural equation modeling (SEM) is a statistical method to verify the effectiveness of the theoretical framework and analyze the relationship between variables based on the covariance matrix of variables [35]. This method is widely used in the fields of psychology, environmental science, pedagogy, and marketing [44]. Considering the inherent logical framework of the residents’ participation behavior in the SOCR belonging to a kind of multiple intermediary model, this study used the SEM to test the relevant hypotheses and analyzed the internal logic of the residents’ participation in the SOCR. In the process of verifying whether the direct effect is supported with the SEM, it is worth specifying that p < 0.001 represents a significant difference, and β greater than 0 represents a positive impact [45]. The maximum likelihood estimate (MLE) method was used to estimate the path coefficients in the internal logic model of the residents’ participation behavior in the SOCR. Meanwhile, mediation effects within the model were tested using the bootstrap method, and the number of the bootstrap samples was set to 1000. Notably, when the bootstrap confidence interval does not contain 0, the mediation effect is proved to exist [36].

4. Results

4.1. Reliability Testing

To establish the internal consistency of the questionnaire, Cronbach’s alpha coefficient and composite reliability coefficient were calculated by conducting a reliability analysis. Values of both coefficients greater than 0.7 are considered a common standard for internal consistency [35]. In this study, SPSS (version 25.0) software was used to perform the reliability analysis. As shown in

Table A3. Results of confirmatory factor analysis.

Variables	Indicators	Cronbach’s Alpha	Composite Reliability	Standardized Factor Loading	AVE
Residents’ participation cognition	RPC1	0.894	0.889	0.743 ***	0.572
	RPC2			0.597 *** (deleted)
	RPC3			0.588 *** (deleted)
	RPC4			0.741 ***
	RPC5			0.754 ***
	RPC6			0.734 ***
	RPC7			0.797 ***
	RPC8			0.757 ***
Residents’ participation attitude	RPA1	0.857	0.858	0.706 ***	0.602
	RPA2			0.782 ***
	RPA3			0.797 ***
	RPA4			0.814 ***
Residents’ participation intention	RPI1	0.858	0.859	0.761 ***	0.603
	RPI2			0.776 ***
	RPI3			0.786 ***
	RPI4			0.783 ***
Residents’ participation behavior	RPB1	0.839	0.839	0.819 ***	0.635
	RPB2			0.775 ***
	RPB3			0.795 ***

Note: significant at *** p < 0.001.

in Appendix A, the Cronbach’s alpha coefficient and composite reliability coefficient for four variables in the current sample were greater than 0.8, indicating that the overall consistency of the variables is good, and the scale meets the reliability requirements.

4.2. Validity Testing

Validity analysis mainly investigates the reliability of the sample results, which usually includes structural validity, convergent validity, and discriminant validity. Structural validity aims to test whether the structure of a multiscale scale adequately reflects the hypothesized dimensions of the structure being measured [46]. As shown in

Table A4. Results of the model fit evaluation.

Fit Indices	Suggested Value	Measured Value
CMIN/DF	$\mathrm{If} 1<{\mathsf{\chi }}^2∕\mathrm{df}$ < 3, the model has a reduced fitting degree; $\mathrm{If} { \mathsf{\chi }}^2/\mathrm{df}>5$, the model needs to be modified [47].	2.412
GFI	If >0.90, the data are ideal [47].	0.988
AGFI	If >0.90, the data are ideal [47].	0.969
RMSEA	If <0.05, the data is ideal; If <0.08, the data are acceptable [47].	0.029
NFI	If >0.90, the data is ideal; If >0.80, the data are acceptable [47].	0.980
TLI(NNFI)	If >0.90, the data are ideal [36].	0.986
CFI	If >0.90, the data are ideal [45].	0.988

in Appendix A, all of the fit indices met the SEM analysis standards, so it can be considered that the fit degree of the model is good. Meanwhile, convergent validity refers to the degree of similarity of measurement results when different measurement methods are used to measure the same construct [47]. Typically, when the standardized factor load of measurement variables on corresponding latent variables is greater than 0.7 and the AVE value of each latent variable is greater than 0.5, it indicates that the scale has good convergent validity [47,48]. As shown in Table A3 in Appendix A, the AVE values of latent variables in the scale were higher than the suggested threshold of 0.5, but the standard load of the RPC2 and the PRC3 were less than the suggested threshold of 0.7. Therefore, the indicators of the RPC2 and the RPC3 were deleted to ensure that the scale has good convergent validity. Moreover, discriminant validity aims to prove that constructs that are not supposed to be related to a preset construct are indeed not related to that construct. It is generally used to test whether there is discrimination between the latent variables of the scale [47]. As shown in

Table A5. Results of discriminant validity analysis.

