Integrating the Theory of Planned Behavior and Motivation to Explore Residential Water-Saving Behaviors

Abstract

Water scarcity in the United States needs to be addressed with demand- and supply-side initiatives. Yet, water is often provided for a low cost to households in residential areas that have the potential to reduce water consumption. The theory of planned behavior (TpB) is a social science theory used to understand why volitional conservation behaviors, including water conservation, are performed by consumers. Additional predictors are included in TpB in water literature, and motivations for behavioral actions may help further explain residential water conservation behavior. This study sought to determine whether TpB and motivation predicted residential water conservation behavior. Data were collected with an online survey (n = 907) in September 2022 from residents of Florida, Georgia, and Alabama using non-probability opt-in quota sampling. The data were analyzed in R using the lavaan package. The results indicate that there was a significant direct effect of attitude and subjective norms on intention. There was a significant direct effect of intention on behavior. There was a significant negative direct effect of extrinsic motivation on behavior. Mediation was present in the model with intrinsic motivation. Perhaps the respondents had not thought about rewards or punishments in the context of water conservation because of the availability of water at a low cost and therefore were not extrinsically motivated to conserve water.

1. Introduction

One of the biggest threats to humankind in the 21st century is water scarcity [1,2], which is predicted to worsen over time because of the growing human population and impacts of climate change [3]. Water scarcity restricts sustainable economic development, future food security, and natural systems [1,2], posing serious challenges for policy makers as they consider human health and fragile ecosystems [4]. Water is often available in private residences at a low cost in the United States (U.S.), with the severity of water scarcity unknown to consumers [4,5]. Yet, groundwater supplies are over-extracted, and there are inadequate water flows in many major river systems [4].

Demand-side approaches, such as household residential interventions, and supply-side approaches, such as improved infrastructure and management, are needed to address water scarcity [3,6]. Households in residential areas have the potential to lessen overall water consumption [5]. Social practices and habitual routines influence the demand for water [7,8]. Most residential water in the U.S. is used for hygiene purposes or outdoor landscapes [5]. In addition, increasing affluence poses challenges to water-saving behaviors [5]. Additional studies are necessary to understand the determinants of water-saving behavior in order to increase voluntary reductions in water use in the home [1].

Residential households that have some capacity to voluntarily conserve water are often unaware of the importance of water conservation or are not knowledgeable about the best ways to reduce water consumption [9]. Often, drought or other triggers like economic incentives initiate water conservation behavior changes, yet the adoption of conservation practices is needed prior to extreme drought conditions [9]. Sharing information about water conservation through communication tools may be an effective approach for targeting behavior, but researchers must first determine ways to target and frame messages to reach specific populations by understanding the drivers of water conservation in residential households [9,10], which are often situational and complex [11].

2. Framework

2.1. The Theory of Planned Behavior

The theory of planned behavior (TpB) is particularly useful in persuasive communication, where individuals are required to centrally process or think about a behavior that they are performing [12]. TpB posits that behavioral intentions are based on attitude, subjective norms, and perceived behavioral control towards a specific behavior [13]. Behavioral intentions are considered the strongest predictor of behavior [13].

Attitude towards a behavior is an individual’s evaluation of the positive or negative outcomes of performing a specific behavior [12,13]. An individual’s subjective norms are their beliefs about whether performing a specific behavior would be approved by people who are important to them [12,13]. Gibson et al. (2021) found that increasing consumers’ subjective norms towards water conservation is important for increasing their intention to engage in water conservation practices in and around the home [14]. Perceived behavioral control is an individual’s perception of how challenging it is to perform a specific behavior and whether they have control over making the decision to perform that behavior [12,13]. When an individual has high self-efficacy and high autonomy towards carrying out a specific behavior, the behavior is more likely to occur than in those who lack autonomy or self-efficacy [12,13].

TpB is frequently used to understand why individuals perform behaviors specific to conservation, e.g., [15,16,17,18], including water conservation. Often, water conservation studies using TpB seek to understand the intention to conserve water by including additional variables to extend TpB, e.g., [1,14,19,20]. For example, Chaudhary et al. (2017) used TpB to examine the effects of attitudes, subjective norms, perceived behavioral control, personal norms, demographic factors, and past behaviors on Florida home landscape irrigation users’ intention to use water-saving irrigation practices [19]. The authors found that all variables other than attitude significantly predicted intention to use water-saving irrigation practices [19]. The extended model predicted 39% of the variance in intention to use water-saving irrigation practices, whereas the foundational variables in the TpB model (attitude, subjective norms, and perceived behavioral control) only predicted 25.1% of the variance [19].

Warner and Diaz (2021) used TpB to explore the effects of attitudes, subjective norms, perceived behavioral control, and connectedness to water on behavioral intent related to landscape irrigation [20]. The authors found that subjective norms, perceived behavioral control, and connectedness to water were positive predictors of behavioral intent related to landscape irrigation [20]. Attitude did not significantly predict behavioral intention related to landscape irrigation [20]. The model predicted 19.6% of the variance in behavioral intent related to landscape irrigation as compared to 15.1% when using only the core TpB variables [20].

Si et al. (2022) used TpB by extending attitude, subjective norms, and perceived behavioral control to include environmental concern, perceived risk, and information publicity to predict general water-saving intention among residents in Jinan, China, who were aware of the city’s water resources and policy [1]. The authors found that attitude and information publicity were the strongest direct predictors of water-saving intentions [1]. Subjective norms and perceived behavioral control also had significant direct effects on water-saving intentions, and environmental concerns and perceived risk had significant indirect effects on water-saving intentions [1]. The model predicted 81.8% of the variance in water-saving intentions, which was greater than the variance explained by the core TpB variables [1]. Adding additional variables explained additional variance in intention, suggesting that factors related to the study context should be added to TpB.

