Supporter Profiling in Recycled Water Reuse: Evidence from Meta-Analysis

Abstract

Recycled water is considered as a viable alternative water source, and its use is of great significance in dealing with the shortage of water resources. However, it is often rejected by the public during its promotion. To identity the common social attributes of proponents of recycled water, it is essential to study the profiled customers’ willingness to accept recycled water during the most difficult initial stage of promoting recycled water use projects. A meta-analysis was conducted in this study to deal with data concerning the influence of different social demographic factors on the public’s acceptance of recycled water use. Three steps of meta-analysis were mainly used in this paper. First, a random-effect model was used to measure the effect size of influencing factors, such as age, gender, and education. It was found that younger women were more likely to accept recycled water use than older men, and individuals with higher education were more willing to accept recycled water. Then, a cumulative meta-analysis showed that it was feasible to select young women with higher education as target profile customers in the initial stage of recycled water promotion. Finally, according to a meta-regression analysis, it was revealed that different research areas and selected model methods have important regulatory effects on the intention of the target population to use recycled water.

1. Introduction

With the continuous development of the social economy, rapid population growth, and people’s continuous pursuit of quality of life, the consumption of natural water resources by human activities has increased significantly. Since the beginning of the 21st century, with the acceleration of urbanization, the gap between the supply and demand of water resources has been an issue of global concern [1]. Statistically, by 2050 there will be 1 billion people living in cities where water is scarce year-round [2]. Human activities have become the most important factors affecting the natural water circulation system, which has exceeded the limit of the ecological carrying capacity and caused irreversible damage to the ecological circulation system. Therefore, in this era of rapid social and economic development, it is necessary to find alternative water sources to solve current and future water shortage problems [3].

Alternative sources of water include desalination, rainwater, and recycled water. Among them, recycled water generated by various types of wastewater is considered as a potential water source, which is economical, reliable, and safe, and which has the potential to solve the problem of insufficient supply and demand [4]. Moreover, recycled water is a stable resource that is not affected by seasons. Its production requires less energy and is more environmentally friendly than seawater desalination; the production of recycled water can also alleviate local water environmental pollution [5,6]. The use of recycled water can be traced back to the mid-19th century; however, at that time, it could only be used for irrigation by farmers in poor areas due to the unavailability of high-efficiency wastewater treatment technology. With the continuous expansion of the application scope of recycled water and the emergence of modern science and technology, such as membrane technology for sewage treatment [7,8], the technical treatment level of recycled water has reached the quality standard for indirect or even direct consumption [9,10]. The technology for recycled water use is no longer the main obstacle hindering the promotion of recycled water; instead, it is the public’s low acceptance of recycled water [11]. Research has found that in many places, such as the United States and Singapore, there have been cases of failed promotion of recycled water, which was due to the public’s low acceptance of recycled water [12]. Therefore, it is extremely important to understand the public’s views on recycled water and to avoid the failure of recycled water use.

Existing studies have found that the public’s response to recycled water use is complex, and the acceptance of recycled water is affected by different dimensions and driving factors [13,14,15,16,17,18]. Among them, sociodemographic factors, such as age, gender, and education level are the most basic variables when scholars studied the factors influencing the public’s willingness to accept recycled water. This indicates that these variables are particularly important in determining the influence of social demographic factors on the public’s willingness to use recycled water during promotion, especially at the difficult initial promotion stage. At the same time, identifying the profile customers is also very important. However, public research on the relationship between recycled water use and various social and demographic characteristics has produced conflicting results. Menegaki et al. (2007) believe that the elderly are more concerned about recycled water than younger individuals [19], and some studies have found that young people are more willing to accept recycled water use than older people [20,21,22,23,24]. Some studies have found that women are more receptive to recycled water than men because they are more influenced by social effects [25,26], while Zhu et al. (2019) found that gender and age had no relationship with the public’s willingness to accept recycled water use [27,28,29,30]. Therefore, to realize the use of recycled water and study people who are more likely to accept its use at the initial promotion stage, it is necessary to determine the influence of social demographic factors such as gender, age, and education level on the public acceptance of recycled water use in order to determine the common characteristics of target customers.