	AVE	RPC	RPA	RPI	RPB
RPC	0.572	0.756
RPA	0.602	0.263 **	0.776
RPI	0.603	0.152 **	0.536 **	0.777
RPB	0.635	0.369 **	0.418 **	0.411 **	0.797

Note: (1) The square roots of AVE values are the bold elements. (2) Significant at ** p < 0.01.

in Appendix A, the correlation coefficients were less than the square root of AVE, indicating that the four latent variables are differentiated, and the discriminant validity of the scale is very good [49].

4.3. Mean Value of Latent Variables and Ranking of Related Observed Indicators

All of the observed indicators were ranked according to their mean values, with a higher mean value indicating a higher rank [50]. The mean value of four latent variables and the ranking of 17 observed variables in different dimensions are shown in

Table 2. The mean value of latent variables and ranking of observed indicators in different dimensions.

Dimensions	Indicators	Rank	Mean Value of Observed Indicators	Mean Value of Latent Variables
Residents’ participation cognition	RPC1	4	3.82	3.83
	RPC4	1	3.90
	RPC5	6	3.70
	RPC6	3	3.84
	RPC7	5	3.81
	RPC8	2	3.89
Residents’ participation attitude	RPA1	4	3.26	3.46
	RPA2	2	3.52
	RPA3	3	3.48
	RPA4	1	3.57
Residents’ participation intention	RPI1	4	3.49	3.56
	RPI2	3	3.53
	RPI3	1	3.61
	RPI4	2	3.60
Residents’ participation behavior	RPB1	2	3.83	3.79
	RPB2	3	3.66
	RPB3	1	3.88

Note: The RPC2 and the RPC3 were deleted to ensure that the scale has good convergent validity.

.

It is suggested that the mean values of latent variables are different, in descending order: the RPC, the RPB, the RPI, and the RPA. Notably, the RPA has the lowest mean value, and it is expressed at a moderate level (the average value is close to 3). The mean value of the RPI is a little higher than that of the RPA, which is 3.83 and 3.56, respectively. Meanwhile, the mean values for the observed indicators in different dimensions are greater than 3.00 but less than 4.00, indicating that these indicators are higher than the moderate level, but there is still room for improvement. In addition, the RPC5, the RPA1, the RPI1, and the RPB2 are the lowest ranking indicators of their respective dimensions, which should be noted.

4.4. Hypotheses Testing

Amos 25.0 specializes in structural equation path analysis and intermedia analysis, so it was selected to analyze the internal logic and path coefficient of residents’ participation behavior in the SOCR [51]. The results are shown in Figure 3. Corresponding standardized and non-standardized coefficients, standard errors, hypothesis testing results, and other data are shown in

Table 3. Results of the hypotheses testing.

Hypothesis	Path Relationship	Standardized Coefficient	Unstandardized Coefficient	Standard Error	T Value	p Value	Hypothesis Testing
H1	RPA ← RPC	0.261	0.261	0.028	9.171	***	Supported
H2	RPI ← RPC	0.009	0.01	0.028	0.355	0.723	Not supported
H3	RPB ← RPC	0.299	0.363	0.033	11.061	***	Supported
H4	RPI ← RPA	0.534	0.575	0.034	16.788	***	Supported
H5	RPB ← RPA	0.202	0.245	0.04	6.105	***	Supported
H6	RPB ← RPI	0.258	0.291	0.037	7.969	***	Supported

Note: Significant at *** p < 0.001.

. Mediation effects within the model were tested using the bootstrap method, and the number of bootstrap samples was set to 1000. The results are shown in

Table 4. Results of the mediation effect.

Hypothesis	Path	Indirect Effect	Mackinnon PRODCLIN2
			95%CI
			Lower	Upper
H4a	RPC → RPA → RPI (Path1)	0.139	0.1133	0.1898
H5a	RPC → RPA → RPB (Path2)	0.052	0.0671	0.1262
H6a	RPC → RPI → RPB (Path3)	----	−0.0119	0.0206
H6b	RPA → RPI → RPB (Path4)	0.137	0.1173	0.2230
H6c	RPC → RPA → RPI → RPB (Path5)	0.092	0.0618	0.1237

.