TpB with intervening factors in the context of residential water conservation is presented in Figure 1. Stern (2018) explained that intervening factors may prevent an individual from performing a behavior, even if their behavioral intention is high (see Figure 1) [12,13]. For example, there may be a financial barrier that prevents them from installing low-flow shower heads or other related water-saving technology [12]. Similarly, they may lack the skills to carry out a behavior, such as planting native plants instead of higher-water-consuming plants [12]. TpB is “open to the inclusion of additional predictors if it can be shown that they capture a significant proportion of the variance in intention or behavior after the theory’s current variables have been taken into account” ([13], p. 199). Although extended models of TpB include additional variables, e.g., [1,19,20], these variables are added to the models to directly or indirectly predict intention rather than to act as intervening factors between intention and behavior, as shown in Figure 1. Including intervening variables as mediators between intention and behavior helps explain the relationship between intention and behavior. Motivation may act as an intervening factor between intention and behavior.

2.2. Motivation for Behavioral Action

Intrinsic and extrinsic motivation are often used to explain the motivation for behavioral actions. Individuals perform actions intrinsically when they are driven by personal interests and the gratification of performing the action [12,21]. The non-tangible rewards of intrinsic actions are increased interest, excitement, and satisfaction about an action, which may later lead to greater performance, persistence, and creativity [12]. Extrinsic motivation arises from external pressure, incentives, or coercion whereby individuals perform an action to attain some external outcome [12,21,22]. A wide range of responses are possible when an action is performed extrinsically, such as reluctance, passive compliance, or personal commitment [12]. Intrinsic and extrinsic motivation do not belong on a continuum and therefore do not have a low–high relationship with one another. It is possible that an action reflects both intrinsic and extrinsic motivation [22]. Intrinsic and extrinsic motivation are operationalized as components of self-determination theory in the present study [23,24,25,26].

Studies have examined intrinsic and extrinsic motivation in the context of sustainability and conservation intention. For example, Li and Wen (2019) explored the effects of extrinsic and intrinsic motivation on respondents’ collaborative consumption intention, specially related to a bicycle-sharing program in 22 proveniences in China [27]. Intrinsic motivation (measured using sustainability, the sense of belonging, trust, and enjoyment) and extrinsic motivation (measured using economic benefits, convenience, and perceived usefulness) predicted 57.7% of the variance in intention to participate in the bicycle-sharing program [27]. Extrinsic motivation was a stronger predictor than intrinsic motivation in intention to participate in the bicycle-sharing program [27].

Ryan et al. (2003) examined factors that motivate farm operators’ adoption of conservation practices along riparian zones in south-eastern Michigan in the River Raisin watershed [28]. The farmers’ adoption of specific conservation practices was found to be intrinsically motived [28]. Economic motivation (extrinsic) did not play a role in the adoption of specific conservation practices [28].

Fatoki (2022) examined energy-saving behaviors among hotel employees using environmental self-identity and intrinsic motivation in the context of self-determination theory [29]. Intrinsic motivation was separated into two dimensions—obligation-based intrinsic motivation and enjoyment-based intrinsic motivation—and these were treated as mediating variables between environmental self-identity and energy-saving behavior [29]. The authors found that both dimensions of intrinsic motivation had a direct effect on energy-saving behavior [29]. In addition, the relationship between environmental self-identity and energy-saving behaviors was mediated by both dimensions of intrinsic motivation [29].

To the best of the authors’ knowledge, extrinsic motivation in relation to self-determination theory has not been widely examined in water literature. However, studies have examined the effects of extrinsic motivators on water consumption. For example, Chappells et al. (2011) examined the social impact of a housepipe ban, where water utilities could turnoff water if a household used too much, which is comparable to extrinsic motivation [7]. The authors found that, although the housepipe ban brought visibility to unsustainable garden practices, the long-term impacts are not straightforward [7]. For example, some participants successfully reduced water consumption, whereas others altered their watering practices, but the new method did not always translate to a reduction in water [7]. Fielding et al. (2012) explored water use determinants in households in Queensland, Australia, to inform campaigns for demand management [30]. The authors found that households used less water when they were in regions that had recently experienced drought conditions and water restrictions, which is comparable to extrinsic motivation [30].

There are a few notable problems with extrinsic motivation for behavioral performance [31]. First, permanent changes are not produced by extrinsic rewards [31]. It is possible, however, to promote a behavior with a reward (extrinsic motivation) and gradually take away the reward as the interest, excitement, and satisfaction in performing the behavior (intrinsic motivation) become more apparent [22].

Second, intrinsic motivation can be crowded out by extrinsic motivations when increased incentives decrease motivation [12,31]. This problem is referred to as motivation crowding, and it is described in motivation crowding theory [12]. For example, providing small farms money to protect water (extrinsic) may crowd out cultural beliefs (intrinsic) to conserve water, and when the payment to conserve water ceases, the initial intrinsic motivation may have been replaced and may be difficult to recover [12]. However, external incentives can benefit intrinsic motivation in specific cases where a barrier, which would otherwise not allow an intrinsically motivated individual to perform a task, is removed [12]. Third, rewards can be controlling because individuals may be tempted by the reward rather than the task itself, so the behavior is controlled by the extrinsic reward [31].

