Using Temporal Framing and Numerosity to Convey Messages of Water Consumption – A Study on How to Reduce Household Water Consumption in the U.S.
A Master’s Thesis
Supervisor: Dr. Mehrad Moeini-Jazani Co-assessor: Dr. Sumaya Albalooshi
By J.J.T. van Putten
S2877718
University of Groningen Faculty of Business and Economics MSc Thesis Marketing Management
11/01/2021 Word count: 7307 Author: J.J.T. van Putten S2877716 MSc Marketing Management Taco Mesdagstraat 23A, Groningen
2 Table of Contents
1. Abstract ……….……3
2. Introduction ……….……4
3. Literature Review ……….…....5
a. Evolution of Literature Regarding Household Water and Energy Consumption………5
b. Non-Quantitative Temporal Framing and Environmental Psychology………...……6
c. Quantitative Temporal Framing and the Numerosity Effect………...……8
4. Research Methodology ………...……….10
5. Results and Discussion ………13
a. Results Part 1……….…13
b. Results Part 2 – Focusing on Temporal Framing without Numerical Information ...………...16
3
Abstract
Our global water supply is quickly depleting, and household water consumption has been steadily on the rise. As current literature on the most effective ways to convey a message of reducing water consumption is limited, this study aims to determine the effect of temporal framing (e.g., “every day” vs. “every year”) and numerosity (e.g., numerical information: absent vs. present) in messages regarding water consumption on pro-environmental intentions.
In a study with 689 participants, respondents observed one of 4 versions of an advertisement regarding average household water consumption in the U.S. Two of these ads differed only on their temporal framing, and no numeric information was communicated: “a significant
amount every day” or “a significant amount every year”. The other two advertisements used similar temporal framings, but also communicated corresponding numerical information: “88 gallons every day” or “32,120 gallons every year”. Subsequently, we measured participants’ pro-environmental intentions together with a series of potential moderators and mediators. We examined whether and how temporal framing and numerical information affected
pro-environmental intentions.
As very little significant effects could be found for numerosity conditions, this study primarily focused on the effect of temporal framing. The study found that for individuals that rank lower in Consideration of Future Consequences (CFC), or have a concrete mindset, the most proximal temporal frame (“every year”) invokes the most concern for the problem. Through tests for mediated moderation, this study found that concern is a significant predictor for pro-environmental intentions.
4
Introduction
Water. Over 75% of the earth is covered with water, yet 1.4 billion people live without clean drinking water. Almost all water is in the earth’s oceans, meaning that only 3% of the earths water can be used as drinking water. Moreover, of that 3%, most is frozen in the polar ice caps. Compared to 1901, global water consumption has increased nearly five-fold (Pacific Institute, 2014). Because of this, and many other statistics, global scientists are worried about the security of drinking water in the future. Unless we change our ways, two-thirds of the world’s population will face water scarcity by the year 2025 (SaveTheWater, 2020). The average person in the United States consumes the most household water per capita in the world, on average 88 gallons every single day (U.S.G.A.O., 2014). Compared to Germany (32 gallons per day), this is nearly three times as much (OECD, cited in Biswas, 2012). In order to secure future drinking water, reducing household water consumption is essential. How can we convey this message most effectively?
The effects of numerosity on human perception have been an interesting field of research for nearly the past 100 years. Over the years it became clear that making use of the numerosity heuristic can be beneficial for marketing activities, resulting in many studies on how consumers are tricked into believing something is more than it actually is (e.g., Pandelaere, Briers & Lembregts, 2011).
Temporal framing has significant effects on how people perceive risk, as day framing appears more threatening when discussing the risk of tobacco use (Chandran & Menon, 2004). The quick depletion of our global fresh water supply is by many scientists perceived as a great risk, hence the utility of understanding the effect of temporal framing when conveying a message of water saving.
Temporal framing can be combined with a form of numerosity, as it might use different ‘currencies’ to tell the same amount (e.g., X per year = X/365 per day). Next to making the issue seem more pressing, the difference in large or small numbers can significantly affect how individuals perceive the magnitude of the problem. The existing literature on temporal framing is still unclear on whether a large number in a year frame, or a small number in a day frame, is most effective when informing about environmental issues.
5
Literature Review and Hypotheses
Evolution of Literature Regarding Household Water and Energy Consumption
According to the author’s best knowledge, this paper will fill a void in the existing literature. In studies performed on household water consumption, and the attempts to reduce household water consumption, we observe an evolution. First, it was assumed that people’s behavior could be altered by providing them information. The base-line suggestion in these studies was that individuals had some notion of knowledge deficiency (e.g., Burgess, Harrison & Filius, 1998). However, in the field of climate change, studies suggest that a better understanding does not translate into behavioral change (Beattie, 2010, Finger, 1994, Leiserowitz, 2006, McKenzie-Mohr, 2008, Owens, 2000, as cited in Seyranian, Sinatra & Polikoff, 2015). Later, studies in the domain of water preservation evolved into more attention to providing feedback by observing studies done on energy preservation. In these studies, ‘boomerang effects’ were also observed. This entailed that households received feedback on their energy consumption that included an average of the neighborhood consumption. When their
household consumption was below the average, these households were prone to consume more than they did before (e.g., Schultz et al., 2007). In studies on feedback mechanisms, both physical and electronic, mixed results were obtained concerning reducing energy consumption (Corral-Verdugo et al., 2012). Hence the need to investigate more. The studies evolved into the present, and Seyranian et al. (2015) focused their current research on the factors invoked in each type of message: social norms, social identity, and personal identity. Seyranian even finds that using the ‘knowledge deficit approach’ results in more water consumption. Therefore, the current focal point of research into water
consumption has now shifted towards personal and social identities.
As we can observe, a structure has emerged in the field of water consumption that is
6
Non-Quantitative Temporal Framing and Environmental Psychology
Chandran & Menon (2004) demonstrate that framing temporal distance has significant effects on an individual’s perception of risk. In their study on tobacco use, the risk of succumbing to heart disease due to smoking was either framed as “per day” or “per year.” They also added a layer of distance, by conveying a “tomorrow” message, and “a year from now.” The results in their study are essential for our understanding in persuading people of risk: ““every day”
framing makes risks appear more proximal and concrete than “every year” framing”.