In terms of the direct influencing factors of the RPB, it can be found from Table 3 that hypotheses H3 (β = 0.299; p < 0.001), H5 (β = 0.202; p < 0.001) and H6 (β = 0.258; p < 0.001) were supported. This means that the RPC, the RPA, and the RPI in the SOCR all have a significant impact on the RPB. The RPC has the greatest impact on the RPB in the SOCR. The RPI has a greater impact on the RPB in the SOCR than the RPA.

Meanwhile, the direct paths related to the mediation effect also were confirmed. Hypotheses H1 (β = 0.261; p < 0.001) and H4 (β = 0.534; p < 0.001) were proven to be supported in the extended TPB model, whereas H2 (β = 0.009; p > 0.05) was not supported. This indicates that the RPC has a significant impact on the RPA but has little impact on the RPI. On the contrary, the RPA has a great impact on the RPI.

The testing results of mediation effects are shown in Table 4. At the 95% confidence level, the confidence intervals of paths 1, 2, 3, and 5 were (0.1133, 0.1898), (0.0671, 0.1262), (0.1173, 0.2230), and (0.0618, 0.1237), respectively. In this case, 0 was not included in the interval, indicating that the hypotheses H4a, H5a, H6b, and H6c had been verified. By contrast, the confidence interval of path3 includes 0 (−0.0119, 0.0206), and hypothesis H6a was not supported. The above results suggest that the RPA in the SOCR can not only act as a mediator in the RPC–RPI relationship (RPC → RPA → RPI) but also act as a mediator in the RPC-RPB relationship (RPC → RPA → RPB). The RPA–RPB relationship is mediated by the RPI (RPA→ RPI → RPB), but the RPC-RPB relationship cannot be mediated by the RPI (RPC × →RPI→RPB). Additionally, the serial mediators, namely RPA and RPI, mediate the RPC-RPB relationship (RPC → RPA → RPI → RPB).

5. Discussion

5.1. Analysis of the Observed Variables

To measure the dimensions of latent variables, the matching observed indicators were designed and verified in the SOCR projects. It is shown that 17 indicators were evidenced by strong goodness of reliability and validity while 2 indicators (the RPC2 and RPC3) demonstrated a lower degree of convergent validity. Compared with public participation indicators proposed in other research fields [52,53], this study provides a comprehensive list of indicators for measuring the RPB and its influencing factors in the SOCR projects.

As a whole, the mean values of latent variables are different, in descending order: the RPC, the RPB, the RPI, and the RPA. Notably, the RPA has the lowest mean value, and it is expressed at a moderate level (the average value is close to 3), which means that most of the respondents may show an inactive attitude towards participation in the SOCR projects. Since the SOCR in China is still in its infancy, there are deficiencies in the understanding of residents regarding their responsibilities and obligations, resulting in a negative attitude toward participation in the SOCR [1,54]. Additionally, lacking a sense of social belonging caused by historical reasons could also affect their attitude [28,54]. Although the mean value of the RPI is a little higher than that of the RPA, the difference was not substantial. This implies that the residents’ willingness to participate in the SOCR is slightly weak. One potential explanation is that there may be a beneficial impact that contributed to residents’ engagement so willingly [41]. Put differently, if sufficient material and spiritual incentives could be provided for residents in old communities, they may invest more cost, time, and effort in the implementation of the SOCR [54]. Additionally, the lowest rank of the RPC5 in the RPC dimension indicates that the importance of participation is yet to be recognized fully [36]. Residents’ understanding of participation-related affairs lacks a set of standardized operating procedures, and the publicity mechanism is not perfect, leading to the SOCR being poorly understood [54]. Similarly, RPB2 ranked lowest in the RPB dimension, showing that residents are less involved in the implementation stage of the SOCR. Thus, in order to encourage residents to participate in the SOCR, these lower-ranking indicators should receive more attention.