2.3. Motivation as an Intervening Factor

Previous research has found intrinsic motivation and extrinsic motivation to be predictors of intention, e.g., [27,28]. To the best of the authors’ knowledge, no studies have assessed the effects of intrinsic motivation and extrinsic motivation on water conservation behavior—as opposed to only intention. Although an individual may have an intention to perform an action based on their attitude, subjective norms, and perceived behavioral control as posited in TpB, they may not have the motivation to perform the behavior. Without the motivation to perform a behavior, it is unlikely that the behavior will occur. Thus, intrinsic motivation and extrinsic motivation may act as intervening variables that add to the relationship of intention with behavior. Tabernero and Hernández (2010) examined the mediating role of intrinsic motivation on the relationship between self-efficacy and environmentally responsible behavior [32]. The authors found a significant mediation effect with intrinsic motivation [32]. Intrinsic motivation and extrinsic motivation have not been examined as mediating variables between intention and behavior in TpB in the environmental literature (see Figure 2). This study seeks to fill that gap.

This study sought to examine TpB with intrinsic motivation and extrinsic motivation as mediating variables in the context of water conservation behaviors in residential areas to help build social science theory. The following objective and hypotheses were used to guide the study:

Objective 1: Describe respondents’ attitude, subjective norms, perceived behavioral control, intrinsic motivation, and extrinsic motivation towards residential conservation intention.

Hypothesis 1. 

Attitude towards residential water conservation will have a direct effect on residential conservation intention.

Hypothesis 2. 

Subjective norms towards residential water conservation will have a direct effect on residential conservation intention.

Hypothesis 3. 

Perceived behavioral control towards residential water conservation will have a direct effect on residential conservation intention.

Hypothesis 4. 

Residential conservation intention will have a direct effect on residential conservation behavior.

Hypothesis 5. 

Intrinsic motivation towards residential water conservation will mediate the relationship between residential conservation intention and residential conservation behavior.

Hypothesis 6. 

Extrinsic motivation towards residential water conservation will mediate the relationship between residential conservation intention and residential conservation behavior.

Hypothesis 7. 

Attitude, subjective norms, and perceived behavioral control towards residential water conservation will have an indirect effect on residential conservation behavior.

3. Materials and Methods

3.1. Participants

A survey administered online via Qualtrics was used to collect data from residents of Florida, Georgia, and Alabama in September 2022. Respondents represented the population based on race and gender in the 2020 U.S. Census. No other quotas were included. Respondents were recruited with non-probability opt-in sampling. Research on public opinions frequently uses non-probability opt-in sampling for data collection, but limitations like sampling bias exist [33]. Examples include limiting participants who do not have internet access or the nature of online surveys to attract specific types of people [33].

Several faculty members at various universities with expertise in survey design, educational research, and water resources reviewed the instrument for content accuracy and face validity. The faculty members were external to the author team and in departments of horticulture and advertising and public relations housed in an external university. Content validity is the extent to which an instrument comprehensively represents the phenomenon that it is supposed to measure [34]. Face validity examines a surface-level view of whether the instrument measures what it is supposed to measure [34]. Qualtrics performs quality checks on the data for speeders, possible bots, duplicate responses, and completion rate. In addition, two attention checks were included at the start of the survey and halfway through the survey, where respondents were promoted to select a specific response. The research design was approved by the University of Georgia Institutional Review Board (IRB #00005553). The survey was piloted with 50 individuals prior to full data collection.

3.2. Study Context

Florida, Georgia, and Alabama were selected as the target states because of the ongoing dispute over two river basins that are shared: the Apalachicola–Chattahoochee–Flint and the Alabama–Coosa–Tallapoosa River Basins [35]. The two river basins supply water needs for metro Atlanta’s growing population, which caused Alabama to sue the U.S. Army Corps of Engineers in 1990 for providing metro Atlanta with additional water. Since then, there have been many other significant developments between the three states related to the river basins. Most recently there was conflict over the equitable apportionment of the water between Georgia and Florida. According to Florida, the substantial use of water by Georgia has caused environmental damage and a collapse in the Apalachicola Bay oyster fishery because of increased salinity due to a lack of freshwater flow. The Supreme Court, however, ruled in favor of Georgia. The unique geographical and political context of this study adds to the literature.

3.3. Measures

The study questionnaire can be found in Supplementary S1. A five-point semantic differential scale from Gibson et al. (2021) was used to measure respondents’ attitudes towards residential water conservation [14]. Respondents were asked to indicate their perception of “engaging in everyday actions to save water around their home and in their home landscape” between two adjectives for seven items ([14], p. 5). The sets of adjectives included “bad (1)/good (5), harmful (1)/beneficial (5), worthless (1)/valuable (5), unpleasant (1)/pleasant (5), not acceptable (1)/acceptable (5), foolish (1)/wise (5), and not essential (1)/essential (5)” ([14], p. 5). Scale reliability was calculated using Cronbach Alpha (α = 0.89).

A five-point Likert scale from Gibson et al. (2021) was used to measure respondents’ subjective norms towards water conservation (1 = Strongly Disagree; 2 = Disagree; 3 = Neither Agree nor Disagree; 4 = Agree; 5 = Strongly Agree) [14]. Respondents were asked to indicate their level of agreement or disagreement with participating in six activities. Scale reliability was calculated using Cronbach Alpha (α = 0.86). The items included “I am expected to I save water around the house and in my home landscape, I feel like there is a social pressure to save water around the house and in my home landscape, the people who are important to me want me to save water around the house and in my home landscape, my neighbors would approve of me saving water around the house and in my home landscape, most people in my life whose opinions I value would approve of me saving water around the house and in my home landscape, and the people that I am close to would approve if I explored ways to reduce my water use around the house and in my home landscape” ([14], p. 5).

A five-point Likert scale from Gibson et al. (2021) was used to measure respondents’ perceived behavioral control towards water conservation (1 = Strongly Disagree; 2 = Disagree; 3 = Neither Agree nor Disagree; 4 = Agree; 5 = Strongly Agree) [14]. Respondents were asked to indicate their level of agreement or disagreement with participating in five activities. Scale reliability was calculated using Cronbach Alpha (α = 0.88). The items included “I am confident that I could save water around the house and in my home landscape if I wanted to, the decision to save water around the house and in my home landscape is in my control, whether or not I save water around the house and in my home landscape is entirely up to me, I am certain that I could save water around the house and in my home landscape if I wanted to, and I have complete control over the decision to save water around the house and in my home landscape” ([14], p. 5).