Chandran & Menon also find that framing a message as “every day” results in higher levels of concern and anxiety about the hazard, a higher level of intentions to exercise precautionary behavior, as well as higher self-risk perceptions (Chandran & Menon, 2004).
These results are in line with the Construal Level Theory, advanced by Trope & Liberman in 2010. According to this theory, people can think about the present or the future, but their primary reference point is in the here and now. Transcending this reference point entails a mental construal, and the further away the object is removed from the present, the higher and more abstract the construal becomes (Trope & Liberman, 2010). According to their analysis, the level of construal affects preferences and actions. The more concrete and lower the construal is, the higher the risk estimate would be (Chandran & Menon, 2004).
Yet, this theory is also criticized by researchers in the domain of environmental psychology. In a 2016 study by Brügger et al. from the University of Bern, the authors propose that focusing on proximal climate change does not increase the willingness to respond. However, they find evidence that the varying psychological distance influences what information people act on (Brügger, Dessai, Devine-Wright, Morton & Pidgeon, 2015). Their study finds that participants with a proximal perspective relied on fear and participants with a distance focus relied on skepticism in order to make decisions regarding climate change.
In the domain of environmental psychology, there are contesting theories and studies
7
Due to the different visions on the effect of temporal framing, it is still quite unclear what the effect of temporal framing will be on household water consumption. However, discussing freshwater shortages in the future invokes elements of risk communication. Based on the current literature, we can assume that a more proximal temporal frame (“every day”) results in a higher sense of urgency. As we assume that conveying a message of risk, which this study inherently does, is not the same as conveying a message of sustainability, we predict that the most proximal temporal frame invokes more concrete levels of thinking. In the health domain, it has been proven that higher risk perceptions increase the likelihood of behavioral change (Ferrer & Klein, 2015). We suspect that this effect will be replicated in this study, but in the domain of the environment. Other studies demonstrated that a low-level, concrete mindset is more effective when paired with a negative message frame (White et al., 2011). This is in line with our research, as our study invokes a negative message frame by conveying a message of water consumption, rather than a message of conservation.
Therefore, drawing from Chandran & Menon (2004) and the construal level theory presented by Trope & Liberman (2010), as well as other papers, this paper hypothesizes the following: H1a: Conveying a message regarding household water consumption in a more proximal temporal frame (“every day”) will result in a higher level of concern.
H1b: Conveying a message regarding household water consumption in a more proximal temporal frame (“every day”) will result in a higher level of concern, which in turn will result in increased pro-environmental intentions.
Not being able to confirm our first hypothesis would open the door for the possibility that Reczek et al.’s (2018) observation holds; a more abstract construal level (“every year”) results in more positive reaction to eco-friendly products (which could be an indicator for intention or behavior). Therefore, hypotheses H1a and H2 are directly competing:
H2: Conveying a message regarding household water consumption in a more abstract temporal frame (“every year”) will result in higher levels of pro-environmental intentions. As we assume that, according to the Construal Level Theory, concrete thinking results in higher levels of concern, we also hypothesize that:
8
Quantitative Temporal Framing and the Numerosity Effect
A major difference in how temporal framing is implemented between studies, is the use of numbers or not. To refer back to the paper of Chandran & Menon (2004), we find that they conducted their study without numbers. This entails that they framed their message as either
“Every day, a significant number of people succumb to heart disease” or “Every year, a significant number of people succumb to heart disease” and is therefore an example of
non-quantitative messaging. Despite this lack of detailed information, they still find an effect of temporal framing.
Others follow a different path. In a 2015 study on residential energy savings, Xu et al. combine temporal framing with numerosity in order to determine its effect on individuals’ attitudes. The authors framed both environmental benefits and economic benefits in either
“per month” or “per year” and adjusted the numbers provided accordingly (i.e., “reduces 660 pounds of carbon emissions (...) in a year” vs. “reduces 55 pounds of carbon emissions (…) in the next month” & “Saves $6 per bulb in a year” vs. “Saves $.50 per bulb in the next month”). The study found that for individuals with higher levels of Consideration of Future
Consequences (CFC) and moderate levels of environmental concern, environmental framing, regardless of the temporal frame, was most effective. However, for participants lower in CFC & environmental concern, the most effective strategy was short-term, economic framing (Xu, Arpan & Chen, 2015). This study has a significant limitation, as the authors do not use equivalent opposites in their temporal framing. The authors frame the numbers either as “per year” or “in the next month”. According to literature, this might create additional
(unnecessary) psychological distance (e.g., Trope & Liberman, 2010).
Now that we have identified that there are at least two types of temporal framing, quantitative or non-quantitative, we must establish which is more effective. In 1981, Petty et al. developed a theory involving two routes that determine the strategy in which people invoke processing information: a central & a peripheral route(Petty, Cacioppo & Goldman, 1981). This theory was later dubbed the Elaboration Likelihood Model (ELM) (Petty & Cacioppo, 1986). However, in 1984 the effect of a quantitative message was already tested along the lines of these two routes. This study found that the presentation of quantitative information stimulated people to rely on a peripheral cue (e.g., the communication source or the expertise of the communicator) as a basis for judgment and that the presentation of non-quantitative
Elmore-9
Yalch, 1984). Based on Yalch & Elmore’s work, we find that a quantitative message will likely invoke the peripheral route, and that non-quantitative messaging will likely invoke the central route in processing information.
Petty et al. (1981) present interesting results for our study, as they find that personal relevance is a moderator for which strategy is used: high personal relevance resulted in a focus on the quality of the argument (central route), whereas low personal relevance predicted a focus on the credibility of the source (peripheral route).