5.2. Direct Paths for Affecting the RPB

The first direct path (H3) was found to be statistically significant at the 0.001 level (p < 0.001) and suggests that the RPC has a significant positive impact on the RPB, and the impact path coefficient is the largest, which has the highest explanatory ability in the extended TPB. Since the RPC dimension of this paper includes residents’ cognition of relevant knowledge, perceived behavior control, and subjective norms, these three factors have a significant positive impact on the RPB. Comparing with the research of Sun-Jung Moon [37] and Gao et al. [36], we found that the cognition of relevant knowledge under the cognitive dimension will also positively affect the RPB. In other words, when residents have a deep understanding of the SOCR (i.e., cognitive improvement), they will know the benefits of sponge renewal and take action to participate in it. Combined with the research of Li et al. [55], the main source of residents’ positive cognition of the SOCR project is the perceived usefulness and perceived enjoyment that the project can bring. This finding further explains the influence of cognition on behavior in the SOCR from the perspective of the TPB.

The second direct path (H5) was found to be statistically significant at the 0.001 level (p < 0.001). Although the RPA has less impact on the RPB than the RPC, its impact on the RPB is still very significant. This finding is consistent with the research results of Scherer et al. [39], and the applicability of the TPB is further supported [29]. In many cases, when residents have a strong positive attitude toward renewal projects, they will take proactive actions to participate in them. As for the SOCR projects, the RPA may be affected by various preconditions, such as the cognition of participation, previous participation experience, and desired effects of participation in the SOCR [56]. In a word, comprehension difficulty recruits a negative attitude. Thus, an in-depth understanding of the SOCR is essential for improving residents’ attitudes and even guiding their participation behavior [1,57].

The third direct path (H6) was found to be statistically significant at the 0.001 level (p < 0.001), which shows that the RPI has a significant positive impact on the RPB. This result is completely consistent with the original hypothesis in the TPB proposed by Ajzen [24]. According to the studies of Gao et al. and Habibah et al. [36,41], the main manifestations of the RPI include residents’ willingness to participate in the SOCR, willingness to promote the SOCR, and willingness to pay costs, time, and effort to implement the SOCR. This means that actions could be taken by residents to participate in the renewal projects if they have a strong willingness and sufficient capital resources to be involved in the SOCR.

5.3. Indirect Paths for Affecting the RPB

The indirect path1 (H4a) is supported by the survey data, suggesting that the RPA acts as a full mediator in the RPC-RPI relationship with the assistance of an insignificant effect of the RPC on the RPI. Meanwhile, the indirect path2 (H5a) indicates that the RPA partially mediates the RPC-RPB relationship because there is a significant effect of the RPC on the RPB. However, the indirect path3 (H6a) cannot be supported, suggesting that the RPI does not mediate the RPC-RPB relationship. That is, the diminished effect exists when the effect of the RPC on the RPB is transmitted through the RPI in the SOCR projects. As for the indirect path4 (H6b), it is shown that the RPI partially mediates the RPA-RPB relationship with the significant effect of the RPA on the RPB. Moreover, the indirect path5 (H6c) can also be supported, indicating that the RPA and the RPI serially and partially mediate the RPC-RPB relationship. In other words, the impact of the RPC on the RPB could be transmitted through the combination of the RPA and the RPI in the SOCR projects.

Specifically, the RPC in the SOCR can not only have an indirect impact on the RPB by influencing the RPA (RPC → RPA → RPB) but also have an indirect impact on the RPB through the RPA and the RPI (RPC → RPA → RPI → RPB). Consistent with the findings of Kim [40] and Habibah et al. [41], this finding shows that residents with a positive attitude towards the SOCR usually have characteristics with a better understanding of techniques used in the SOCR, more insight into methods of participation, more awareness of the importance of participation, substantial knowledge about the SOCR, and sufficient trust in others and themselves. This positive attitude directly encourages them to take action to participate in the whole process of the SOCR. Meanwhile, the probability for residents to engage in the SOCR projects increases with the strengthening of the RPI [47]. Thus, the improvement of residents’ cognition of the SOCR is a necessary precondition for promoting residents’ participation in these projects.

5.4. Limitations

However, two limitations need to be considered regarding the present study. First, a cautious explanation is required for insignificant results since the five pilot cities selected cannot fully represent all cities of China. Second, as the implementation of the SOCR is a dynamic process, residents’ participation cognition, attitude, and intention will change over time. The longitudinal variables are recommended to be analyzed with more applicable and feasible models.

6. Practical Implications

According to the results of Table 3, there are four supported indirect paths (H4a, H5a, H6b, and H6c). Through them, the RPC contributes to the RPB in the SOCR projects. Strategies that enhance the RPC inevitably induce developments in the RPA and RPI, which can further promote the RPB in the SOCR projects. Hence, several practical implications can be obtained according to these findings.