A five-point Likert scale from Gibson et al. (in review) was used to measure respondents’ self-reported residential conservation intention (1 = Very Unlikely; 2 = Unlikely; 3 = Undecided; 4 = Likely; 5 = Very Likely) [36]. The items were initially adapted from Gibson et al. (2021) and Owens and Lamm (2017) [14,37]. Respondents were asked to indicate how likely or unlikely they were to participate in five activities in the future. Respondents had the option to select not applicable if an item did not apply to them. For example, not applicable would be selected by a respondent if they did not own a dishwasher. A mean score was calculated for respondents who selected not applicable based on the number of items that they answered. Scale reliability was calculated using Cronbach Alpha (α = 0.81). The items included “Only run the washing machine when it is full, only run the dishwasher when it is full, only water your lawn in the morning or evening, reduce the number of times a week you water your lawn, and sweep patios and sidewalks instead of hosing them down” ([14], p. 6).

A five-point Likert scale adapted from Owens and Lamm (2017) was used to measure respondents’ self-reported residential conservation behavior (1 = Very Unlikely; 2 = Unlikely; 3 = Undecided; 4 = Likely; 5 = Very Likely) [37]. Respondents were asked to indicate how often they participated in five activities. Respondents had the option to select not applicable if an item did not apply to them. A mean score was calculated for respondents who selected not applicable based on the number of items that they answered. Scale reliability was calculated using Cronbach Alpha (α = 0.68). The items included “I let my sprinklers run when it has rained or is raining (recoded [RC]), I let my sprinklers run when rain is predicted in the forecast (RC), I only run the washing machine when it is full, I only run the dishwasher when it is full, and I hose down my driveway (RC)” ([37], p. 61).

A five-point Likert scale adapted from the Center for Self-Determination Theory (n.d.) and guided by Deci and Ryan (2000) was used to measure respondents’ intrinsic motivation towards water conservation (1 = Strongly Disagree; 2 = Disagree; 3 = Neither Agree nor Disagree; 4 = Agree; 5 = Strongly Agree) [23,38]. Respondents were asked to indicate their level of agreement or disagreement with participating in seven activities. Scale reliability was calculated using Cronbach Alpha (α = 0.87). The items included I enjoy conserving water; conserving water is fun; conserving water is a boring task (RC); I do not pay attention to conserving water (RC); conserving water is very interesting; conserving water is quite enjoyable; and while conserving water, I think about how much I enjoy it.

A five-point Likert scale from Tabernero and Hernández (2010) and guided by Deci and Ryan (2000) was used to measure respondents’ extrinsic motivation towards water conservation (1 = Strongly Disagree; 2 = Disagree; 3 = Neither Agree nor Disagree; 4 = Agree; 5 = Strongly Agree) [23,32]. Respondents were asked to indicate their level of agreement or disagreement for three statements about why they conserve water. Scale reliability was calculated using Cronbach Alpha (α = 0.88). The items included receiving a reward, avoiding a penalty, and gaining social acceptance.

All measures in this study were self-reported by the respondents. Social sciences literature frequently uses self-reported measures, although they are not without limitations. For example, one study found discrepancies in participants’ self-reported water consumption and their actual water consumption, including underestimating outdoor consumption and overestimating indoor consumption [39]. There is limited literature that investigates both self-reported and actual measures of water conservation, likely due to time and economic barriers.

3.4. Demographics

Responses were collected from 907 respondents. The majority of the respondents self-reported being white (76.5%), having at least a 2-year college degree (52.6%), and living within the city or town limits. Demographics are reported in

Table 1. Demographics of Respondents (n = 907).

	Total (n = 907)		Florida (n = 309)		Georgia (n = 311)		Alabama (n = 287)
Baseline Characteristic	F	%	F	%	F	%	F	%
Sex
Female	483	53.3	194	62.8	133	42.8	156	54.4
Male	424	46.7	115	37.2	178	57.2	131	45.6
Age
18–34 years	206	22.6	81	26.2	66	21.2	59	20.6
35–54 years	287	31.6	92	29.8	101	32.5	94	32.8
55+ years	414	45.8	136	44.0	144	46.3	134	46.7
Race *
White	694	76.5	212	68.6	232	74.6	250	87.1
Black	130	14.3	56	18.1	49	15.8	25	8.7
Asian	55	6.1	30	9.7	20	6.4	5	1.7
American Indian or	21	2.3	10	3.2	3	1.0	8	2.8
Alaska Native
Other	42	4.6	23	7.4	12	3.9	7	2.4
Ethnicity
Hispanic	153	16.9	111	35.9	25	8.0	17	5.9
Non-Hispanic	754	83.1	198	64.1	286	92.0	270	94.1
Family Income Level
Less than USD 24,999	195	21.5	62	20.1	58	18.6	75	26.1
USD 25,000–USD 49,999	237	26.1	86	27.8	71	22.8	80	27.9
USD 50,0000–USD 74,9999	191	21.1	67	21.7	61	19.6	63	22.0
USD 75,000–USD 149,999	218	24.0	77	24.9	85	27.3	56	19.5
USD 150,000–USD 249,999	48	5.3	10	3.2	28	9.0	10	3.5
USD 250,000 or more	18	2.0	7	2.3	8	2.6	3	1.0
Home Ownership
Rent	577	63.6	116	37.5	97	31.2	93	32.4
Own	306	33.7	184	59.5	208	66.9	185	64.5
Other	24	2.6	9	2.9	6	1.9	9	3.1
Education
Less than 12th grade	27	3	7	2.3	5	1.6	15	5.2
High school diploma	197	21.7	68	22	70	22.5	59	20.6
Some college	206	22.7	57	18.4	59	19	90	31.4
2-year college degree	116	12.8	47	15.2	37	11.9	32	11.1
4-year college degree	232	25.6	84	27.2	82	26.4	66	23
Graduate or professional degree	129	14.2	46	14.9	58	18.6	25	8.7
Rurality
Within city or town limits	596	65.7	232	75.1	183	58.8	181	63.1
Outside city or town limits	311	34.3	77	24.9	128	41.2	106	36.9

Note: * race was a check-all-that-apply item.