Next to the Elaboration Likelihood Model determining the effect of numbers in general, we must also consider the numerosity effect when specifically investigating the numbers used. Numerosity research suggests that numerical units impact an individuals’ judgment, as the unit in which the number presented is often disregarded (Pelham et al., 1994). This is famously demonstrated in research on monetary values (Raghubir et al., 2002). Converting exchange rates is difficult, and therefore $16HKD might appear to be “more expensive” than $2USD, despite being equally expensive (Chan, 2018). However, currencies do not
necessarily be monetary, different methods to explain time can be seen as currencies as well. Framing a message as a smaller number per day or a larger number per year can (despite their total being the same) have a significant impact on the interpretation of the number, as
consumers focus on the number rather than the type of unit in which the number is expressed (Pandelaere, Briers & Lembregts, 2011). According to Pandelaere, who focused their research on different types of scales (1-10 vs. 1-1,000), consumers tend to rank the attributes of the higher number as greater, e.g., a larger quantity or quality. For our research, this might indicate that the larger number (year frame) will be more pressing, as the larger number per year appears to be more despite being equal in total. This is opposite to the effect of temporal framing, in which the more proximal frame will most likely have the strongest impact on behavioral intentions.
However, others argue that the presentation of numbers expressed in a finer granularity increases confidence in accuracy, as the number provided seems more precise (Zhang & Schwarz, 2012). The combination of precision and higher risk estimates could prove to be more effective than the larger number in a year frame.
10
number will create the highest perceived value and therefore seem more pressing, or a smaller number in a finer granularity will invoke more precision and higher risk perceptions. This results in the following (competing) hypotheses:
H4: Conveying a quantitative message regarding household water consumption in a more proximal temporal frame (“88 gallons every day”) will result in a higher level of pro-environmental intentions, especially among those with lower levels of climate concern. H5: Conveying a quantitative message regarding household water consumption in a long-term temporal frame (“32,120 gallons every year”) will result in a greater perception of problem magnitude, as the numerosity effect creates the illusion that the total is greater.
H6: In line with the Elaboration Likelihood Model, the numerosity effect will be stronger for individuals with no personal experience or relevance.
Research Methodology
Design. In order to test our hypotheses, we used a 2 (temporal frame: day vs. year) x 2
(numerical information: present vs. absent) between-subject design. Every respondent was randomly assigned to one of four conditions, and the possible effects of the different
conditions on the dependent variable was studied. The four conditions were presented in the form of an advertisement.
Procedure. To create a large sample size during the COVID-19 pandemic, we used
11
In order to test the numerosity effect, consumption was described as “(…) 88 gallons of
household water every day” or “(…) 32,120 gallons of household water every year”.
Figure 1. Scenario 1 (day frame, numerical information absent) Figure 2. Scenario 2 (year frame, numerical information absent)
Figure 3. Scenario 3 (day frame, numerical information present) Figure 4. Scenario 4 (year frame, numerical information present)
12
Time. After measuring the effects on the dependent variable, several potential moderators were included in the survey: Consideration of Future Consequences; Ecological Value; Geographical Area and Personal Experience; Perceived Accuracy, Believability, Credibility, and Relevance; Concrete and Abstract Mindsets; and Demographics. The total survey duration was, on average around 10 minutes.
Measures
In order to determine the perception of problem magnitude, questions were either created by the author self or adapted from the Climate Change Risk Perception Scale by Kellstedt et al. (2008).
The format for questioning behavioral intentions was slightly derived from L. Duchi et al. (2020). In their paper, behavioral intentions were split into four categories. In our paper, we split behavioral intentions into two parts (‘Behavioral Intentions’ and ‘Willingness to Donate Time’). Willingness to Donate Time was recorded as the percentage of free time a participant was willing to spend promoting the message of water saving.
Measures for Consideration of Future Consequences were derived from H. Bruderer Enzler et al. (2019), in which they used a shortened version of the CFC survey, involving only 6
questions that get to the heart of CFC. We opted for this due to survey duration constraints. Ecological value was tested along 6 7-point Likert Scale questions, derived from the New Ecological Paradigm Scale proposed by Dunlap et al. (2000).
Geographical Area and Experience was tested by 4 “yes/no” questions, intended to determine whether the participant experienced water shortages before.
Concrete and Abstract Mindsets were tested with the Action Identification Task designed by Vallacher & Wegner (1989), which includes 10 questions determining the participant’s mindset.
The requested demographics included e.g., age, gender, education level, income, employment status, family size, political orientation, and Conservatism (measured through a 7-point scale). Respondents were also asked to elaborate on their experience during this survey (e.g., type of device used, leaving their PC, or quietness) and their experience on the Amazon platform in general before answering this survey (e.g., numbers of surveys completed that day).
13
Results - Part 1
After exclusions, our total sample size included 689 (mean age of 40 years old, S.D. = 13.34) respondents, of which the majority (55,9%) was female. In this sample, 63,4% claimed to hold at least a College or University Degree, 76,2% claimed to be currently employed, and 54% had a personal annual income of 50,000USD or less. The majority of respondents self-identified as Democrat (45,6%), as opposed to 21,6% identifying as Republican and 27,7% identifying as Independent. A complete overview can be found in Appendix K.
Cronbach’s Alpha tests were performed on all the variables, and all were above the threshold of ,6. Results can be found in Appendix B.
In order to test the effects of both Temporal Framing and the Numerosity Effect (non-quantitative and (non-quantitative messaging) two-way ANOVAs were performed. A two-way ANOVA, including Temporal Framing and Numerosity, revealed no significant interaction on Behavioral Intentions (F(1,685)=,127; Pinteraction=,721). No significant interaction effect could
be demonstrated when testing for Willingness to Donate Time (F(1,685)=,417;
Pinteraction=,493) or Average Concern (F(1,685)=,339; Pinteraction=,561). This could indicate
that the Numerosity Effect has no significant effect on our dependent variables.
Figure 5. Estimated means for (l-t-r) Behavioral Intentions, Willingness to Donate Time, and Average Concern (red = numerosity present)
One significant effect was observed. The effect of Temporal Framing on Willingness to Donate Time in the two-way ANOVA was significant, R2 = ,007; (F(1,685)=3,472; p=,063*.
14
Figure 6. The effect of Temporal Framing on Willingness to Donate Time.
To make sure no interactions were missed, Hayes’ PROCESS model 3 was used to test for three-way interactions, with Temporal Framing as Independent Variable and numerosity as moderator (W). The three-way models were conducted on Time Donation; Behavioral
Intentions; and Average Concern, and included either Consideration of Future Consequences, Ecological Value, or Conservatism as second moderator (Z).