A meta-analysis mainly focuses on the results obtained by different researchers under the same topic for a specific research topic. By integrating the results of various studies, their significance and influence direction can be verified to avoid differences in results caused by sampling errors and other factors in a single study [31]. The findings of existing studies are diverse in terms of how sociodemographic factors such as gender, age, and education level affect the specific impact of public acceptance of recycled water use. Therefore, meta-analysis was an acceptable method used in this paper to quantitatively analyze the specific impact of gender, age, education, and other social demographic variables on the public’s willingness to use recycled water. By using a meta-analysis, the profile customer group with common social attributes in the early stage of recycled water use can be obtained, facilitating initial small-scale promotion of recycled water use, and then gradually attracting more people to accept recycled water use with the help of effective marketing and communication strategies, to ultimately achieve the goal of widespread public acceptance of recycled water.

For the above issues, this paper proposes the following research hypotheses. Hypothesis one states that because females are more sensitive to recycled water pathogens [32], and males are more likely to accept the technical risks that may exist during recycled water treatment, men are more willing to accept the use of recycled water than women. Furthermore, with age, the public has a stronger awareness of health and is more risk averse. Therefore, compared with the elderly, the young public is more willing and more likely to accept the use of recycled water [33,34]. Thus, hypothesis two states that young people are more willing to accept the use of recycled water. It is generally believed that higher education is positively related to the public’s willingness to accept recycled water [35,36]. Therefore, hypothesis three states that higher education is positively correlated with the acceptance of recycled water use. Finally, hypothesis four states that intermediate variables such as different regions and research models have moderating effects on gender, age, and education level.

Following the above overview, this paper analyzes existing research on the public’s influence on the willingness to use recycled water by meta-analysis, as shown in Figure 1. First, we searched existing documents in the Web of Science (WOS), Scopus, and CNKI databases by keywords to determine an initial document library (literature database). The retrieved literature was further screened in the order of identification, screening, eligibility, and inclusion, which also prepared the library for a subsequent meta-analysis to extract the original data. During screening, documents that were not reviewed by the same level and conference types were also supplemented to prevent omissions. To better explain the results of the meta-analysis in Part 3 and analyze its guiding significance in facilitating the promotion of recycled water use, a basic overview and the main steps of the meta-analysis, such as the heterogeneity test and publication bias, are introduced in Part 2. The fourth section mainly discusses targeting young female groups with higher education levels in water-poor areas as profile customers in the early stages of the recycled water use promotion. Then, with the help of the “herding sheep effect” to drive other groups, the use of recycled water on a larger scale can eventually be realized. Moreover, to improve the public’s acceptance of recycled water, countermeasures and suggestions are provided. Finally, based on the limitations of this study and the conclusions of previous studies, directions for future research are determined.

2. Material and Methods

To verify these hypotheses and determine the common social attributes of profile groups that are more likely to accept the use of recycled water in the initial stage of recycled water promotion, this study analyzes existing research on the public’s influence on the willingness to use recycled water.

2.1. Data Materials and Literature Search

To identify the relevant research literature, this study used

Table 1. Keywords considered for search.

Term Type	Search Term
Technical terms	recycl*, recla*, alternative water resources, non-conventional water resources, water reuse, wastewater, greywater
Social terms	Percept, attitude, belief, acceptance, resist, disgust, support, oppos, yuck factor, agreement, impact, benefit, participat*, public

to search the three literature databases of WOS, Scopus, and CNKI, and retrieved original literature based on two different categories of keywords related to the technological and social use of recycled water. We used the “or” operator for technical terms and the “and” operator for social terms during each database search. In addition to the database search, references related to the study were evaluated to ensure as much as possible that no other studies were overlooked during the search, and then the final dataset was imported into Endnote for further analysis.

