Can the Process of Evaluation and Recognition Effectively Promote Water Conservation in Cities? Evidence from China

Abstract

Water-saving construction is a crucial technique for China to address water resource scarcity and tackle the water-use issues. Can evaluation and recognition result in urban water-saving construction, and can it produce a more effective water-saving effect with the help of smart city construction? Based on the unbalanced panel data of prefecture-level cities in China from 2006 to 2019, this paper empirically discusses the impact of China’s National Water-saving City Selection on urban water conservation by using the multi-time point difference-in-differences method, and on this basis, it includes the consideration of smart city construction, so as to investigate the strengthening role of urban digital management in urban water-saving construction. The findings show the following: (1) the National Water-saving City Selection has significantly improved water conservation in the evaluated cities, and it has contributed to the positive effect via direct and indirect channels. (2) The water-saving effect generated by the policy varies with the total urban water supply or the number of people with a higher education in the city. (3) The implementation of the smart city pilot policy has significantly strengthened the city’s existing water-saving construction, and a further mechanism analysis shows that it not only strengthens the existing direct impact channels but also compensates for past construction shortcomings.

1. Introduction

The uneven distribution pattern between the southern and northern regions, combined with the influence of climatic factors, contributes to the disparities in precipitation distribution in China. Under the rough development model, the substantial degradation of regional water quality resulting from industrial and agricultural practices, combined with a lack of awareness about water conservation, has intensified China’s water resource scarcity [1]. The limited per capita water supply, elevated water demand, and competition for water resources among diverse industries have presented substantial obstacles to advancing both production and daily life in China [2]. In light of this, China has not only consistently improved national and regional water resource management measures and implemented the most stringent water resource management system compared to the previous system in China to advance wastewater treatment, bolster water usage control, expedite water conservation projects, and accomplish ecosystem restoration, but also uses the power of evaluation and recognition to promote the transformation of production and living water conservation in the region, and the National Water-saving Selection is one of them. According to documents from the Ministry of Housing and Urban-Rural Development of the People’s Republic of China (MOHURD), the National Water-saving City Selection commenced in 2002 and has been conducted 11 times to date, resulting in the successful designation of approximately 140 cities. The policy’s primary objective is to tackle the underlying issue of water resource scarcity in China through water conservation initiatives, with the ultimate goals of safeguarding regional aquatic ecosystems and the water environment, while also ensuring the quality and safety of the residential water supply. The policy expressly stipulated that the assessment and verification of water-saving cities should be conducted every two years, with a comprehensive review every four years. The assessment criteria primarily encompass four categories, including fundamental management indicators, technical assessment indicators, and incentivizing indicators. These criteria encompass a wide range of projects, such as regulatory and institutional development, investments in water-saving technology, and urban wastewater treatment. Simultaneously, in the era of global digitalization, the integration of digital transformation with urban governance has driven the development of new urban infrastructure and the further optimization of urban governance models. In 2012, China enthusiastically launched the National Smart City pilot program, paving the way for innovative urban planning and management models that fully leverage modern technology.

Based on the above, this paper attempts to explore the following issues: whether the National Water-saving City Selection, for the purpose of evaluation and recognition, can promote water conservation in cities, through what mechanism it will achieve the effect, and whether the effect will be different due to different external factors. On this basis, whether the new urban digital governance model of the smart city can have a significant strengthening effect on the water-saving effect of the policy is determined.

In accordance with the National Water-saving City Selection policy, this study employs a multi-time point difference-in-differences method and utilizes non-balanced panel data from 2006 to 2019 for prefecture-level cities. It assesses the impact and efficacy of the recognition policy on regional water conservation efforts, taking into account the diversity of factors and their potential influencing mechanisms. Additionally, by incorporating the smart city pilot policy, this research investigates the potential benefits of urban digitization and intelligent management for regional water conservation initiatives and the enhancement of related avenues.

The paper’s main contributions are as follows: first, we broaden the scope of evaluation indicators employed in assessing the effects of the selection and recognition policy. This expansion goes beyond the previous focus on economic indicators, like the regional GDP and the labor force, as well as environmental pollution indicators, such as industrial SO₂ emissions and wastewater emissions. It now includes a resource utilization indicator, namely the regional water conservation. This extension allows for an examination of the potential impact of the evaluation and recognition policy on the regional water conservation status. Second, building on the influence of the evaluation and recognition policy, this paper takes into account the augmenting role of smart city construction within the context of the ongoing national trend of digital innovation. This provides evidence of the positive impact of urban digitization and intelligent management systems on urban resource allocation and utilization. Third, we employ the widely-used multi-time point difference-in-differences method, a commonly used policy evaluation approach in academia, to empirically assess the water-saving effect and its underlying mechanisms in National Water-saving Cities. This enriches the quantitative research on water conservation.

2. Literature Review

2.1. Evaluation, Recognition, and Environmental Governance

The region’s development needs and unavoidable “free rider” behavior make it easy for the local government to exhibit an obvious “bottom-by-bottom competition” phenomenon in the process of environmental governance [3,4], making a horizontal neighborhood behavior comparison an important factor affecting the local government’s environmental governance efforts [5]. As a result, while optimizing the evaluation of local environmental performance, it has gradually become an effective means of improving local environmental governance by forming a benign interaction between localities and between localities and the central government and encouraging “top-to-top competition” in environmental governance between neighboring places [6,7]. In this regard, China’s evaluation and recognition policy has become an important tool for the central government to strengthen local environmental governance and overcome the inherent weaknesses of local governance through ideal selection criteria and a tight assessment system. Most of the existing studies have discussed its environmental effects from the following aspects: Liu et al. (2023) found that the brand effect brought by the selection of civilized cities can significantly reduce the emission of PM2.5, SO₂, and CO₂ in selected cities by improving the level of regional technological innovation and promoting the upgrading of local industrial structure, and the effect is more obvious with the improvement of the urban expansion level and economic level [8]; Li et al. (2023) confirmed that the measures taken by local governments to strengthen environmental regulation, improve the innovation ability of enterprises, and enhance public participation in environmental protection, in order to obtain the honorary title issued by the central government, will generate low-carbon effects and promote local green development [9]. Simultaneously, some studies have examined the impact of appraisal and recognition on regional sustainable development based on the establishment of an urban sustainable development index. It has been discovered that the designation of a civilized city can support local sustainable development by successfully driving the level of urban industrial agglomeration [10]. Furthermore, the level of urban green innovation will be enhanced as a result of the increased scale of local investment brought about by the city selection [11]; not only that, but participation in the selection has effectively supported the growth of local green total factor productivity, thereby improving the degree of local economic development while optimizing the quality of the local environment [12], and on this premise, the resource-based cities who participated in the selection and were chosen have greatly improved their resource allocation effectiveness [13]. Finally, local governments’ participation in city selection activities will encourage local businesses to boost their investment in environmental protection construction and actively assume environmental protection obligations [14].