.

In the 2020 U.S. census for Florida, most reported as white (76.8%) and half as female (50.8%). In the 2020 U.S. census for Georgia, a little over half reported as white (59%) and half as female (51.2%). In the 2020 U.S. census for Alabama, most reported as white (68.9%), and around half reported as female (51.4%). One study limitation is that only quotas for over 18 years of age, race, and gender were included in the data collection. Other demographics, like age or education level, may not meet census quotas, and, therefore, the study sample may not be representative of the census population across all demographic characteristics.

3.5. Data Analysis

3.5.1. CFA

Data were analyzed using R Studio with the lavaan package. The item–factor structure was determined using a confirmatory factor analysis (CFA), with items having standardized factor loadings below 0.50 removed from the model [40]. Based on this criterion, one item was removed from subjective norms (“I feel like there is a social pressure to save water around the house and in my home landscape”, λ = 0.44; [14], p. 5), and two items were removed from conservation behavior (“I only run the washing machine when it is full”, λ = 0.17; “I only run the dishwasher when it is full”, λ = 0.19) for the final measurement model [37], p. 61. After removing the items, CFA model fit indices were assessed using the comparative fit index (CFI = 0.91), the Tucker–Lewis Index (TLI = 0.90), and the root mean square error of approximation (RMSEA = 0.06). The model fit was deemed acceptable based on existing model fit criteria in the literature: TLI and CFI > 0.90, and RMSEA < 0.07 [41,42]. Chi-squared (χ²(483) = 2004.47, p < 0.000) was not used as a measure of model fit because of the influence of large sample sizes on the results [43], but it is reported here as suggested by Barrett et al. (2007) [44]. Additional information about model fit indices is reported in the Section 5.1.

The Sobel test was conducted as part of the CFA to examine whether the mediation of intention on behavior through intrinsic motivation and extrinsic motivation was present in the model [45]. There was a significant indirect effect (standardized = −0.05, p < 0.01) of intrinsic motivation. The indirect effect of extrinsic motivation (standardized = −0.00, p = 0.93) indicated that mediation was not present. Therefore, intrinsic motivation remained a mediator in the model between intention and behavior, but extrinsic motivation was only examined as a direct effect on behavior.

3.5.2. Reliability and Validity

The index reliability of each construct in the final measurement model was calculated, and all exceeded 0.70. The measurement model constructs included attitude (α = 0.89), subjective norms (α = 0.87), perceived behavioral control (α = 0.88), residential conservation intention (α = 0.81), residential conservation behavior (α = 0.79), intrinsic motivation (α = 0.90), and extrinsic motivation (α = 0.88).

Convergent validity and discriminant validity [40] were analyzed using R Studio with the semTools package. Convergent validity was examined for the structural model and deemed acceptable if the average variance extracted (AVE) exceeded 0.50 or if the AVE exceeded 0.40 and the corresponding composite reliability exceeded 0.60 (see

Table 2. Constructs’ convergent validity.

Construct	AVE	Composite Reliability
Attitudes	0.55	0.90
Subjective norms	0.57	0.86
Perceived behavioral control	0.60	0.88
Intrinsic motivation	0.65	0.90
Extrinsic motivation	0.70	0.87
Residential conservation intention	0.46	0.80
Behavior	0.56	0.80

) [40]. Discriminant validity was examined for the structural model with the Fornell and Larcker criterion, examining whether factor correlations were less than the square root of their corresponding AVE (see

Table 3. Constructs’ discriminant validity using the Fornell and Larcker criterion.

Construct	Attitudes	Subjective Norms	Perceived Behavioral Control	Intrinsic Motivation	Extrinsic Motivation	Residential Conservation Intention	Behavior
Attitudes	0.55	-	-	-	-	-	-
Subjective norms	0.43	0.57	-	-	-	-	-
Perceived behavioral control	0.42	0.49	0.61	-	-	-	-
Intrinsic motivation	0.19	0.46	0.13	0.65	-	-	-
Extrinsic motivation	0.00	0.07	0.00	0.39	0.70	-	-
Residential conservation intention	0.49	0.18	0.32	0.40	0.00	0.46	-
Behavior	0.08	0.00	0.05	−0.23	−0.52	0.15	0.56

Note: numbers on the diagonal are the square root of their corresponding AVE.

) [40].

3.5.3. Structural Model

The final structural model fit indices were assessed using the comparative fit index (CFI = 0.91), the Tucker–Lewis Index (TLI = 0.90), and the root mean square error of approximation (RMSEA = 0.06). The model fit was deemed acceptable [41,42]. Again, chi-squared (χ²(484) = 2004.48, p < 0.000) was not used as a measure of model fit.

Direct and indirect effects were assessed in the model. Direct effects are the paths from an exogenous variable (independent) to an endogenous variable (dependent). For example, there is a direct effect of attitudes on intention in the structural model. Indirect effects are the pathways from an exogenous variable to an endogenous variable through a mediator. For example, there is an indirect effect of intention on behavior with intrinsic motivation as the mediator. The results present the indirect effects of the variables. For example, attitudes have an indirect effect on behaviors through intention and through intention and intrinsic motivation. The total indirect effect is the indirect effects added together.