Only two three-way interactions proved significant. Firstly, the interaction between Temporal Framing, numerosity, and Conservatism significantly predicted Time Donation (F=1,5867;
Pinteraction=,0225**). In general, framing a message in a “year” frame has significant negative
effects for those identifying as Conservative (effect=-6,7815; p=,0374**). Further investigation in the conditional effects reveals the effect of numerosity in this model: communicating “88 gallons every day” is more effective for Conservatives than communicating “32,120 gallons per year” (effect=-5,1376; p=,0273**). For Liberals, numerosity has no effect in this model. It is revealed, however, that for Liberals when the Numerosity Effect is not present, a more proximal Temporal Frame is more effective
(effect=-4,0337; p=,0609*). A conceptual model of this effect can be found in Appendix C.
The other significant three-way interaction included Personal Experience as a moderator (F=4,4205; Pinteraction =,0769*). However, in the conditional effects, no significant effects
could be observed.
15 Independent Variable Dependent Variable Moderator (W) Moderator (Z) F-value P-value Interaction Time Donation Temporal Framing Numerosity Effect CFC F=1,3776 Pinteraction =,9853 Time Donation Temporal Framing Numerosity Effect Ecological Value F=1,1795 Pinteraction =,6377 Time Donation Temporal Framing Numerosity Effect Conservatism F=1,5867 Pinteraction =,0225** Time Donation Temporal Framing Numerosity Effect Personal Experience F=4,9421 Pinteraction =,9932 Behavioral Intentions Temporal Framing Numerosity Effect CFC F=24,0821 Pinteraction =,9939 Behavioral Intentions Temporal Framing Numerosity Effect Ecological Value F=,6359 Pinteraction =,1159 Behavioral Intentions Temporal Framing Numerosity Effect Conservatism F=4,7544 Pinteraction =,3287 Behavioral Intentions Temporal Framing Numerosity Effect Personal Experience F=4,4205 Pinteraction =,0769* Average Concern Temporal Framing Numerosity Effect CFC F=16,6349 Pinteraction =,3735 Average Concern Temporal Framing Numerosity Effect Ecological Value F=,6410 Pinteraction =,4751 Average Concern Temporal Framing Numerosity Effect Conservatism F=12,0491 Pinteraction =,4142 Average Concern Temporal Framing Numerosity Effect Personal Experience F=3,9006 Pinteraction =,2548 Table 1. Interaction Results Discussion
16
With these results, we can conclude the following about our hypotheses:
• H4: no significant evidence could be found to support our fourth hypothesis. The combination of a more proximal Temporal Frame and the Numerosity Effect does not result in higher levels of pro-environmental intentions (or willingness to donate more time) regardless of the levels of climate concern.
• H5: our results provided no significant evidence to support our fifth hypothesis. No significant effects could be found to demonstrate that a long-term proximal frame will result in a greater perception of problem magnitude (or concern).
• H6: Some significant results were found that personal experience moderates the effect of numerosity. However, these results could only be found for behavioral intentions and could not be replicated. Also, conditional effects demonstrated only marginal significant results, which indicated an opposite effect than our hypothesis. Therefore, our sixth hypothesis could not be confirmed with this study.
It is clear that the direct and three-way effects on our dependent variables were minimal. Because previous research has focused on temporal framing without numerical information (e.g., Chandran & Menon, 2004; Xu et al., 2015), we also turned our focus on the two conditions in our analysis where no numerical information was communicated for the
subsequent analysis. However, it is noteworthy that for the sake of completion, the upcoming tests have been conducted on our data for the conditions with numeric information. Results of those analyses can be found in the Appendix (see Appendix F).
Results Part 2 – Focusing on Temporal Framing without Numerical
Information
Demographics
After excluding the 345 participants who were exposed to numerical conditions, our
17
Main Effect of Temporal Framing on Dependent Variables
New ANOVA’s were performed after removing the numerosity conditions. No direct effects between Temporal Frame and Time Donation were observed (R2= ,003; F(1,342)=,779; P=,378). Similarly, no direct effects between Temporal Frame and Behavioral Intentions
(R2= ,003; F(1,342)=,939; P=,333) and between Temporal Frame and Average Concern (R2=
,005; F(1,342)=1,656; P=,199) were revealed.
We therefore decided to test potential moderators and mediators:
Figure 7. Overview of mediated moderation models tested below.
Exploring the Moderating Effect of CFC on The Link Between Temporal Framing and Dependent Variables
Hayes’ Model 1 analysis demonstrated that CFC moderates the effect of Temporal Framing on Average Concern (F=27,8525; p=,0439**). This is especially significant for those who score lower in CFC: moving from an “every day” frame to an “every year” frame significantly decreases the average concern (effect= -,4528; p=,0122**).
18
An ANOVA revealed that Average Concern has significant effect (R2=,450;
F(33,344)=7,684; p=,000***) on Behavioral Intentions. The more concerned an individual,
the better the Behavioral Intentions. Similarly, another ANOVA revealed that Average Concern has a significant effect (R2=,185; F(33,344)=2,127; p=,001***) on Willingness to
Donate Time (Appendix H).
Figure 9. The effects of concern on Behavioral Intentions (l) and Time Donation (r)
We therefore decided to test for moderated mediation with Hayes’ Process Model 8, which demonstrated significant results for both ‘Willingness to Donate Time’ and ‘Behavioral Intentions’. In our first model, Consideration of Future Consequences moderated the effect of Temporal Framing on Average Concern, and Average Concern mediated the effect of
Temporal Framing on Behavioral Intentions. As predicted, Temporal Framing, moderated by CFC, significantly predicts Concern. As predicted, concern significantly predicted Behavioral Intentions. The entire model was proven to be significant as the upper and lower confidence intervals of the indirect effects did not include 0 (LLCI=-,3664; ULCI=-,0152). Similarly, the entire model proved to be a significant predictor for Willingness to Donate Time
(LLCI=-4,0734; ULCI= -,146). We can conclude that Model 8, where CFC moderates the effect of
Temporal Framing on the mediator Concern, is a significant model for predicting both Behavioral Intentions and Willingness to Donate Time.