2.2. Assessment of Water-Saving Effect of Existing Policies

Some scholars have discussed the water-saving effect of water-saving or energy-saving policies by means of an index calculation or input–output model. For example, Zarezadeh et al. (2023) evaluated Iran’s water-saving policies at the macro level from the perspective of water footprint sustainability indicators [15]; Alun et al. (2014) used an input-output table to study the association between China’s industrial energy consumption and water usage against the backdrop of China’s “Eleventh Five-Year” energy-saving program, and discussed the prospect of synergistic savings between the two [16]. Some studies have analyzed the water-saving effect of the construction of a water-saving society using the difference-in-differences method or the hydrological and economic input–output model at the basin scale based on the city level or the basin level, based on the policy background of the construction of a water-saving society in China [17,18]. Furthermore, some researchers employ the difference-in-differences method to assess the influence of water-saving programs on agricultural water use; for example, using the release of China’s national agricultural water-saving outline in 2012 as the research backdrop, Xu et al. (2021) discovered that the outline’s implementation did not effectively reduce agricultural water consumption, but instead resulted in a sharp increase in agricultural water extraction due to the expansion of the irrigation area [19]. Simultaneously, highlighting the benefits of water restriction and enhancing the fairness and justice of policy implementation would enhance farmers’ attitudes about the implementation of command-type water-saving regulations based on an agricultural water-saving outline [20]. However, Zhang et al. (2021) discovered that a water rights pilot project implemented between Chinese provinces in 2014 can reduce local agricultural water use by improving agricultural technology innovation and promoting the transfer of water rights to high water-consuming industries [21]. Furthermore, by improving the industrial structure and increasing environmental management, the trial application of water rights can reduce regional wastewater output while accomplishing water conservation [22]. Finally, starting from the demand side, Stavenhagen et al. (2018) used four cities in Europe as research samples, employing a mixed qualitative and quantitative research method to confirm that demand-side water management policies can significantly reduce residents’ domestic water consumption, thereby improving urban water conservation [23].

In addition to the above research, some scholars have also considered the differences in economic and geographical environments, and then based on the diversity assessment method, they have conducted in-depth discussions on different types of water-saving policies in different countries and the possible formulation directions of policies. For example, Lee et al. (2013) used sampling surveys and other methods to evaluate the incentive water-saving measures implemented for specific groups in Miami-Dade County, Florida, the United States, from the perspective of user use, and thus affirmed the positive impact of this measure on residents’ water conservation and the important role of participant satisfaction in the implementation of water-saving measures [24], and based on the problem of agricultural water use in Brazil, some scholars used the calibrated and validated hydrological model SWAP/WOFOST to explore the important impact of water-saving irrigation management on agricultural water saving and water productivity improvement [25]. Qureshi et al. (2011) took the Murray-Darling Basin in Australia as the research object and compared the subsidized investment in irrigation water and its alternative measures, such as the purchase of water in the water market from the perspective of water use efficiency and economic efficiency, and found that the cost-effectiveness of the former was relatively lower than that of the latter [26]. At the same time, some scholars found that under different water prices, the contradiction between European agricultural policies and water management policies will have different effects on water saving in the production processes of different crop species [27]. In addition, Soula et al. (2021), taking the Mahdiya region of Tunisia as the research object, demonstrated the pressure of the existing water management system on the exploitation of groundwater resources in the case of an imperfect collective action policy setting and the necessity of promoting an efficient collective action formulation, promoting water price reform and strengthening water-saving sustainability [28]. Due to the word limitations of this article, more related research will not be discussed here.

Although the existing research has analyzed the impact of the evaluation and recognition policy, the discussion of its environmental quality improvement is mostly focused on air pollutants, low-carbon development, etc., and the policy background is mostly China’s national civilized city selection. Therefore, the improvement effect of resource utilization and the implementation effects of other evaluation and recognition policies are not analyzed in depth. In terms of a water-saving effect evaluation, few studies consider the impact of China’s top-down evaluation and recognition policies on regional water-saving effects. Therefore, the discussion on it will expand the relevant research on the effect of a top-down evaluation and recognition policies on regional resource utilization and conservation. At the same time, on the basis of the evaluation and recognition policies, few studies consider other policies, especially the urban digital construction, which promotes the upgrading of urban resource management methods and may have a complementary role in the implementation of the evaluation and recognition policies or may have a policy synergy effect with the evaluation and recognition policies.

3. Research Hypothesis

According to the essential requirements outlined in the “Declaration and Evaluation Method for Water-saving Cities and the Evaluation Standards for Water-Saving Cities”, a veto system is introduced in the assessment of system and institutional construction. This is intended to ensure the fundamental effectiveness of urban water conservation efforts. Furthermore, the designation of National Water-Saving City plays a critical role in signaling the advancement of local officials [29], motivating them to strictly follow the selection criteria, and scoring requirements when implementing urban water conservation projects. Additionally, the four-year review mechanism secures the long-lasting efficacy of the policy. Based on the above content, this paper proposes the following:

Hypothesis 1.

The National Water-saving City Selection policy has the potential to enhance the water-saving practices in the selected cities.