(1)

Strengthening the residents’ participation cognition in the SOCR projects. First, regular publicity of the SOCR can enable residents to fully understand the practical changes brought about by such projects so as to gradually improve their participation cognition. In the publicity process of the SOCR project, community management departments, neighborhood committees, and community service centers are encouraged to provide more education and training opportunities for residents so that they can thoroughly understand the construction of their sponge city and regard their participation in renewal activities as their obligations. Second, considering the characteristics of later maintenance of sponge facilities, the residents could be provided with the necessary facilities maintenance guidance to improve their ability to participate in these projects as well as enhance their cognition of participating in public affairs. Finally, communication platforms in old communities could build a bridge between community residents and the public sector [55]. Therefore, it is suggested to establish a feasible communication platform and host regular community dialogue meetings, such as community forums and community roundtables. Through continuous communication between all stakeholders in the community, it would be possible for residents to keep an in-depth understanding of the SOCR projects. In this case, their cognition of participation could be improved.

(2)

Enhancing the residents’ participation attitude towards the SOCR projects. First, defining the obligations and responsibilities of all participants in the life cycle of the SOCR project, especially emphasizing the participation obligations of residents, could help residents understand that it is their duty to participate in the SOCR so as to fundamentally improve their participation attitude. Second, residents could be cultivated with a sense of community belonging so that they can emotionally endorse the SOCR and consciously participate in the renewal projects. It is proposed that the community environment could be improved by increasing investment in infrastructure to meet the diverse needs of residents and increase the level of residents’ satisfaction with the community. Moreover, community committees can provide opportunities to enhance mutual understanding and trust by organizing regular community activities. Finally, learning from the multi-party cooperation system established in the rainwater management project of Australia [57], residents should be given more rights to engage in the SOCR cooperation system composed of government, local enterprise, community organizations, and residents.

(3)

Improving the residents’ willingness to participate in the SOCR projects. On the one hand, material incentives are the most direct way to activate enthusiasm for residents’ participation. It is suggested to propose targeted incentive measures to meet the specific needs of residents for the SOCR project, such as, for example, providing more activity spaces and parking spaces to address residents’ demand for community space. Furthermore, by learning from the experience of rainwater utilization in Kronsberg of Germany, the rainwater collected by sponge facilities could be used for free for residents to wash their cars or water plants. Ulteriorly, subsidies are suggested to be provided for residents who participate in the opinion exchange meetings on the SOCR in the long term. On the other hand, channels that facilitate residents’ participation could be provided by the government. The ‘unintentional participation’ of residents in the SOCR project is usually caused by the fact that residents do not know the channels of participation. In the rainstorm management project of the United States, the government uses e-mail as a common way of communication, while Australia uses telephone consultation [58,59]. Learning from this case, it is suggested that more convenient online communication methods could be developed, such as online complaint portals, online discussion boards, and online forums.

7. Conclusions

Using the large survey data from five cities in the Yangtze River Delta, this study, for the first time, develops a conceptual framework of the determinants of the RPB in the SOCR based on the extended TPB and then explores the roles of the RPC, the RPA, and the RPI in influencing the RPB. There are several key findings in our research. First, the RPA has the lowest mean value of the three independent variables, so improvement may be especially important in the RPA compared to other variables in the SOCR projects. Furthermore, those lower ranking indicators of the RPC (e.g., RPC5), the RPI (e.g., RPI1), and the PRB (e.g., RPB2) should be given more attention. Second, the RPC, the RPA, and the RPI all have direct positive effects on the RPB. Third, the RPA and the RPI serially mediate the RPC-RPB relationship. Particularly, the RPA partially mediates the RPC-RPB relationship, while the RPI does not mediate this relationship. These findings not only expand the application boundary of the TPB but also enrich the knowledge system of residents’ participation and the SOCR. Further, more extended models could be developed based on this conceptual framework for residents in other countries in terms of their participation in urban renewal projects. In addition, inspired by this study, policymakers could be better equipped to understand the determinants of residents’ participation behavior and make appropriate decisions that could promote residents’ participation in the renewal projects. Further research will focus on the residents’ participation behavior and its determinants in the SOCR projects in other cities and analyze the mechanisms using longitudinal data when they become available. Additionally, these practical implications proposed by this research will be validated and evaluated in the follow-up study.