19
Exploring the Moderating Effect of Conservatism on The Link Between Temporal Framing and Dependent Variables
Model 1 of Hayes’ Process did not reveal a significant interaction effect between
Conservatism and Temporal Frame on concern (F=1,3490; Pinteraction= ,2463). However, we
find interesting results in the conditional effects. Our data indicates that the more Liberal an individual is, the less concern a long-term temporal frame invokes (effect= -,3118;
p=,0537*).
As demonstrated, concern remains a significant predictor for behavioral intentions. Hayes’ Model 8, where Conservatism moderates the effect of Temporal Framing on concern, proved to be a significant overall model of moderated mediation for Behavioral Intentions, as the upper- and lower confidence intervals for Liberals did not include 0 (LLCI=-,2229;
ULCI=-,0181). Similarly, this entire model proved to be a significant predictor for Willingness to
Donate time amongst Liberals, as confidence intervals did not include 0 (LLCI=-2,6200;
ULCI=-,1812).
Figure 11. Visualization of Hayes’ Process 8 moderated by Conservatism.
An ANOVA revealed a significant effect of Conservatism on CFC. Individuals high in Conservatism rank significantly lower in CFC (R2= ,046; F(6,682)=5,489; p=,000***)
(Appendix I).
Exploring the Moderating Effect of Abstract or Concrete Mindsets on The Link Between Temporal Framing and Dependent Variables
In the third significant model 8 analysis, ‘Abstract or Concrete Mindsets’ was used as moderator between Temporal Frame and Average Concern. The interaction effect was significant (F=1,9280; Pinteraction=,0937*) on a 90% confidence level for predicting Average
20
Conditional effects demonstrated that for those with a concrete mindset, implementing a proximal Temporal Frame (“every day”) resulted in significantly higher levels of concern (effect=,4251; p=0,0370**), which in turn results in greater behavioral intentions and willingness to donate more time. The entire model was proven to be significant, as the upper and lower confidence intervals of the indirect effects did not include 0 (LLCI=-,4235;
ULCI=-,0037), for Behavioral Intentions. However, the entire model did not prove to be a
significant predictor for Willingness to Donate Time (LLCI=-3,2902; ULCI=,0700).
Figure 12. Visualization of Hayes’ Process 8 moderated by ‘Abstract vs. Concrete Mindset’.
Discussion
Despite excluding those assigned to the numerosity conditions, still no direct effects could be found for Temporal Framing. Hayes’ Process revealed that mediated moderation explains the differences in our data. How Temporal Framing affects concern is significantly moderated by both Consideration of Future Consequences, and whether the individual has an abstract or concrete mindset. For Liberals, it is significantly better to frame a message as “every day”, as that increases the level of concern. In order to convey the most effective message of water conservation for those low in CFC, it is wise to frame the message as “every day” as it significantly increases their concern about the problem. The same strategy can be applied for individuals with a concrete mindset. As concern is a significant mediator, implementing these strategies will result in greater levels of pro-environmental intentions. CFC appears to be the strongest moderator for predicting the effect of Temporal Framing on concern.
21
With these results, we can conclude the following about our hypotheses:
• H1a: No direct effects of a more Temporal Frame (“every day”) on levels of concern were observed. However, we have found significant effects of Temporal Framing on levels of concern through moderator CFC and the type of mindset (concrete vs. abstract). We therefore find sufficient evidence to support the claim that a more proximal Temporal Frame (“every day”) results in higher levels of concern, for individuals with a concrete mindset and/or low in CFC.
• H1b: We have found evidence that the level of concern mediates the effect of
Temporal Framing on pro-environmental intentions, depending on levels of CFC and the mindset of an individual (concrete vs. abstract). Our models demonstrate that a more proximal Temporal Frame (“every day”) results in higher levels of concern. The more concerned an individual is, the more behavioral intentions increase. There is sufficient evidence to support this claim.
• H2: No direct effects of a long-term Temporal Frame on pro-environmental intentions were observed. Therefore, we find no sufficient evidence to support this hypothesis. • H3: We find sufficient evidence with regards to the hypothesis that for individuals
with a concrete mindset, a more proximal Temporal Frame (“every day”) results in higher levels of concern. We did not find evidence to support that the total amount of concern in a “every day” frame will be higher for individuals with concrete mindsets than for individuals with abstract mindsets.
General Discussion
The goal of this thesis is to add to existing literature ways in which messages aiming to reduce water consumption can be improved. Current literature has undergone great evolutions over the last decades and is currently focused on personal and social identities in providing feedback regarding water consumption. With this study a new layer is presented, where temporal framing is used to demonstrate the effects of “every day” or “every year” framing, with and without numerical information, on pro-environmental behavioral intentions.
22
either be because our measurements were wrong, or our statistical sample was incorrect. As effect sizes in environmental studies tend to be rather small, the possibility arises that a larger sample size (and more statistical power) would have resulted in significant results. Especially considering the fact that some effects were observed (but were insignificant) in this particular study.
Despite not finding direct results, we did observe significant models where the effect of numerosity on the link between temporal framing and pro-environmental intentions was affected by personal traits. From what we can determine, if the goal is to increase
pro-environmental behavior (in this case: reduce water consumption) amongst Conservatives, the best way to move forward is to frame the message as “88 gallons every day”, rather than “32,120 gallons every year”. It is very well possible that this effect mainly holds for
Conservatives as Liberals are willing to improve their ecological footprint regardless of the temporal frame in which the message is portrayed. Another explanation is that Liberals are more future oriented than Conservatives (Robinson et al., 2015).
As there appeared to be more interesting observations in our data, we excluded all those that received numeric information and were therefore exposed to the numerosity effect. Therefore, only non-numeric temporal framing was tested for the following results. We decided to only include temporal framing, as that has proven to be a significant factor in the domain of health. Finding similar results for messages regarding water conservation would be a novel addition to the current literature.