4. Results

4.1. Descriptive Statistics

The means, standard deviations, skew, and kurtosis for attitudes, subjective norms, perceived behavioral control, intrinsic motivation, extrinsic motivation, residential conservation intention, and residential conservation behavior are reported in

Table 4. Descriptive statistics for attitudes towards residential water conservation.

	Mean	SD	Skew	Kurtosis
Bad: good	4.54	0.83	−2.10	4.58
Harmful: beneficial	4.53	0.84	−2.03	4.16
Worthless: valuable	4.45	0.90	−1.77	2.86
Unpleasant: pleasant	4.17	0.98	−1.04	0.51
Not acceptable: acceptable	4.52	0.83	−1.97	3.87
Foolish: wise	4.48	0.91	−1.96	3.53
Not essential: essential	4.39	0.91	−1.52	1.81

,

Table 5. Descriptive statistics for subjective norms towards residential water conservation.

	Mean	SD	Skew	Kurtosis
I am expected to save water around the house and in my home landscape	3.68	1.08	−0.72	−0.02
The people who are important to me want me to save water around the house and in my home landscape	3.34	1.16	−0.34	−0.63
My neighbors would approve of me saving water around the house and in my home landscape	3.65	0.97	−0.45	0.10
Most people in my life whose opinions I value would approve of me saving water around the house and in my home landscape	3.85	0.94	−0.76	0.58
The people that I am close to would approve if I explored ways to reduce my water use around the house and in my home landscape	3.87	0.93	−0.82	0.81

,

Table 6. Descriptive statistics for perceived behavioral control towards residential water conservation.

	Mean	SD	Skew	Kurtosis
I am confident that I could save water around the house and in my home landscape if I wanted to	4.24	0.79	−1.36	2.99
The decision to save water around the house and in my home landscape is in my control	4.14	0.89	−1.17	1.57
Whether or not I save water around the house and in my home landscape is entirely up to me	3.97	1.00	−0.98	0.57
I am certain that I could save water around the house and in my home landscape if I wanted to	4.18	0.84	−1.26	2.26
I have complete control over the decision to save water around the house and in my home landscape	3.98	1.04	−1.02	0.52

,

Table 7. Descriptive statistics for residential conservation intention.

	Mean	SD	Skew	Kurtosis
Only run the washing machine when it is full	4.32	0.96	−1.66	2.38
Only run the dishwasher when it is full	4.42	0.86	−2.13	4.83
Only water your lawn in the morning or evening	4.27	0.88	−1.88	4.06
Reduce the number of times a week you water your lawn	4.26	0.87	−1.74	3.45
Sweep patios and sidewalks instead of hosing them down	4.39	0.85	−1.96	4.32

,

Table 8. Descriptive statistics for residential conservation behavior.

	Mean	SD	Skew	Kurtosis
I let my sprinklers run when it has rained or is raining (R)	4.37	0.88	−2.19	4.93
I let my sprinklers run when rain is predicted in the forecast (R)	4.20	0.90	−1.83	3.84
I only run the washing machine when it is full	4.28	0.87	−1.42	2.23
I only run the dishwasher when it is full	4.39	0.81	−2.00	4.81
I hose down my driveway (R)	4.08	1.02	−1.37	1.70

Note: (R) indicates a reverse coded item.

,

Table 9. Intrinsic motivation towards water conservation.

	Mean	SD	Skew	Kurtosis
I enjoy conserving water	3.82	0.84	−0.36	0.00
Conserving water is fun	3.41	0.92	0.03	−0.24
Conserving water is very interesting	3.33	0.99	−0.22	−0.10
Conserving water is quite enjoyable	3.32	0.94	−0.12	−0.03
While conserving water, I think about how much I enjoy it.	2.98	1.09	0.02	−0.48

and

Table 10. Extrinsic motivation towards water conservation.

	Mean	SD	Skew	Kurtosis
Receiving a reward	1.91	1.10	1.18	0.66
Avoiding a penalty	2.09	1.17	0.83	−0.31
Gaining social acceptance	2.11	1.15	0.79	−0.34

. Several items in residential conservation behavior were recoded prior to calculating descriptive statistics. Skewness did not exceed ±3 and kurtosis did not exceed ±10 for any of the items, which is acceptable under the guidance of Kline (2015) [43].

4.2. Structural Model

The first hypothesis that attitude towards residential water conservation will have a direct effect on residential conservation intention was supported. Attitudes (standardized = 0.35, p < 0.001) had a significant direct effect on residential conservation intention (see

Table 11. Results of the adapted TpB model, including the standardized path coefficients for the direct and indirect effects.

Effect	M	S.D.	Standardized Direct Effect (95% CI)	Standardized Total Indirect Effect (95% CI)	Standardized Indirect Effects through Intention (95% CI)	Standardized Indirect Effects through Intrinsic Motivation (95% CI)
Residential conservation intention	4.33	0.67	-	-	-	-
Attitudes	4.44	0.69	0.35 ***	-	-	-
Subjective norms	3.68	0.82	0.30 ***	-	-	-
Perceived behavioral control	4.10	0.76	0.03	-	-	-
Behavior	4.22	0.78	-	-	-	-
Intrinsic motivation	3.37	0.81	−0.13 ***	-	-	-
Extrinsic motivation	2.03	1.02	−0.47 ***	-	-	-
Residential conservation intention	4.33	0.67	0.21 ***	−0.14 ***	-	−0.14 ***
Attitudes	4.44	0.69	-	0.05 ***	0.07 ***	−0.02 ***
Subjective norms	3.68	0.82	-	0.05 ***	0.06 ***	−0.02 **
Perceived behavioral control	4.10	0.76	-	0.00	0.01	0.00
Intrinsic motivation	3.37	0.81	-	-	-	-
Residential conservation intention	4.33	0.67	0.40 ***	-	-	-