Second, according to the published scoring criteria, cities aspiring to achieve the status of a National Water-Saving City should not lag behind in technical evaluation indicators. Such initiatives can encourage regions to promote water resource reuse and strengthen the overall water supply capacity, consequently increasing regional water savings. Simultaneously, the primary goal of public infrastructure construction, focused on improving water quality, ensuring water supply safety, and optimizing regional water conservation, undoubtedly enhances the quality of local public service provisions and bolsters the region’s capacity to attract foreign labor [30]. This promotes the concentration of high-skilled labor and innovative talent. Therefore, stringent evaluation criteria will compel local enterprises to innovate and thrive [31]. The combination of the two elements above will propel the development of local comprehensive innovation skills while also assisting the region in generating a water-saving effect. Based on this, the second hypothesis of this paper is proposed:

Hypothesis 2.

The National Water-saving City Selection policy will improve the regional water-saving effect by raising regional recurring water consumption, strengthening regional comprehensive water supply capacity, and fostering regional comprehensive innovative capabilities.

In addition, the potential impacts of the policy may vary based on the region’s overall water supply and the percentage of residents with higher education. Concerning the total water supply, government departments in regions with limited water resources or relevant stakeholders in areas with an abundant water supply tend to prioritize regional water conservation, thereby affecting the strength of the evaluation’s influence. The water conservation consciousness of the local population plays a crucial role in promoting regional water-saving initiatives, and educational attainment is a key factor influencing individual water-saving behavior. Individuals with higher levels of education exhibit greater water-saving awareness compared to others [32]. Consequently, this program is likely to yield more pronounced water-saving effects in areas with a relatively lower proportion of the population having a higher education. Based on these observations, this paper proposes the following:

Hypothesis 3.

Due to differences in the total amount of the regional water supply and the number of persons receiving higher education in different areas, the selection of National Water-saving Cities will have a diverse impact on the regional water conservation.

Finally, the smart city pilot initiative requires each pilot area to adhere rigorously to the specifications outlined in the “National Smart City (District, Town) Pilot Index System (Trial)” for the completion of local smart city construction. This initiative aims to enhance urban public service infrastructure through digital management and a diverse array of modern information technologies. Its objective is to achieve the comprehensive management and monitoring of the water supply, water conservation, heating, lighting, air quality, and water quality, ultimately enhancing the overall livability of the city. Consequently, the advent of smart city construction will drive government agencies to optimize the utilization of local science, technology, and human resources. Simultaneously, it will expedite the transmission and processing of company information, providing an intrinsic impetus for local innovation [33]. Thus, this paper proposes the following:

Hypothesis 4.

In the process of participating in or being evaluated as a National Water-saving City, if the region also obtains the pilot qualification of smart city construction, the latter will strengthen the urban water-saving construction and improve the local water-saving effect to a certain extent. It will not only strengthen the existing water-saving path in the regional water-saving construction but also supplement the shortcomings in the water-saving construction on the basis of improving the information infrastructure.

4. Empirical Strategy

4.1. Regression Model

In this paper, a series of empirical tests will be completed by using the multi-time point difference-in-differences method. The reason for selecting this evaluation method is to take into account the implementation characteristics of the multi-period development of the policy, the final results of selection and non-selection, and the recognition of the multi-time point difference-in-differences method by relevant scholars in the evaluation of the effect of similar multi-period selection policies [8,9,10,11]. The exact settings of each model are provided in Models (1) and (2), with reference to the current literature [34]:

y_it = α₀ + β₁treatw_i × postw_t + δX_it +λ_i + μ_t + ε_it

(1)

y_it = α₁ + β₂treatwz_i × postwz_t + θX_it +λ_i + μ_t + ε_it

(2)

Among them, Model (1) is used to evaluate the impact of National Water-saving City Selection on regional water conservation, while Model (2) is used to consider the change in regional water conservation under the dual policy of National Water-saving City Selection and the smart city. In Model (1), the subscript i represents the individual city, and t represents the year; y_it is the explained variable; treatw_i×postw_t is the core explanatory variable in this model; X_it is the control variable at the city level in the model; ε_it is a random disturbance term; and λ_i and μ_t are individual fixed effect (CITYFE) and time fixed effect (YEARFE), respectively. The specific setting of each variable will be described in detail below. Finally, β₁ is the core explanatory variable coefficient, that is, the effect of the National Water-saving City Selection on urban water conservation. In addition to changing the core explanatory variable, Model (2) is adjusted for the interaction term of treatwz_i and postwz_t, and the remaining settings are the same as in Model (1).

4.2. Variable Description

This paper chooses urban water saving (waters) as the explanatory variable in the model to directly reflect the amount of urban water conservation. The interaction term between treatw_i, which indicates if the city is a National Water-saving City, and postw_t, which represents whether the year is assessed and the year following it is evaluated, is the core explanatory variable in Model (1). Among them, this paper takes the first year of the evaluation year as the starting year of the policy, taking into account the provisions of the “Water-saving City Declaration and Assessment Method” for the assessment period and the relevant practices of the existing literature for the selection year in the selection and recognition study [35]. The core explanatory variables in Model (2) refer to relevant practices from the existing literature [36,37,38], with the interaction term of treatwz_i and postwz_t, which represents whether the representation is the postwzt of the most recent implementation year and subsequent years in the two policies. The National Water-saving City’s selection year is set as Model (1), and the smart city’s year setting is based on the practice of Song et al. (2023) [39].

This paper controls the per capita GDP (rpgdp) and its square term in the model, as well as the actual total investment in fixed assets of the entire society (rinv), regional fiscal decentralization (fin), that is, the ratio of fiscal expenditure to fiscal revenue, urban population density (pod), and urban water supply pipeline density (alltubel). Furthermore, in the robustness test, this article also adds the annual average precipitation (rain), annual average temperature (temp), and water use penetration rate (wp).

In addition to the above settings, this paper also deals with the variables as follows: first, in addition to the ratio variable and the measurement index less than 1, this paper logarithmizes the variables with absolute values and the measurement index greater than 1, that is, the urban water saving, repeated water consumption, actual per capita GDP, actual total fixed asset investment in the whole society, leakage water, urban innovation index, average annual precipitation, and annual average temperature in the model, and logarithmizes some of the variables containing 0 values, that is, water saving, repeated water consumption, and leakage water, after adding 1; second, using 2003 as the base year, the actual per capita GDP and total fixed asset investment of the entire society are deflated by the GDP index and fixed asset investment price index.