23
Our study could not replicate these overall significant models with moderators Conservatism and Ecological Value. However, despite not being a significant model in general, significant conditional effects were found for Conservatism as a moderator on the mediation between temporal framing and pro-environmental intentions. These conditional effects suggest that the more liberal an individual is self-described, the less concern a long-term temporal frame (“every year”) invokes. However, since the total model is not significant, more research is required to determine what effect Conservatism has on the effect of temporal framing on concern.
A potential explanation for the significant conditional effects was found when we tested the link between CFC and Conservatism. Our study found that the Conservatives in our study scored significantly lower on Consideration of Future Consequences. Similarly, we found that men ranked significantly lower in CFC, as well as those that were less educated (Appendix I). It appears that CFC differs among different groups of people, and that this difference affects which temporal frame is most effective. More research needs to be done to confirm this. Appendix J includes a conceptualization of the complete potential CFC model we found in our Study.
Limitations
24 Future Recommendations
For future research, we recommend investigating the effect political orientation has on temporal framing. Understanding what frame fits which group of people best in order to deliver an effective message is crucial in our fight against the environmental issues facing us today. We also recommend a larger study on the effect of numerosity in the field of
environmental messaging. Our study found a significant direct effect of the numerosity effect (“88 gallons per day” vs “32,120 gallons per year”) on Willingness to Donate Time, but this effect could not be replicated for Behavioral Intentions (see Appendix G). We believe that a larger sample size and a study solely focused on the numerosity effect might find significant results. This research was performed in the domain of household water conservation, we recommend studying the effect of temporal framing in all domains of environmental action. Also, we recommend testing the results of this study cross-culturally. The U.S. might be the biggest consumer of household water, plenty of countries worldwide can cut back on their consumption.
Conclusion
This is the first study in the domain of water conservation strategies that investigated the effect of temporal framing on pro-environmental behavioral intentions. It is also the first study in the domain of water conservation that combined temporal framing with numerosity. The implications of our findings are important. Our survey has demonstrated that
Consideration of Future Consequences is an important character trait in determining which temporal frame is most effective. Based on our dataset and analyses, we find that
25
References
Ajzen, I., & Fishbein, M. 1977. Attitude-Behavior Relations: A Theoretical Analysis and Review of Empirical Research. Psychological Bulletin, 85(5).
Ansbacher, H. 1937. Perception of number as affected by the monetary value of the objects. Archives of Psychology, 215.
Bathke, D. 2021. U.S. Drought Monitor. Droughtmonitor.unl.edu. https://droughtmonitor.unl.edu.
Berliner Wasserbetriebe. n.d. Berliner Wasserbetriebe - Water Usage at Home. Berliner
Wasserbetriebe. https://www.bwb.de/en/2266.php.
Biswas, A. 2012. Water prices in Europe need to rise substantially to encourage more sustainable water consumption. LSE.
Https://blogs.lse.ac.uk/europpblog/2012/10/31/europe-water-prices/.
Bruderer Enzler, H. 2013. Consideration of Future Consequences as a Predictor of
Environmentally Responsible Behavior. Environment and Behavior, 47(6): 618-643. Brügger, A., Dessai, S., Devine-Wright, P., Morton, T., & Pidgeon, N. 2015. Psychological
responses to the proximity of climate change. Nature Climate Change, 5(12): 1031-1037. Burgess, J., Harrison, C., & Filius, P. 1998. Environmental Communication and the Cultural
Politics of Environmental Citizenship. Environment and Planning A: Economy and
Space, 30(8): 1445-1460.
Chan, E. 2018. Climate change is the world's greatest threat – In Celsius or Fahrenheit? Journal of Environmental Psychology, 60: 21-26.
Chandran, S., & Menon, G. 2004. When a Day Means More than a Year: Effects of Temporal Framing on Judgments of Health Risk. Journal of Consumer Research, 31(2): 375-389. Corral-Verdugo, V., Frías-Armenta, M., Tapia-Fonllem, C., & Fraijo-Sing, B. 2012.
26
Demarque, C., Apostolidis, T., & Joule, R. 2013. Consideration of future consequences and pro-environmental decision making in the context of persuasion and binding
commitment. Journal of Environmental Psychology, 36: 214-220.
Dunlap, R., Van Liere, K., Mertig, A., & Jones, R. 2000. New Trends in Measuring
Environmental Attitudes: Measuring Endorsement of the New Ecological Paradigm: A Revised NEP Scale. Journal of Social Issues, 56(3): 425-442.
Lo, S., Smith, S., Taylor, M., Good, A., & von Wagner, C. 2012. The Effect of Temporal Framing on Behavioral Intentions, Expectations, and Behavior: The Case of Healthy Eating. Journal of Applied Biobehavioral Research, 17(3): 202-213.
Pacific Institute. 2014. The World's Water Vol. 8. https://www.worldwater.org.
Pandelaere, M., Briers, B., & Lembregts, C. 2011. How to Make a 29% Increase Look Bigger: The Unit Effect in Option Comparisons. Journal of Consumer Research, 38(2): 308-322.
Pelham, B., Sumarta, T., & Myaskovsky, L. 1994. The Easy Path from Many to Much: the Numerosity Heuristic. Cognitive Psychology, 26(2): 103-133.
Petty, R., & Cacioppo, J. 1986. The Elaboration Likelihood Model of Persuasion. Advances in
Experimental Social Psychology, 123-205.
Petty, R., Cacioppo, J., & Goldman, R. 1981. Personal involvement as a determinant of argument-based persuasion. Journal of Personality and Social Psychology, 41(5): 847-855.
Raghubir, P., & Srivastava, J. 2002. Effect of Face Value on Product Valuation in Foreign Currencies. Journal of Consumer Research, 29(3): 335-347.
Reczek, R., Trudel, R., & White, K. 2018. Focusing on the forest or the trees: How abstract versus concrete construal level predicts responses to eco-friendly products. Journal of
Environmental Psychology, 57: 87-98.
Robinson, M., Cassidy, D., Boyd, R., & Fetterman, A. 2015. The politics of time:
27
SaveTheWater. 2020. Water Facts. Savethewater.org. https://savethewater.org/water-facts/. Schultz, P., Nolan, J., Cialdini, R., Goldstein, N., & Griskevicius, V. 2007. The Constructive,
Destructive, and Reconstructive Power of Social Norms. Psychological Science, 18(5): 429-434.