Note: *** p < 0.001; ** p < 0.01

and Figure 3). The second hypothesis that subjective norms towards residential water conservation will have a direct effect on residential conservation intention was supported. Subjective norms (standardized = 0.30, p < 0.001) had a significant direct effect on residential conservation intention. The third hypothesis that perceived behavioral control towards residential water conservation will have a direct effect on residential conservation intention was rejected. Perceived behavioral control (standardized = 0.03, p = 0.48) did not have a significant direct effect on residential conservation intention.

The fourth hypothesis that residential conservation intention will have a direct effect on residential conservation behavior was supported. Residential conservation intention (standardized = 0.21, p < 0.001) had a significant direct effect on residential conservation behavior. The fifth hypothesis that intrinsic motivation towards residential water conservation will mediate the relationship between residential conservation intention and residential conservation behavior was supported. The Sobel test was examined during the CFA to determine whether the mediation of intention on behavior through intrinsic motivation was present in the model. There was a significant indirect effect (standardized = −0.05, p < 0.001). The direct effect of intention on behavior (standardized = 0.20, p < 0.01) increased slightly when the mediator was present (standardized = 0.21, p < 0.01).

The sixth hypothesis that extrinsic motivation towards residential water conservation will mediate the relationship between residential conservation intention and residential conservation behavior was rejected. The Sobel test was conducted during the CFA to examine whether the mediation of intention on behavior through extrinsic motivation was present in the model. The indirect effect of extrinsic motivation (standardized = −0.00, p = 0.93) indicated that mediation was not present and was removed from the model. Therefore, the sixth hypothesis was rejected. Extrinsic motivation was examined as a direct effect on behavior (standardized = −0.47, p < 0.001).

The hypothesis that attitude, subjective norms, and perceived behavioral control towards residential water conservation will have an indirect effect on residential conservation behavior was partly supported. Attitudes (standardized = 0.05, p < 0.001) and subjective norms (standardized = 0.05, p < 0.001) had a significant indirect total effect on residential conservation behavior. Subjective norms (standardized = 0.00, p = 0.484) did not have a significant indirect total effect on residential conservation behavior.

5. Discussion

Intrinsic motivation towards residential water conservation was a mediating variable between residential conservation intention and residential conservation behavior. However, the relationship between intrinsic motivation and residential conservation behavior was negative, which was unexpected. Future studies should conduct interviews with consumers to determine why the relationship was negative or whether new scales are needed to measure intrinsic motivation and residential conservation behavior. The findings contradicted those of Li and Wen (2019), who found that intrinsic motivation had a positive relationship with respondents’ intention to use a bicycle-sharing program [27]. It is possible that the contradictory findings are due to intrinsic motivation being used as a mediating variable between intention and behavior in the present study, whereas Li and Wen (2019) used intrinsic motivation as a direct effect on intention [27]. In addition, Li and Wen (2019) used trust, sustainability, the sense of belonging, and enjoyment as measures of intrinsic motivation, whereas the present study used a scale adapted from Deci and Ryan (2000) [23,27]. Moreover, water conservation is less tangible of an experience than bicycle sharing, which may have caused the respondents to inaccurately measure their residential conservation behavior. Comparing different environmental behaviors rather than the same environmental behavior may also have caused the discrepancy. This may have been further exacerbated by self-reported versus actual measures of water conservation behaviors, as previous studies have found a discrepancy between conditions [39], and this should be treated as a limitation of this study. To strengthen the findings about water conservation behaviors, studies would benefit from measuring actual water consumption with smart meters or water bills.

Other studies have used attitudes as a measure of intrinsic motivation. For example, Bopp et al. (2019) described attitudes as individuals’ dispositions to perform a behavior based on their negative or positive evaluations of the behavior [46]. Intrinsic motivation was described as actions that individuals perform for the personal enjoyment of the activity itself [47]. Attitude and intrinsic motivation are theoretically different and therefore should not be used interchangeably, as seen in the present study considering the positive indirect effect of attitude on behavior and the negative direct effect of intrinsic motivation on behavior. It is possible that attitude and intrinsic motivation are highly correlated, and future studies should examine both variables to determine whether they have similar predictive abilities. Although all scales were deemed reliable, e.g., [27,28,46], determining the best measures for intrinsic motivation related to environmental sustainability remains a challenge.

Extrinsic motivation towards residential water conservation had a negative direct effect on residential conservation behavior, suggesting that water conservation behavior may decrease as extrinsic motivation increases. The mean extrinsic motivation towards residential water conservation indicated that the respondents disagreed that they have extrinsic motivation to conserve water. Perhaps the respondents had not thought about rewards or punishments in the context of water conservation because of the availability of water at a low cost and therefore were not extrinsically motivated to conserve water. For example, in Athens-Clarke County in Georgia, the monthly water bill for 5000 gallons of water is USD 40.48 (see https://dashboards.efc.sog.unc.edu/ga (accessed on 10 August 2023) for additional water prices in Georgia). The cost of water was not analyzed in this study in relation to extrinsic motivation and should be considered in future studies. It is possible that residents in other states who have experienced fines or other penalties related to the overuse of water may have different perceptions of extrinsic motivation towards residential water conservation. Future studies should examine water policies in different states to identify components of water conservation policies that could influence extrinsic motivation. In addition, future studies should conduct interviews or focus groups to explore nuances in whether/how the policy impacts motivation to conserve water. It is possible that extrinsic motivators for water conservation will not alter consumers’ consumption behaviors and that other variables like social practices and habitual routines need to be targeted, e.g., [7,8].