4.3. Data Sources and Descriptive Statistics

The data in this research are mostly derived from the China Urban Construction Statistical Yearbook, the China Statistical Yearbook, Kou et al.’s calculated and published China Urban Innovation Index [40], and the National Meteorological Science Data Sharing Service Platform. Among them, the China Urban Construction Statistical Yearbook, which serves as the primary data source for this paper, has some deficiencies in urban water conservation and repetitive water usage. Therefore, this paper first eliminates the individuals that are completely missing; then, the linear interpolation method is used to fill in the above variables and eliminate the abnormal values and their previous interpolation terms in continuous interpolation to ensure the effectiveness of interpolation. Finally, on the basis of the above data, individuals that have only appeared once in the sample are eliminated.

Table 1. Variable descriptive statistics.

Variable	N	Mean	Sd	Min	Max
lnwaters	2273	6.7602	1.548	0	12.3919
treatw	2273	0.3528	0.478	0	1
postw	2273	0.2684	0.4432	0	1
treatwz	2273	0.2618	0.4397	0	1
postwz	2273	0.1373	0.3442	0	1
lnrpgdp	2273	10.3377	0.7499	7.8746	12.5179
lnrpgdp2	2273	107.4311	15.5625	62.01	156.6986
lnrinv	2273	14.4954	0.9922	12.1089	17.3689
fin	2273	2.4258	1.5051	0.6488	18.025
pod	2273	0.0212	0.0312	0.0003	0.2759
alltubel	2273	0.2959	0.6717	0.0007	8.5463
lnrain	2273	6.5636	0.4436	5.035	7.6545
lntemp	2273	2.5487	0.4387	0.2123	3.2456
wp	2273	96.3452	6.9021	46.54	107.15
lnrecycle	2273	8.7956	2.5112	0	13.5524
lnrecycle_ind	2265	8.6844	2.5601	0	13.5440
lnallcapacity	2273	4.0585	1.063	1.2754	8.0706
lnncreate	2273	0.9817	1.966	−4.1819	7.5823
info	2273	0.0498	0.0647	0.0005	0.7736
lnallloss	1405	7.0377	1.354	0	10.809

shows the descriptive statistics for each variable data in this work, and some variable settings will be detailed in depth when used later.

5. Regression Results and Analysis

5.1. Benchmark Regression Results and Analysis

The benchmark regression estimation outcomes in this study are displayed in

Table 2. Benchmark regression results.

	(1)	(2)
	lnwaters	lnwaters
treatw × postw	0.236 **	0.266 **
	(2.041)	(2.265)
lnrpgdp		3.561 ***
		(3.359)
lnrpgdp2		−0.179 ***
		(−3.580)
lnrinv		2.944 *
		(1.736)
fin		−0.107 ***
		(−2.780)
pod		0.167
		(0.035)
alltubel		0.276 **
		(2.435)
CITYFE	YES	YES
YEARFE	YES	YES
cons_	6.697 ***	−53.367 **
	(181.116)	(−2.266)
N	2273	2273
adjR2	0.658	0.663

Note: ***, **, and * are significant at the 1%, 5%, and 10% significance levels, respectively, and the t value is in parentheses.

based on Model (1):

Column (1) is the result of controlling the individual fixed effect and time effect without adding any control variables. Column (2) adds the relevant control variables mentioned above on the basis of Column (1). The regression results of the core explanatory variables in each column of the table show that the coefficient of the core explanatory variable treatw × postw is significantly positive, regardless of whether the control variables are added or not, indicating that the National Water-saving City Selection effectively improves the water conservation of the city. At the same time, participation in the selection and receipt of the National Water-saving City title has boosted the regional water conservation by an average of 26.6%, according to the results of Column (2), assuming all other parameters remain constant. The content of Hypothesis 1 in this paper is verified based on the aforesaid analysis.

5.2. Robustness Test

5.2.1. Parallel Trend Test

Referring to the existing literature [41], this paper conducts a parallel trend test on the effect of the policy to ensure the applicability of the difference-in-differences method. The results are shown in Figure 1.

Among them, the policy period, that is, the 0th period, is utilized as a reference, with a 90% confidence interval. Figure 1 shows that the coefficients of the core explanatory variables are not significant before the policy, indicating that there is no obvious gap between the intervention and control groups before the policy, which conforms with the basic assumption of a parallel trend.

5.2.2. Change in Time Point of Policy

In this paper, the time point of policy occurrence is advanced by 2 and 3 periods, respectively, to generate the false core explanatory variables treatw × postw_2 and treatw × postw_3, so that a counterfactual test on the benchmark regression results can be performed by changing the policy time point. Columns (1) and (2) of

Table 3. Change in time point of policy.

	(1)	(2)
	lnwaters	lnwaters
treatw × postw_2	0.224
	(1.627)
treatw × postw_3		0.099
		(0.627)
CONTROL	YES	YES
CITYFE	YES	YES
YEARFE	YES	YES
cons_	−55.201 **	−55.266 **
	(−2.327)	(−2.312)
N	2273	2273
adjR2	0.663	0.662

Note: ** is significant at the 5% significance level, and the t value is in parentheses.

show the regression results:

5.2.3. Variable Adjustment

This part makes the following changes to the variables in the model and regresses them. The remaining variables, with the exception of the key explanatory variables, are winsorized by up or down 1%, and the regression results are displayed in Column (1) of

Table 4. Variable adjustment.

	(1)	(2)	(3)	(4)	(5)
	lnwaters_w	lnwaters	lnwaters	lnwaters	lnwaters
treatw × postw	0.237 **	0.266 **	0.268 **	0.269 **	0.268 **
	(2.130)	(2.276)	(2.276)	(2.293)	(2.285)
CONTROL	YES	YES	YES	YES	YES
CITYFE	YES	YES	YES	YES	YES
YEARFE	YES	YES	YES	YES	YES
cons_	−6.034	−47.536 **	−53.312 **	−51.413 **	−48.474 **
	(−0.367)	(−2.007)	(−2.266)	(−2.195)	(−2.091)
N	2273	2273	2273	2273	2273
adjR2	0.694	0.664	0.663	0.663	0.663

Note: ** is significant at the 5% significance level, and the t value is in parentheses.