A total of 752 relevant studies were preliminarily searched, of which 625 were obtained from the Web of Science, 87 from Scopus, and 40 from CNKI. The search also included literature that had not been peer-reviewed, such as meeting minutes and book chapters. First, 38 papers were excluded due to repletion and used systematic evaluation, 559 papers were removed by reading titles and abstracts, and 2 papers were supplemented by references. The sample size of this study was finally determined to be 88. Then, the full text was assessed based on the following criteria to finally determine the sample size for this study: first, the writing language had to be Chinese or English; second, the research content was the public’s willingness to reuse or the factors which affected public use of recycled water during reuse; third, the literature had to have enough data information to directly calculate or transform to obtain the effect size and standard error or confidence interval of this research, and a clear sample size was required; fourth, because of the screening conditions in article 3, the research method had to be quantitative, and studies based on qualitative research and systematic evaluation were excluded; fifth, for research published in different journals by the same or different authors using the same data source, literature with more comprehensive information was selected. The amount of literature screened at each stage can be found in Figure 2.

2.2. Document Coding and Quality Evaluation

Different independent studies may cause differences in research results due to different research areas and empirical methods. Therefore, while encoding the original text, we strove to reflect the characteristics of each independent study in detail, including the first author’s name, publication date, research area, sample size, research method, and sampling method. The coding results are shown in

Table 2. Research sample, coding, and quality evaluation score.

Author	Year	Country	N	Model	Method	Score
Boyer et al. [20]	2017	US	486	Probit	Random	15
Savchenko et al. [4]	2019	US	760	Logit	Random	14
Suri et al. [38]	2019	US	746	Logit	Random	16
Redman et al. [23]	2019	Nevada	474	Logit	Random	15
Massoud et al. [29]	2018	Beirut	297	Logit	Random	16
Lazaridou et al. [39]	2019	Greece	302	Multiple linear regression analysis	Random	13
Etale et al. [2]	2020	Australia	480	Multiple linear regression analysis	Random	16
Fielding et al. [2]	2020	South Africa	467	Multiple linear regression analysis	Random	15
Savchenko et al. [24]	2018	US	393	Logit	Random	14
Glick et al. [15]	2019	US	1000	Logit	Random	13
Ellis and Savchenko [26]	2019	US	907	Logit	Random	14
Rice et al. [17]	2016	US	1329	Multiple regression analysis	Random	15
Dean and Fielding [40]	2016	Australia	5194	Regression analysis	Random	14
Wester et al. [32]	2015	US	207	Regression analysis	Random	18
Dolnicar and Hurlimann [41]	2011	Australia	3000	Regression analysis	Random	14

Note: N is the sample size; Score is calculated according to the JBI (Joanna Briggs institute) scale.

.

The quality of the included literature was evaluated using the JBI scale, the cross-sectional study migration risk assessment criteria developed by the Joanna Briggs institute (JBI). The literature quality was evaluated from ten aspects. Generally, a score of >14 is considered to be high quality. The specific evaluation results are shown in Table 2. Except for the research score of Lazaridou et al. (2019) and Glick et al. (2019), the score was 13 points. In addition, the JBI score of all the other included studies was 14 points or above, indicating that the quality of the fifteen literature sources included in this study was relatively high, and can ensure the reliability of the meta-analysis results.

In addition, while coding the included literature, it was found that most of the included studies used women as the reference category and analyzed the actual age and the different education levels of the public. For example, Wester (2015) divided the education level of the public into five levels: lower than high school, high school degree, a university master’s degree, master’s degree, and higher than master’s degree to study the impact of different knowledge levels on the willingness of the public to use recycled water [32]. Therefore, based on existing literature research, females were still selected as the “control group” for the variable of gender, and data analysis was conducted on this factor by means of dichotomous variable measurement. For age and education, the data were analyzed by measuring actual age and different levels of education. Additional variables, such as region and publication time were extracted as moderating variables to help explain the research heterogeneity.