The direct effect of extrinsic motivation on residential water conservation was stronger than the direct effect of intrinsic motivation on residential water conservation, aligning with the results of Li and Wen (2019), who found that extrinsic motivation was a stronger predictor in the intention of participating in the bicycle-sharing program than intrinsic motivation [27]. Exploring alternative measures of extrinsic motivation to determine whether they are consistent with the measure used in this study may create a more robust model for predicting water conservation behavior. For example, a future study may consider measuring extrinsic motivation with a self-report scale and with an intervention, like providing a reward for not using a certain amount of water over one month as documented by their water bill.

The TpB variables explained 31.2% of the variance in residential conservation intention, and the independent variables only explained 30.7% of the variance in residential conservation behavior. The minimal predictive ability of intention surrounding behavior is noteworthy, as studies often measure intention, although this may be isolated to residential water conservation studies due to the non-tangible nature of water conservation. It is possible that intention is a better predictor of actual water conservation behaviors (e.g., water bills or smart meters) than self-reported water conservation behaviors and should be examined in future studies. Moreover, respondents may inaccurately gauge their residential water conservation intention and behavior because they do not have concrete comparisons of their water consumption to what is considered sustainable. Signage around communities, which includes the average water bill for households in one month, may help respondents gauge what a normal water bill is in their area, e.g., [48]. Many water companies have autopayments for consumers, where they do not need to look at their water bill or the associated gallons of water that they consume per month, so respondents may not be able to interpret gallons as easily as cost. An experimental design where consumers belong to a control group or receive signage related to the typical cost of water bills and then answer intention and behavior questions may help determine whether residential water conservation behavior is gauged inaccurately due to the lack of concrete comparisons or if there is still a disconnect between the intention and behavior constructs.

There was a significant direct effect of attitude on water conservation intention, contradicting the findings of Chaudhary et al. (2017) and Warner and Diaz (2021), who found that attitudes were not a significant predictor of intention [19,20]. It is possible that attitude plays a larger role in predicting residential conservation in the home, as Chaudhary et al. (2017) and Warner and Diaz (2021) examined landscape irrigation intention [19,20]. However, the findings are similar to those of Si et al. (2022), who broadly examined water-saving intentions and found that attitude was one of the strongest predictors of intention [1]. Although water-saving behaviors are often grouped together (e.g., household and landscape), they are not affected by the same constructs and therefore should be treated as separate components when crafting messaging targeted at reducing water consumption. This may also help explain the small amount of variance accounted for in residential conservation behavior by the independent variables mentioned previously, as the behavior questions were related to outdoor water consumption, whereas the intention items included both indoor and outdoor water consumption questions. Future studies should further examine attitudes in relation to water conservation due to their predictive nature in regard to intention and behavior. For example, antecedents of attitudes posited in the affective, behavioral, and cognitive model—or the ABC model of attitudes—can help identify components of attitude to target educational interventions and/or communication messaging.

There was a significant direct effect of attitudes and subjective norms on intention, but perceived behavioral control did not predict intention. This finding contradicts that of Si et al. (2022) and Gibson et al. (2021), who found that attitudes, subjective norms, and perceived behavioral control had an effect on water conservation intention [1,14]. It is possible that attitudes, subjective norms, and perceived behavioral control interact differently with consumers’ conservation intention depending on geographic location. For example, Florida is a water-scarce state and frequently receives communication regarding water conservation practices. Future studies should examine whether one facet of TPB is the most effective for addressing audiences in specific geographic locations.

5.1. Limitations

The limitations of this study need to be addressed. First, all measures were self-reported and may be limited by social desirability bias. Social desirability bias happens when respondents answer items in the way that they believe is viewed favorably by others or how another would want them to respond.

The chi-squared statistic in the CFA and structural model was significant, indicating that the proposed model was not suitable. There is a debate in the literature as to the most appropriate model fit tests, with some arguing that a non-significant chi-squared statistic is needed and others arguing for model fit indices [44,49]. Larger sample sizes can cause significant chi-squared statistics, and, therefore, chi-squared was not utilized when examining the model fit [43]. Moreover, the model fit for RMSEA only indicated an acceptable model fit. Model fit recommendations have shifted throughout the years, with Hu and Bentler (1999) indicating an acceptable model fit for RMSEA below 0.06 [42], Brown and Cudeck (1992) indicating a close model fit below 0.05 and an acceptable model fit below 0.08 [50], and Steiger (2007) indicating an acceptable model fit below 0.07 [51].

The items used in the final behavior scale only addressed outdoor water conservation behaviors after removing items due to the CFA results. The other TpB variables addressed both indoor and outdoor water conservation behaviors and therefore may be theoretically different from the behavior variable. Future studies should consider measuring indoor and outdoor water conservation behaviors as separate dimensions for TPB variables. In addition, irrigation behaviors like watering the lawn may be distinctly different from outdoor behaviors like hosing down the driveway and should also be considered. Indoor conservation behaviors like dishwasher use may be theoretically distinct from personal hygiene factors like toilet flushing.

6. Conclusions

The study findings suggest that intrinsic motivation and extrinsic motivation may not be the most effective variables to help explain the relationship between intention and behavior in water conservation. Thus, using motivational frames in water education programs or communication messages is not recommended. However, the study results are preliminary and may vary due to numerous factors, such as geographic location and timeframe. The TpB variables predicted self-reported intention and self-reported behavior, but the additional variance in the constructs needs to be explained. Moving forward, additional variables such as identity or cultural norms need to be explored to help understand consumers’ water conservation behavior.