. The relevant variables in the benchmark regression control variables that are not logarithmically treated and are logarithmically processed, and the regression results are shown in Table 4 Column (2); considering the utilization and construction of rainwater resources in the region, the annual average temperature and the size of water penetration rate may have an impact on water use and water supply in the region, and the regional water penetration rate, annual average precipitation, and annual average temperature are included in the control variables in turn. The regression results are shown in columns (3) to (5) of Table 4, of which Column (3) is the regression result after adding the water penetration rate, Column (4) includes the annual average precipitation of the region on the basis of the former, and Column (5) considers the annual average temperature of the region on the basis of the former two.

From the above regression results, although the coefficients of each key explanatory variable fluctuate slightly in the above regression results, they are significantly positive, confirming the robustness of the prior results to some extent.

5.2.4. Sample Adjustment

Considering the unbalanced characteristics of the data and the relevant changes in the policy, this paper processes the samples as follows to test the robustness of the previous results. To ensure that the intervention individuals who have been processed, except for the sample period, have one period of data before and after the policy is implemented, this paper adds the total number of times the individual appears in the data at each processing time point, then eliminates the individuals who appear less than six times, and performs regression on this basis. The results are shown in Column (1) of

Table 5. Sample adjustment.

	(1)	(2)	(3)	(4)	(5)	(6)
	lnwaters	lnwaters	lnwaters	lnwaters	lnwaters	lnwaters
treatw × postw	0.230 **	0.269 **	0.267 **	0.277 **	0.270 **	0.273 **
	(1.973)	(2.272)	(2.224)	(2.273)	(2.300)	(2.312)
CONTROL	YES	YES	YES	YES	YES	YES
CITYFE	YES	YES	YES	YES	YES	YES
YEARFE	YES	YES	YES	YES	YES	YES
cons_	−51.228 **	−78.258 ***	−56.705 **	−83.200 ***	−53.184 **	−62.779 **
	(−2.180)	(−3.087)	(−2.311)	(−3.139)	(−2.285)	(−2.519)
N	2190	1985	2142	1854	2292	2179
adjR2	0.657	0.675	0.649	0.660	0.668	0.669

Note: *** and ** are significant at the 1% and 5% significance levels, respectively, and the t value is in parentheses.

. Although the sample period for this article concluded in 2019, the building of the cities that will be evaluated in 2020 should start in 2019. Thus, the inhabitants of the cities assessed in 2020 are eliminated here, and the regression is run once more. The outcomes are displayed in Column (2) of Table 5; according to Yao et al. (2023)’s research, the initial batch of National Water-saving Cities was not truly generated by the evaluation [42], so it was excluded from the sample, and the findings are shown in Table 5 (3); finally, the cities evaluated in 2020 were removed from the first batch of cities at the same time, and the results are shown in Column (4) of Table 5; additionally, when the explained variables are continuously interpolated, this paper eliminates the outliers and their previous sub-items. The previous sub-items are maintained and regressed here to ensure the robustness of the results. Column (5) of Table 5 displays the outcomes. Simultaneously, the amount of urban water saved that has not been interpolated is added as an explanatory variable in the model, and the result is displayed in Column (6) of Table 5. The coefficients of the core explanatory variables in the table are all significantly positive, ensuring the robustness of the benchmark regression results to some extent.

5.2.5. Consideration of the Impact of Relevant Policies

The pilot construction of the first ecological civilization demonstration area was carried out nationwide in 2013, which also put forward the relevant requirements of promoting the water-saving transformation of the first demonstration area and accelerating the construction of local water-saving societies. Simultaneously, the pilot work of financial support for sponge city construction carried out in 2016 further promoted the level of local water culture construction on the basis of promoting rainwater resource discharge control in the pilot area and optimizing the utilization of local rainwater resources. This paper is based on the implementation time of each policy and the pilot cities, and it employs the same setup procedure as the core explanatory variables. The model is being expanded to include the ecological civilization pilot demonstration area (plotece) and the sponge city construction pilot (plotsc). The regression findings are shown in

Table 6. Consideration of the impact of relevant policies.

	(1)	(2)	(3)
	lnwaters	lnwaters	lnwaters
treatw × postw	0.264 **	0.266 **	0.264 **
	(2.239)	(2.259)	(2.238)
plotece	YES		YES
plotsc		YES	YES
CONTROL	YES	YES	YES
CITYFE	YES	YES	YES
YEARFE	YES	YES	YES
cons_	−52.981 **	−52.779 **	−52.520 **
	(−2.251)	(−2.243)	(−2.233)
N	2273	2273	2273
adjR2	0.663	0.663	0.663

Note: ** is significant at the 5% significance level, and the t value is in parentheses.

in Columns (1) through (3), where Columns (1), (2), and (3) are the regression results of the ecological civilization pilot demonstration area, the sponge city construction pilot, and the combination of the two. The coefficient of the main explanatory variable is still significantly positive based on the results of each column, indicating that the results obtained above are not generated from the impact of relevant policies.

5.2.6. Placebo Test

In order to ensure that the policy effect is not affected by other non-observed omission factors, the placebo test is carried out on the benchmark regression results; that is, the intervention group is randomly selected and the time point of intervention is randomly specified, and then, the false core explanatory variables are constructed on this basis, and then, the regression is carried out in turn with the help of Model (1). At the same time, the process is repeated 1000 times. The final probability distribution density diagram of the false coefficient and its p value distribution are shown in Figure 2. It can be seen from the figure that the probability distribution density of the coefficient value of the false core explanatory variable presents a normal distribution form centered on 0, and its p value is almost greater than 0.1 (the horizontal dash line in the Figure 2). At the same time, the real policy effect (0.236) is not in this distribution. Therefore, to a certain extent, it proves the authenticity of the policy effect.