To ensure the reliability of the screening process and avoid subjective bias, the processes of literature retrieval, coding, and quality evaluation were completed by two researchers independently. When disputes arose, they were resolved by discussion. If there were still differences, the assistance of a third party for judgment was sought, and if there were not enough literature data during the coding process, the original author was contacted for supplementation; otherwise the literature was excluded. After completing the initial coding of the included literature, meta-analysis was conducted using the open source “metaphor” package [37] in “R” and the “metan” package in Stata.15.

2.3. Meta-Analysis

Meta-analysis is a quantitative literature review method, which was initially mainly used in the medical field to analyze the effect of a certain drug or therapeutic effects [42]. Since the 1870s, this method has been widely used in the fields of psychology, sociology, and economics. The reason this study adopts the method of meta-analysis is that many quantitative studies have been conducted on the relationship between gender, age, and education level in social population variables and the public’s willingness to accept recycled water, which can satisfy the requirements of meta-analysis on the amount of literature. Meanwhile, meta-analysis can combine the diversified quantitative results and qualitative conclusions between social demographic variables and the public’s willingness to accept recycled water, to achieve scientific induction from individual to general conclusions.

What is different from experimental disciplines such as the natural sciences and medicine is that in quantitative analysis there are very few experimental groups and control groups that have the same conditions, except for the treatment effect, and the relationship is more bivariate. Therefore, according to the recommendation of Pigott (2012), Hedges d, which is unrelated to the sample size, is selected as the effect size in this study [43]. When Hedges d > 0, it means that the combined effect of this variable has a positive effect on the willingness to use recycled water. Hedges d < 0 is a negative effect, and Hedges d = 0 means no effect. Currently, the significance of Hedges d is verified by a significance test. In addition, due to the inter-study heterogeneity among the included analysis, random effects are chosen instead of fixed effects to calculate the effect size [44]. The meta-analysis of random effects is a special case of general linear models. Although it cannot solve the heterogeneity problem of included studies, it is helpful to deal with the diversity of research and can be used for a wide range of studies [45,46], indicating the degree of heterogeneity of studies.

The heterogeneity in meta-analysis can be divided into two types: intra-study heterogeneity and inter-study heterogeneity. To assess the heterogeneity of included studies more accurately, Q, Q_p, t, and I² statistics are used in this study. The Cochran’ Q is the overall heterogeneity, also known as the chi-square statistic, and its calculation formula is

$$Q={\displaystyle \sum }_{i=1}^n{w}_i{\left(E_i-\overline{E}\right)}^2$$

where $\overline{E}$ effect is the comprehensive effect size; E_i is the effect size of the i-th study, namely Hedges d; w_i is the reciprocal of the corresponding variance of the i-th study, which is used as the weight of the study; and Q_p is used to indicate the significance of the statistic Q.

However, when only a few studies are included in the meta-analysis, the Q statistic cannot detect the heterogeneity in real data. Therefore, in addition to Cochran’ Q, we also report the statistic I² expressing as a percentage, namely the heterogeneity between studies that observed the effect of the total change in estimate, which provides a quantitative basis for judging whether there is heterogeneity in the study [47]. It is generally considered that heterogeneity is high when I² > 50%. In addition, the analysis of the three variables of gender, age, and education in this study has generated a forest plot and a funnel plot, and we also conducted Egger’s Test for each analysis to detect the asymmetry of the funnel plot and provide a testing basis for judging whether there is publication bias or not. Unlike the funnel plot and the loss-of-safety coefficient, which are used to determine whether there is publication bias by subjective observation (Sterne, 2001) [48], the Egger test can conduct quantitative detection of publication bias [49,50]. Publication bias is a kind of sampling bias, which refers to “the tendency to prepare, submit and publish research results according to the nature and direction of research results” [51,52,53], which refers to statistically significant “positive study results” that are more likely to be published than those meaningless “negative study results” or invalid study results [54], which will lead to the inability to obtain truly representative research samples. Therefore, to avoid sample selection bias caused by the fifteen included literature cases in this study, it was necessary to conduct the publication bias test.