Seyranian, V., Sinatra, G., & Polikoff, M. 2015. Comparing communication strategies for reducing residential water consumption. Journal of Environmental Psychology, 41: 81-90.
Trope, Y., & Liberman, N. 2010. Construal-level theory of psychological distance. Psychological Review, 117(2): 440-463.
United States Government Accountability Office. 2014. Freshwater: Supply Concerns
Continue, and Uncertainties Complicate Planning. Washington, DC: United States
Government Accountability Office.
Vallacher, R., & Wegner, D. 1989. Action Identification Theory. Handbook of Theories of Social Psychology: Volume 1, 327-348.
White, K., Macdonnell, R., & Dahl, D. 2011. It is the Mind-Set that Matters: The Role of Construal Level and Message Framing in Influencing Consumer Efficacy and Conservation Behaviors. Journal of Marketing Research, 48(3): 472-485. WorldoMeter. 2015. Water Use Statistics - Worldometer. Worldometers.info.
https://www.worldometers.info/water/.
Xu, X., Arpan, L., & Chen, C. 2015. The moderating role of individual differences in responses to benefit and temporal framing of messages promoting residential energy saving. Journal of Environmental Psychology, 44: 95-108.
Yalch, R., & Elmore-Yalch, R. 1984. The Effect of Numbers on the Route to Persuasion. Journal of Consumer Research, 11(1): 522.
28
Appendices
Appendix A
Below structure of the conducted survey: Intro & Consent Form
Part I: Main Study
“All around our country, freshwater shortages are becoming more and more common. Scientists and policymakers are increasingly worried about individuals' and households' overconsumption of water, and accordingly, new initiatives have begun to protect our drinking water security in the future.
Recently, a new NGO was formed to do exactly that. SaveFreshWater has launched a factual advertising campaign to increase the public’s awareness of our water shortages due to
household consumption.
On the next page, you will see one of their latest printed advertisements. Please take your time to see the ad. You will then answer a few questions about the ad and your opinion of it.” [4 different advertisements, randomly assigned]
29
Figure 3: Scenario 3 (day frame, numerosity effect) Figure 4: Scenario 4 (year frame, numerosity effect) [measuring the processing time of the scenario ]
Perception of consumption magnitude (1 = strongly disagree, 7 = strongly agree) (6) 1. I believe the average American consumes too much water
2. I believe future freshwater shortages are inevitable
3. In my opinion, freshwater shortages will have a noticeable negative impact on my health (over the next 25 years)
4. In my opinion, freshwater shortages will have a noticeable negative impact on my economic and financial situation (over the next 25 years)
5. I believe a future with freshwater shortages will soon happen 6. I am deeply concerned about freshwater shortages
Pro-Environmental Behavioral Intentions (1= very unlikely, 7= very likely) (9) 1. I am willing to create a plan to use less water
2. I am willing to take shorter showers
3. I am willing to install water-saving shower heads
4. I am willing to turn off the water while brushing my teeth 5. I am willing to turn off water while shaving
6. I am willing to use my dishwasher or washing machine for full loads only 7. I am willing to turn off the water while doing dishes by hand
8. I am willing to install flow restrictors on all my faucets
30 Willing to donate time (1)
“People often have little knowledge about actions they can take to help reduce the freshwater shortage. SaveFreshWater aims to increase public awareness by disseminating customized guidelines through social media, email lists, and other online platforms. The NGO needs volunteers’ help to increase its content reach and its information dissemination speed. Using the slider below, please indicate what percentage of your free time you would be willing to donate to help the NGO with this mission.”
Scale of 0 – 100% Willingness to receive email (1)
As a token of appreciation for your participation in this survey and for sharing your opinions with us, we would like to provide you with easy tips and tricks to save water at home. Would you be willing to receive an email from us with this information after the survey?
Yes or No Part II: Moderators
Ecological Value Scale (6) (1 = strongly disagree, 7 = strongly agree)
1. Humans have the right to modify the natural environment to suit their needs 2. Humans are severely abusing the planet
3. Plants and animals have the same right to exist as humans
4. Nature is strong enough to cope with the impacts of modern industrial nations 5. Humans were meant to rule over the rest of nature
6. The balance of nature is very delicate and easily upset CFC (6) (1 = strongly disagree, 7 = strongly agree)
1. I consider how things might be in the future
2. I am willing to sacrifice now in order to achieve future outcomes
3. I think it is important to take warnings about negative outcomes seriously even if the negative outcome will not occur for many years
4. I mainly act to satisfy my immediate concerns, figuring the future will take care of itself
31
6. I only act to satisfy immediate concerns, figuring that I will take care of future problems that may occur at a later date
Geographical Area and Personal Experience (4) (yes or no) 1. I have experienced freshwater shortages before
2. I live in an area where freshwater shortages are common 3. I have family that experienced freshwater shortages before 4. I have lived under drought restrictions before
Perceived Accuracy and Believability (4) (1 = strongly disagree, 7 = strongly agree) 1. I perceived the information provided to me as accurate
2. I perceived the information provided to me as believable 3. I perceived the information provided to me as credible
4. I perceived the information provided to me as personally relevant Concrete or Abstract Mindset (10)
“In the following, you will see 10 different kinds of actions followed by two alternatives. Please indicate which alternative describes the behavior the best. There are no right or wrong answers, we are only interested in your opinion.”
1. Making a list
a. Getting organized b. Writing things down 2. Reading
a. Following lines of print b. Gaining knowledge 3. Eating
a. Getting nutrition
b. Chewing and Swallowing 4. Cleaning the house
a. Showing one’s cleanliness b. Vacuuming the floor 5. Paying the rent
a. Maintaining a place to live b. Writing a check
6. Growing a garden a. Planting seeds
32 7. Locking a door
a. Putting a key in the lock b. Securing the house 8. Chopping down a tree
a. Wielding an axe b. Getting firewood 9. Traveling by car
a. Following a map b. Seeing the countryside 10. Picking an apple
a. Getting something to eat b. Pulling an apple off a branch Demographics (23)
Affect; age; gender; education level; employment status; ethnicity; family size; personal income; household income; political orientation; Conservatism; Leaving computer during survey; alone during survey; musical distractions during survey; quietness surrounding; cellphone use during survey; type of device used in survey; level of English; attentiveness; experience on MTurk; confusion during survey.