5.2.7. PSM-DID

Finally, this paper computes the individual propensity score based on the average value of all control variables included in the benchmark regression in a given year or years, employs the 1:2 nearest neighbor matching method to match the intervention and control groups, and then estimates the PSM-DID on this basis. The following years or year intervals are chosen by this paper to complete the propensity score calculation and matching: 2006, 2006–2008, 2006–2010, 2006–2012, 2006–2014, 2006–2016, 2006–2018, and 2006–2019. Columns (1) to (8) of

Table 7. PSM-DID.

	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)
	2006	2006–2008	2006–2010	2006–2012	2006–2014	2006–2016	2006–2018	2006–2019
treatw × postw	0.212 *	0.267 **	0.257 **	0.238 **	0.248 **	0.246 **	0.246 **	0.246 **
	(1.791)	(2.260)	(2.185)	(2.039)	(2.117)	(2.102)	(2.106)	(2.101)
CONTROL	YES	YES	YES	YES	YES	YES	YES	YES
CITYFE	YES	YES	YES	YES	YES	YES	YES	YES
YEARFE	YES	YES	YES	YES	YES	YES	YES	YES
cons_	−87.396 ***	−63.388 **	−67.100 ***	−85.095 ***	−75.879 ***	−79.619 ***	−79.573 ***	−79.683 ***
	(−3.039)	(−2.479)	(−2.631)	(−3.349)	(−2.985)	(−3.140)	(−3.131)	(−3.110)
N	1723	2002	2002	1976	1959	1968	1958	1945
adjR2	0.620	0.618	0.614	0.608	0.615	0.621	0.624	0.624

Note: ***, **, and * are significant at the 1%, 5%, and 10% significance levels, respectively, and the t value is in parentheses.

include the regression results. It is clear that, while the change in the number of matches generates a minor fluctuation in the coefficient, it has no effect on the significance level of any result, indicating, to some extent, that the prior study results are robust.

5.3. Mechanism Analysis

Based on the mechanism path suggested above, this work refers to Jiang’s (2022) pertinent remarks on mechanism analysis [43] and completes the following mechanism analysis using Model (3). The overall repeated water consumption (recycle) and industrial repeated water consumption (recycle_ind), the entire production capacity of the urban water supply (allcapacity), and the urban innovation index (ncreate) are among them. The remaining parameters are the same as in Equation (1), and the regression results are shown in

Table 8. Mechanism analysis.

	(1)	(2)	(3)	(4)
	lnrecycle	lnrecycle_ind	lnallcapacity	lnncreate
treatw × postw	0.853 ***	0.916 ***	0.067 **	0.257 ***
	(5.975)	(6.050)	(2.427)	(6.219)
CONTROL	YES	YES	YES	YES
CITYFE	YES	YES	YES	YES
YEARFE	YES	YES	YES	YES
cons_	−6.514	−3.856	4.637	−6.505
	(−0.242)	(−0.147)	(0.883)	(−0.976)
N	2273	2265	2273	2273
adjR2	0.813	0.820	0.943	0.979

Note: *** and ** are significant at the 1% and 5% significance levels, respectively, and the t value is in parentheses.

columns (1) to (4).

m_it = α₂ + β₃treatw_i × postw_t + δX_it +λ_i + μ_t + ε_it

(3)

The core explanatory variable treatw×postw is significantly positive in the regression results of each column in Table 8, indicating that the National Water-saving City Selection improves the urban water conservation by increasing repeated water consumption and enhancing the city’s comprehensive water supply capacity. At the same time, it promotes the improvement in urban innovation on the basis of the above mechanism, establishing a firm platform for regional water-saving technology innovation and promotion, and indirectly boosting the region’s water conservation. So far, this paper’s Hypothesis 2 has been confirmed.

5.4. Heterogeneity Analysis

The possible heterogeneous effects of the policy are evaluated here using grouping regression based on the material described in Section 3. Referring specifically to the research concepts of Chen et al. (2019) [44], the degree of the numerical level is divided by the 25% or 75% quantile. In this paper, the 25% quantile of total urban water supply is used as the boundary, and areas with total water supply less than or equal to this number are considered to have less water supply, while areas with a higher water supply are considered to have more water supply. Similarly, the proportion of ordinary undergraduates and above in the total number of people in the region at the end of the year is used to reflect the number of people receiving higher education in the region, and with the 75% quantile as the boundary, the region with a proportion greater than or equal to this value is regarded as having a relatively large number of people receiving higher education and vice versa. Finally,

Table 9. Heterogeneity analysis.

	Water Supply		High Education Level
	(1)	(2)	(3)	(4)
y: lnwaters	Low	High	High	Low
treatw × postw	0.069	0.306 **	−0.188	0.427 ***
	(0.250)	(2.270)	(−1.082)	(2.897)
CONTROL	YES	YES	YES	YES
CITYFE	YES	YES	YES	YES
YEARFE	YES	YES	YES	YES
cons_	−169.331 **	−21.079	94.337 **	−82.181 ***
	(−2.500)	(−0.847)	(2.160)	(−2.943)
N	562	1698	561	1698
adjR2	0.573	0.586	0.686	0.649

Note: *** and ** are significant at the 1% and 5% significance levels, respectively, and the t value is in parentheses.

columns (1) to (4) provide the regression findings for each group.

It is clear from the results of (1) to (4) in the table that the coefficient of core explanatory factors in areas with more water supply is much higher than in places with less water supply. Simultaneously, when compared to locations with more higher education, the policy has a more noticeable positive influence in areas with less higher education. The aforementioned findings are essentially consistent with the correlation analysis in Section 3 of this study, which validates the substance of Hypothesis 3 in this paper.

5.5. Expanded Analysis

5.5.1. Water-Saving Effect under Double Policies

In order to consider the strengthening effect of smart city construction on urban water-saving construction on the basis of National Water-saving Cities, so as to complete the test of the relevant content of Hypothesis 4 above, this paper conducts a corresponding regression based on Model (2) and presents the results in Column (1) of

Table 10. Water-saving effect under double policies.