Subsequently, cumulative meta-analysis and meta-regression were used to study data heterogeneity. With the development of society and changes in the environment, the impact of social demographic variables on public acceptance of recycled water use may gradually change [39]. Therefore, the sociodemographic variables included in this study, namely gender, age, and education level, were cumulatively analyzed based on the year of publication. Meanwhile, generating the cumulative forest graph to reflect the various variables of recycled water use that the public may accept affects intensity changes. Different studies lead to research heterogeneity due to different model methods, research time, sampling methods, and research area, and these factors leading to heterogeneity do not originate at the individual level and need to be controlled by the introduction of regulatory variables. Therefore, this study used meta-regression analysis to take the two heterogeneous factors of gender and age in the social demographic factors as the dependent variables, and selected the moderation variables with moderating effects, such as research area, model methods, and journal impact factors as independent variables, to study the sources of heterogeneity within the data.

Contrary to a traditional systematic review, meta-analysis mainly focuses on a specific research subject and studies the results obtained by different researchers through quantitative analysis. In spite of possessing a higher processing level and accuracy, this method still has drawbacks. First of all, a large quantity of preliminary work, such as database selection and document coding, is required to be completed manually. Compared with the system evaluations, which can be processed by software [55], meta-analysis consumes more upfront time. Moreover, only Chinese and English literature sources were considered in this paper, resulting in a narrow scope of the literature search. Secondly, meta-analysis is a quantitative literature review method, whose data originates in the existing literature. However, few authors present the original data in the published literature. Therefore, the factors of the analysis are inadequate due to the lack of necessary data and strict screening conditions of meta-analysis.

3. Results and Analysis

3.1. Descriptive Statistics

Based on the research on public recycled water use as the data source, fifteen articles were finally determined by keyword retrieval and screening of three literature databases including Web of Science, Scopus, and CNKI, and 15,064 samples and 34 effect sizes were provided for meta-analysis.

All studies were conducted by means of random sampling, most of which collected data through questionnaires, and these data were used for subsequent analysis (n = 12). While Lazaridou (2019) and Rice (2016) chose a deeper method of face-to-face interviews to collect data to obtain the thinking process of the participants when filling out the questionnaire [17,38]. Savchenko (2018) used a field experiment with significant preference to recycled water from the public [24]. The willingness preference data used were scientifically measured. The data came from seven different countries; more than half of the studies were conducted in the United States (53%), and most of the remaining studies were conducted in Australia (20%). Gender and age were the two variables with the most complete effect values (twelve studies each), and only ten studies reported effect values related to education.

The random effects model was used to estimate the Hedges d of the willingness to use recycled water, and a 95% confidence interval was reported, as shown in Figure 3. Half of the studies believed that the male groups were more willing to accept the use of recycled water than the female groups. However, the results of quantitative analysis found the opposite result (E = 0.034 > 0). The effect values of age were mostly distributed below the 0 axis, indicating that most scholars believed that young people were more willing to accept the use of recycled water (75%). Among them, Rice (2016) commented on this conclusion [17]. The results of the study were the most significant, while Suri (2019) and Dean and Fielding (2016) held the opposite view, believing that the elderly were more inclined to accept the use of recycled water [40,56], and their contradictory results further confirmed the applicability of the meta-analysis method.

3.2. Main Effect Analysis of Sociodemographic Variables on Public’s Willingness to Accept Recycled Water

Data were processed in Stata and R-Studio software, and meta-analysis was conducted on gender, age, and education level in turn. The results are shown in

Table 3. Summary of the analysis results of social demographic variables.

Variable	E	Confidence Interval		Q	PQ	I2/%	n	N	Weight	Egger’s Test Results
		Lower 95% CI	Upper 95% CI
Gender	0.034	−0.124	0.193	43.37	0.0001	74.6	12	12,443	7.41	t = −0.579, p = 0.575
Age	−0.008	−0.019	0.002	67.51	0.0001	83.7	12	15,064	78.35	t = −1.003, p = 0.257
Education	0.003	−0.06	0.065	19.38	0.0221	53.6	10	11,039	14.21	t = 0.094, p = 0.927

Note: E is the comprehensive effect size; Q is the statistic of the heterogeneity test; P_Q is the significance of Q; I² is the proportion of heterogeneity in the overall change; n is the number of the effect size; and N is the sample size.