	(1)	(2)	(3)	(4)	(5)
	lnwaters	lnwaters	lnwaters	lnwaters	lnwaters
treatwz × postwz	0.334 ***	0.287 ***	0.386 ***	0.462 *	0.424 ***
	(3.268)	(2.886)	(2.906)	(1.919)	(3.333)
treatw×postw		0.135
		(1.200)
CONTROL	YES	YES	YES	YES	YES
CITYFE	YES	YES	YES	YES	YES
YEARFE	YES	YES	YES	YES	YES
cons_	−58.848 **	−57.509 **	−131.013 ***	−168.957 ***	−42.554
	(−2.471)	(−2.435)	(−3.283)	(−3.111)	(−1.059)
N	2,273	2,273	802	367	615
adjR2	0.664	0.664	0.677	0.640	0.688

Note: ***, **, and * are significant at the 1%, 5%, and 10% significance levels, respectively, and the t value is in parentheses.

. Simultaneously, in order to assess the variations in the impact between the two policies and the single policy, this study first incorporates Model (1) and Model (2)’s core explanatory variables into the model and regression on the basis of the entire sample. Table 10 Column (2) represents the results, while Model (4) depicts the specific form. Furthermore, only persons who have received the title of a National Water-saving City are selected and employed as samples for regression based on the relevant practices of the available literature [37,38]. The results are shown in Table 10 Column (3); following that, this paper examines the possible impact of the two policies on the level of urban water saving when the smart city pilot is earlier or later than the National Water-saving City Selection, while still limiting the sample to individual cities, based on the above analysis. The results are listed in Table 10, Column (4) and Column (5), of which Column (4) is the regression result of the smart city pilot earlier than the regional participation and the National Water-saving City title, while Column (5) is the regression result of the smart city pilot later than the participation and the national water-saving city title, both of which exclude individuals whose evaluation and recognition in the same year is the smart city pilot policy.

m_it = α₃ + β₄treatwz_i × postwz_t + β₄treatw_i × postw_t + ηX_it +λ_i + μ_t + ε_it

(4)

Table 10 shows that smart city construction does contribute to a further improvement in urban water conservation, which is more direct in Columns (2) and (3); that is, the introduction of the variable treatw×postw or the sample size is limited to the evaluated city, which does not affect the significance level of the core explanatory variable, the treatwz × postwz coefficient. At the same time, the results of Columns (4) and (5) show that regardless of when the pilot policy occurs, it has played a significant role in boosting the city’s water-saving building. As a result, part of Hypothesis 4 of this study is confirmed.

5.5.2. Mechanism Analysis under the Dual Policy

Based on the analysis of Section 5.5.1, the explanatory variables in Section 5.3 are replaced by the explanatory variables of Model (3) in order to further explore the water-saving effect under the dual policy, based on the relevant setting of Model (2) and the practice of relevant literature [36], and the results are shown in

Table 11. Mechanism analysis under the dual policy I.

	(1)	(2)	(3)	(4)
	lnrecycle	lnrecycle_ind	lnallcapacity	lnncreate
treatwz × postwz	0.604 ***	0.603 ***	0.134 ***	0.099 ***
	(4.828)	(4.685)	(5.782)	(3.377)
CONTROL	YES	YES	YES	YES
CITYFE	YES	YES	YES	YES
YEARFE	YES	YES	YES	YES
cons_	−18.557	−16.180	2.668	−9.163
	(−0.687)	(−0.613)	(0.506)	(−1.355)
N	2273	2265	2273	2273
adjR2	0.812	0.818	0.944	0.979

Note: *** is significant at the 1% significance level, and the t value is in parentheses.

(1) to (4). Simultaneously, in order to compare the mechanism of the dual policy and the single policy effectively, the corresponding regression with the mechanism variable as the explanatory variable is performed under the condition of the entire sample, and the results are displayed in

Table 12. Mechanism analysis under the dual policy II.

	(1)	(2)	(3)	(4)
	lnrecycle	lnrecycle_ind	lnallcapacity	lnncreate
treatwz × postwz	0.360 ***	0.332 ***	0.131 ***	0.010
	(2.998)	(2.703)	(5.451)	(0.302)
treatw × postw	0.688 ***	0.764 ***	0.007	0.253 ***
	(4.916)	(5.135)	(0.251)	(5.424)
CONTROL	YES	YES	YES	YES
CITYFE	YES	YES	YES	YES
YEARFE	YES	YES	YES	YES
cons_	−11.716	−8.607	2.738	−6.648
	(−0.433)	(−0.326)	(0.519)	(−0.989)
N	2273	2265	2273	2273
adjR2	0.814	0.820	0.944	0.979

Note: *** is significant at the 1% significance level, and the t value is in parentheses.

(1) to (4).

Although the results of Table 11 show that the two policies significantly improved urban repeated water consumption, comprehensive water supply capacity, and the urban innovation level, the core explanatory variable, the treatwz × postwz coefficient, only passed the significance level test in the first three columns after the introduction of treatw × postw. As a result of the selection and recognition, the smart city pilot’s new urban infrastructure construction and urban digital management model have increased urban repeated water usage and its comprehensive water supply capacity. However, it has not increased the degree of regional innovation beyond what it is now.

Simultaneously, given that the management and improvement of urban water supply leakage require the strong support of monitoring and management based on modern information technology [45], smart city construction requirements not only put forward the use of information technology to realize the whole process supervision of the water supply process, but also pointed out that the digital comprehensive management and monitoring of an urban underground pipeline network should be realized. As a result, under smart city pilot construction, it may achieve effective regional leakage water management on the basis of existing water-saving construction and then compensate for the weaknesses of water-saving construction. Based on the preceding research, this paper first investigates the influence of National Water-saving City Selection on urban leakage (allloss) using Model (1). Column (1) of

Table 13. Mechanism analysis under the dual policy III.