, in which women were more inclined to accept the use of recycled water than men (E = 0.034 > 0), and I², Q, and P_Q (74.6%, 43.37, and 0.0001) indicated that the variables had significant heterogeneity. That is, in the current study, gender divergence in the research significantly affected the public’s willingness to accept recycled water, which is consistent with the prerequisites of this research using the meta-analysis method for analysis. With increasing age, the public’s acceptance of recycled water decreased (E = −0.008), and age had the largest weight on public willingness to accept recycled water (78.35%), indicating that age is an important explanatory variable among the influential factors affecting the public’s willingness to accept recycled water. The heterogeneity of education was low (I² = 53.6%), and it had a positive effect on promoting public’s willingness to accept the use of recycled water (E = 0.003 > 0). That is, the higher the level of education, the higher the public’s willingness to accept the use of recycled water.

To avoid the possible failure of obtaining truly representative research results due to publication level, publication bias tests were conducted on fifteen studies through Egger’s test and Tri-and-Fill [57]. Furthermore, the accurate deviation detection results were obtained, which are shown in

Table 4. Meta-analysis results of heterogeneous variables.

Variable	Moderator	B	SE	p	Confidence Interval		I2/%
					Lower 95% CI	Upper 95% CI
Gender	Publication time	0.241	0.372	0.541	−0.669	1.151	69.81
	US	0.514	0.472	0.318	−0.0641	1.67
	Non-logit model	0.368	0.571	0.544	−1.03	1.765
	Impact factors	0.159	0.152	0.333	0.211	0.53
Age	Publication time	0.075	0.451	0.148	−0.035	0.185	67.12
	US	0.352	0.136	0.041	0.02	0.684
	Logit	0.422	0.143	0.026	0.072	0.773
	Impact factors	−0.006	0.005	0.278	−0.177	0.006

. The Tri-and-Fill method can effectively reduce the impact of missing data on the research results, as shown in Figure 4. The solid circle in the figure represents the true value, while the hollow circle indicates the missing literature that are filled. Observing the three funnel diagrams a, b, and c in Figure 4, it was found that the points were concentrated on both sides of the line in the funnel diagram and distributed at the top, indicating that the publication bias errors of the three factors of gender, age, and education were small, and Egger’s test results showed that these three variables all passed the Egger test and inspection, further confirming that there was no publication bias in gender, age, and education.

3.3. Effect of Sociodemographic Variables on Public Acceptance of Recycled Water Reuse over Time

Meta-analysis was mainly used to analyze the differences among the fifteen literature studies. However, with time and the addition of new studies, the influence of gender, age, and education level on the public’s willingness to receive recycled water may change. Therefore, through cumulative meta-analysis, the research included in this study is regarded as a continuous unity to analyze the impact of new research results on existing research results. Meanwhile, a forest plot was used to indicate the impact on the comprehensive results [55], and the results are shown in Figure 5.

The results show that the combined effect and 95% confidence interval of gender (specifically female) and age on the public’s willingness to accept the use of recycled water gradually narrowed and tended to approach one over time. This indicates that the existing scholars’ conclusion that women are more willing to accept the use of recycled water is increasingly robust, which also provides quantitative support for this study to conclude that “young women are more likely to accept the use of recycled water than older men”. Among them, it was found under the selected test standard that the initial statistical significance was the study of Boyer et al. (2017). When performing a cumulative meta-analysis on age, it was found that the research conclusions of Rice et al. (2016) deviated slightly from the overall cumulative meta-analysis results over time [17]. The reason may be that the sampled age was higher than the average age of all fifteen study samples. Therefore, it affected the outcome, which is that young people are more willing to accept the use of recycled water than older people. Education also used Boyer et al.’s (2017) research as a node, and its impact on the public acceptance of recycled water gradually stabilized, and with the improvement in the level of education, the public’s willingness to accept recycled water gradually increased. In addition, education eventually approached “1”, indicating that with the improvement in public education level, the willingness to accept recycled water gradually increased.

3.4. Analysis of the Moderating Effect of Moderating Variables on Social Population Variables

Compared with the heterogeneity test and cumulative meta-analysis that can only be analyzed by a single factor, meta-regression can simultaneously analyze multiple factors affecting the relationship to the public’s willingness to use recycled water to determine the adjustment factors that affect the relationship factors between the two, and discuss the contribution of each factor to the relationship. Therefore, meta-regression analysis was conducted on the aforementioned variables with heterogeneity of gender and age to determine the heterogeneity source.

With the addition of research samples and the generation of new conclusions, different publication times may have an impact on the social population variables and effect size of the public’s willingness to use recycled water. Therefore, publication time was selected as one of the moderating variables. Etale and Fielding et al. (2020) found that different regional cultures play a central role in the public’s use of recycled water [2]. That is, differences among regions can represent differences in public culture and water resources abundance. Therefore, the study region was included as a moderator variable. The selection of different estimation models may impart bias to the research results, and different journal impact factors may also cause differences at the publication level. Therefore, the estimation model and journal impact factors were also introduced as regulating variables. The publication date of the literature was taken as the publication time instead of the online publication time. Limited by the sample size of the included literature, the study area was divided into two major regions, namely, the United States and non-United States. The estimation model was the logit model or non-logit model, and the impact factor was the actual value obtained through the JCR query in the Web of Science.

Meta-regression results with two heterogeneous variables, gender and age, are listed in Table 4. The results show that gender heterogeneity was reduced after controlling for publication time, region, and other regulatory variables (I² = 69.81%). However, the significance of each variables was weak (p > 0.05), indicating that there may be potential regulatory factors that have not been explored. Among them, the regional influence was greater (p = 0.318), which means that the region should be further divided into different regions in future research to further analyze the possible impact of the differences in different regional cultures and water resources abundance on this factor. After controlling the four moderators, the age heterogeneity was significantly reduced (I² = 67.12%) and was significantly different in the region (p = 0.041) and the model estimation method (p = 0.026), indicating that the region and the model estimation method were the reasons for the large age heterogeneity. Compared with the United States, the influence of this variable in other regions was smaller. Compared with the logit model, other model estimation methods had a weaker influence on this variable.

4. Discussion

To analyze the impact of social demographic variables on public willingness in the initial stage of recycled water use promotion, this study conducts a meta-analysis, and the empirical results highlight the main research objectives of this study.

First, young women with middle and high education levels are more willing to use recycled water. This study used a random effects meta-analysis model to evaluate public social population attributes of gender, age, and education, and found that women who are more risk averse are more willing to accept the technology risk of recycled water (E = 0.034) [58]. This is contrary to our hypothesis that men are more willing to accept recycled water than women. It is speculated that the reason may be that the influence of this factor on public willingness to accept recycled water is non-linear. In addition, it may be that female groups are more likely to be persuaded by the existing evidence about the cleanliness of recycled water than male groups, indicating that during promotion of recycled water use, the preferred target group should be women. Compared with older people, young people are more willing to try recycled water (E = −0.008). According to hypothesis two, young people are more willing to accept the use of recycled water, which is corroborated by our analysis. This conclusion is related to the fact that young people are more sensitive to changes in a series of external factors, such as rapid population growth, increasing urbanization rates, and severe climate change, and pay more attention to the phenomenon of serious water scarcity. Meanwhile, as older people pay more attention to health issues, they are less willing to receive recycled water due to their concern about the risk attributes of recycled water. With the improvement of the level of public education, the clearer the understanding of the technological process and treatment process of recycled water, such as membrane treatment technology, the lower the fear of new things, and the higher the willingness to accept recycled water [59], which provides a basis for the verification of hypothesis 3 (E = 0.003).