	(1)	(2)	(3)	(4)	(5)
	lnallloss	lnallloss	info	info	lnallloss
treatwz × postwz				0.007 **	−0.171 ***
				(2.513)	(−2.805)
treatw × postw	−0.071	0.039	−0.003
	(−0.730)	(0.340)	(−1.576)
treatw × postw × info		−2.531 *
		(−1.831)
info		2.469 *
		(1.716)
CONTROL	YES	YES	YES	YES	YES
CITYFE	YES	YES	YES	YES	YES
YEARFE	YES	YES	YES	YES	YES
cons_	−19.722	−18.614	0.022	−0.045	−16.408
	(−1.141)	(−1.080)	(0.060)	(−0.127)	(−0.945)
N	1401	1401	2273	2273	1401
adjR2	0.867	0.867	0.907	0.907	0.867

Note: ***, **, and * are significant at the 1%, 5%, and 10% significance levels, respectively, and the t value is in parentheses.

shows the results. Then, on the basis of Model (1), this paper further introduces the interaction term, the next term of the core explanatory variable, and the urban digital infrastructure construction index and makes a regression. The digital infrastructure construction index (info) of the regional digital information infrastructure construction level draws on the practice of relevant literature [46] and takes into account the availability of relevant data of prefecture-level cities. Based on the total number of urban telecommunications services, the number of mobile phone users, and the number of Internet users, the urban digital infrastructure construction index is calculated by using entropy weighting and summation. The specific setting of the model is shown in Model (5), and the results are shown in Column (2) of Table 13. Next, in Column (3) of Table 13, the explanatory variable of Model (1) is substituted for the digital infrastructure construction index to examine the potential influence of the selection activities on the regional digital infrastructure. Finally, the digital infrastructure construction index and urban water leakage are employed as explanatory variables in Model (2) to test the influence of the dual policy on regional digital infrastructure construction and urban water supply leakage. Table 13 Columns (4) and (5) reveal the results.

m_it = α₄ + β₅treatw_i × postw_t × info + β₆treatw_i × postw_t + β₇info + ξX_it +λ_i + μ_t + ε_it

(5)

The results of Column (1) in Table 13 show that the selection of National Water-saving Cities had no significant negative impact on the amount of regional leakage water, whereas the coefficient of interaction term treatw×postw×info in Column (2) is significantly negative, indicating that improving the level of digital infrastructure construction is beneficial to improving the level of regional leakage water management in the water-saving process. Second, based on the results of Column (3), the selection had no substantial positive impact on the creation of regional digital infrastructure since it did not present clear prerequisites for the construction of a regional digital government. However, based on the results of Column (4) and (5), implementing a smart city pilot on the basis of the selection will improve the level of digital infrastructure construction in the region, effectively improving the governance effect of regional water supply leakage, thus reducing the amount of regional water supply leakage. At this time, the entire substance of Hypothesis 4 in this study has been validated.

6. Conclusions

This paper examines the impact and mechanism of the National Water-saving City Selection on urban water conservation using the urban unbalanced panel data from 2006 to 2019 and the multi-time point difference-in-differences method. It then considers the potential strengthening impact of urban digital management construction brought about by smart city construction on urban water-saving construction. The analysis presented above led this research to the following conclusions: (1) the National Water-saving City Selection has significantly improved the water conservation of the evaluated cities, and the results show that participating in the selection and obtaining the National Water-saving City title has increased the regional water conservation by 26.6% on average. At the same time, the benchmark results are still valid after a series of robustness tests. (2) The selection actions not only directly increase regional water conservation by increasing regional repeated water consumption and comprehensive water supply capacity but also indirectly aid urban water-saving construction by activating urban innovation kinetic energy. (3) The policy has resulted in more visible water-saving efficiency in locations with a greater water supply or less higher education. (4) Based on the selection of National Water-saving Cities and the assessment of regional smart city pilots, it is discovered that the growth of smart cities has greatly improved the region’s existing water conservation, with the effect being more visible in direct channels. Simultaneously, the digital management construction enabled by it adds to the management of urban water supply leaks, thus filling the short board of water-saving construction.

Based on the above results, on the basis of the existing literature affirming the water-saving effect of social water-saving pilot construction and market-based water-saving regulation [17,18,21,22], the conclusions obtained in this paper further affirm the effectiveness of China’s National Water-saving City Selection in water conservation to a certain extent. On the basis of the research on the effect of the evaluation and recognition policies [8,9,13], the conclusions above further prove the positive impact of the evaluation and recognition policies guided by the central government on improving the enthusiasm of local governments to accelerate the construction of resource conservation and weakening the “race to the bottom” phenomenon of the local government environment and resource regulation between neighboring places. At the same time, this paper takes the pilot of the smart city into consideration in the expansion part and then considers the complementary role of digital city construction on the basis of the National Water-saving City Selection, so as to further verify the synergistic effect of the two policies to a certain extent.

The above results also provide the following policy implications: first of all, they give full play to the positive impact of evaluation and recognition policies on environmental governance and resource management. We can not only improve the selection system, but expand the selection content, strengthen the qualification review, and adhere to the post-evaluation review to ensure that the selection activities are implemented. At the same time, we can also use extensive publicity and effective incentives to improve the initiative and enthusiasm of local governments and urban residents to participate and form a benign interaction of “promoting governance by evaluation”, so as to make up for the inherent deficiencies in traditional command control and the deficiencies in market-based regulation. Secondly, we should effectively play a positive role in digital information technology in resource management and utilization, improve the construction of new urban infrastructure in the construction of a smart city, and realize the whole process supervision and whole process monitoring of regional resource management and utilization with the help of a digital management platform, so as to ensure the effective and efficient use of various resources. Thirdly, from the national level, although it has not been strictly demonstrated, considering the different levels of economic development and political systems of various countries, the evaluation and recognition policies may not be able to bring similar positive effects to different countries. However, from the mechanism analysis, heterogeneity analysis, and expanded analysis results of this paper, it can be seen that the positive impact of technological upgrading, talent training, and digital transformation of regional management on resource utilization cannot be ignored. Therefore, different countries may consider making efforts from the above levels to improve their water conservation.

Of course, in addition to the above analysis and demonstration, due to the availability of data and the feasibility of calculation methods, there are still some deficiencies in this paper; that is, the possible impact of China’s National Water-saving City Selection on water-saving sustainability and the optimal utilization of water resources have not been evaluated in-depth using an ecological footprint or water footprint. In this regard, relevant scholars may further discuss and analyze it.