Second, the profile of recycled water use supporters is stable and does not change over time. The cumulative meta-analysis in this study finds that the existing studies on the influence of gender (I² = 74.6%) and age (I² = 83.7%) on the public’s willingness to use recycled water is slightly divergent. However, the group willing to use recycled water will not change significantly long term (see Figure 4); therefore, the young female group is still chosen as the main profile customers for the promotion of recycled water use. In addition, through the impact analysis of degree of education, we found that with the increase of new research published and new samples the effect on existing conclusions is almost negligible. Therefore, water recycling use can be promoted initially from colleges and universities as a pilot, during the promotion implementation, and then be popularized in broader society with the help of teachers and students.

Finally, regional and model estimation methods have a regulatory effect on the willingness of recycled water use supporters. In this study, meta-regression was used to analyze the potential moderator that may affect the public’s use of recycled water. The results show that the influence of different regional cultures on social and demographic variables are consistent with previous studies [2,14,21,32]. That is, women in different regions have different degrees of willingness to accept recycled water. For example, compared with American female groups, Indian groups are more likely to accept the use of recycled water [59], and the gender’s willingness to accept the use of recycled water is also affected to a certain extent by the differences in potential variables such as the degree of water resource abundance and different economic development conditions. Different cultures, differences in the degree of water resources richness, and different model estimation methods in different research areas are the main reasons for the varying conclusions with regard to the public’s willingness to use recycled water at different ages. Therefore, it is wise to select the public in water-scarce areas for initial promotion of recycled water use.

5. Conclusions

As an alternative water source, recycled water plays an important role in alleviating water shortages and promoting environmental protection. The promotion of recycled water has been of wide concern among scholars worldwide. In contrast to existing studies that investigate the acceptance factors of recycled water from an individual level, this study uses the meta-analysis method to evaluate gender, age, and education as social demographic variables, which provide substantial evidence of the willingness to accept recycled water by the public, by searching literature data from 2006 to 2020 through keywords. Additionally, by collating the data of 15,064 participants from seven countries, it was finally concluded the target customers in the initial promotion stage of recycled water in areas with water shortages.

The results show that young female consumers with higher education levels are more likely to accept the use of recycled water in areas with water shortages. Therefore, to the public, especially to female groups, promote and popularize the concept, provide education on the characteristics and technology of the recycled water treatment process, expand the popularity of recycled water in public life, further strengthen the advantages of low energy consumption, environmental protection, and reliable technical treatment of recycled water in the public mind, and reduce public’s fear of the newness of recycled water. By improving the utilization rate of recycled water in the daily life of the female group, the beneficial publicity of recycled water is carried out by virtue of the perceptual characteristics of the female group and the “interpersonal communication” mode. Meanwhile, promoting use of recycled water in a small range in younger groups, such as student groups or young teachers with a higher ability to accept new things. This leads to auxiliary water saving consciousness of the public at the same time, and reduces the public risk perception of recycled water, thereby facilitating recycled water acceptance with a gradually expanding scope to the older group, and eventually realizing the goal of improving the recycled water utilization ratio.

In addition, this study finds that the influence intensity of many other variables decreases, indicating that there may be other potential important variables that have not attracted enough attention. For example, the gender’s willingness to accept recycled water use is also affected by potential variables such as differences in water resources abundance and different economic development conditions. Therefore, in future research, it is necessary to further verify whether the public in areas with richer water resources and developed economic development levels are less willing to accept recycled water. Most existing studies use structured or semi-structured questionnaires, and the questionnaire design contains more subjective parameters, which may omit more important factors. Therefore, future research should be conducted from the perspective of the public and based on open questionnaires, such as social complex network and grounded theory, to find more important factors. Due to the lack of necessary data, the common characteristics of profile groups are analyzed only from the perspective of public sociodemographic factors. Therefore, in future studies, more efforts should be made to explore how the public’s psychological factors surrounding recycled water, such as the risk perception, understanding, and aversion to use recycled water, to better realize the promotion of recycled water use.