Attention Checks (4)
1. “Please indicate whether you have experienced any of the following events in your life.” 1. Stung by a wasp
2. Dancing in a party 3. Playing cards
4. Suffered a fatal heart attack 5. Reading a novel
6. Answering a survey
7. Bitten by a great white shark
2. In the advertisement that you read at the beginning of the survey, what was the name of the NGO?
3. In the advertisement that you read at the beginning of the survey, in what way was the amount of water consumed described?
4. Please tell us what day of the week was yesterday and briefly write about what you did. Suggestions and code
33
Appendix B
Reliability Analysis Variables
Measure Cronbach’s
Alpha
Number of Items Perception of Problem Magnitude – Concern ,882 6
Behavioral Intentions ,870 9
Ecological Value ,795 6
Action Identification Task – Concrete vs Abstract Mindset ,806 10
Consideration of Future Consequences ,826 6
Personal Experience ,740 4
Appendix C
34
Appendix D
Models conceptualizing significant Hayes’ Process Model 8 analyses:
Figure 1. Conceptualization of moderator CFC on the effect of temporal framing on average concern. Average concern predicts Behavioral Intentions and Willingness to Donate Time.
Figure 2. Conceptualization of the moderator Conservatism on the effect of temporal framing on average concern. Average concern predicts Behavioral Intentions.
35
Appendix E
Results Hayes’ Process Model 8 and 15 – The effects of Temporal Framing on Behavioral Intentions and Time Donation:
Model 8 Y X M W F Model Pinteraction Behavioral Intention Temporal Framing Concern CFC 27,8525 ,0439** Behavioral Intention Temporal Framing Concern Political 16,8955 ,2463 Behavioral Intention Temporal Framing
Concern Eco Value 48,0711 ,4336 Behavioral Intention Temporal Framing Concern AIT 1,9280 ,0937* Time Donation Temporal Framing Concern CFC 27,8525 ,0439** Time Donation Temporal Framing Concern Political 16,8955 ,2463 Time Donation Temporal Framing
Concern Eco Value 48,0711 ,4336 Time Donation Temporal Framing Concern AIT 1,9280 ,0937* Model 15 Y X M W F Model Pinteraction Behavioral Intention Temporal Framing Concern CFC 55,4008 ,1457 Behavioral Intention Temporal Framing Concern Political 45,1044 ,4895 Behavioral Intention Temporal Framing
Concern Eco Value 49,8941 ,2560 Behavioral Intention Temporal Framing Concern AIT 45,7856 ,6690 Time Donation Temporal Framing Concern CFC 9,6273 ,8310 Time Donation Temporal Framing Concern Political 10,7392 ,1322 Time Donation Temporal Framing
Concern Eco Value 8,5064 ,7368 Time
Donation
Temporal Framing
36
Appendix F
Results Hayes’ Process Model 8 and 15 – The effects of Numerosity on Behavioral Intentions and Time Donation:
Model 8 Y X M W F Model Pinteraction Behavioral Intention Numerosity effect Concern CFC 10,5807 ,4361 Behavioral Intention Numerosity effect Concern Political 10,6920 ,9788 Behavioral Intention Numerosity effect
Concern Eco Value 43,9251 ,4726 Behavioral Intention Numerosity effect Concern AIT 3,5986 ,7303 Time Donation Numerosity effect Concern CFC 10,5807 ,4361 Time Donation Numerosity effect Concern Political 10,6920 ,9788 Time Donation Numerosity effect
Concern Eco Value 43,9251 ,4726 Time Donation Numerosity effect Concern AIT 3,5986 ,7303 Model 15 Y X M W F Model Pinteraction Behavioral Intention Numerosity effect Concern CFC 47,4348 ,5920 Behavioral Intention Numerosity effect Concern Political 35,1264 ,4883 Behavioral Intention Numerosity effect
Concern Eco Value 37,7300 ,9366 Behavioral Intention Numerosity effect Concern AIT 36,1826 ,5890 Time Donation Numerosity effect Concern CFC 12,4080 ,2361 Time Donation Numerosity effect Concern Political 12,8078 ,7146 Time Donation Numerosity effect
37
Appendix G
Our data indicated a significant effect of numerosity on the willingness to donate time to charity. We did not include this result in our Results section, as we decided to focus our research on Temporal Framing rather than the Numerosity Effect. However, based on our study, we can say that framing a message as “88 gallons every day” significantly increases (R2 = ,008; F(1,343)=2,928; p=0,088*) the amount of time an individual is willing to donate to
the cause compared to framing a message as “32,120 gallons every year”.
38
Appendix H
ANOVA results reveal the effect of Concern on Behavioral Intentions and Willingness to Donate Time.
The effect of Average Concern on Behavioral Intentions(R2=,450; F(33,344)=7,684; p=,000***):
39
Appendix I
Our data suggests that CFC is a very important and significant moderator in the effect temporal framing has on average concern. However, we observed a “ripple effect” of CFC. Levels of CFC can be explained for the reason why three other moderators partially work as well: Conservatism, Gender, and Educational Level. We performed ANOVA’s in order to determine these connections, and created the following graphs:
CFC and Conservatism
ANOVA Result: R2= ,046; F(6,682)=5,489; p=,000***
CFC and Educational Level
40 CFC and Gender
ANOVA Result: R2=,019; F(2,344)=3,380; p=,035**
Appendix J
41
Appendix K
Complete Overview of Demographics in Survey Sample Results 1
Demographic Subcategories N (total 689) % Cumulative %
Age 18-30 187 27,1 27,1 31-40 233 33,9 61,0 41-50 116 16,8 77,8 51-60 80 11,6 89,4 61-70 63 9,1 98,5 71-75 10 1,5 100,0 Gender Male 303 44,0 44,0 Female 385 55,9 99,9 Other 1 ,1 100 Education Level
Less than high school 4 ,6 ,6
High school diploma 140 20,3 20,9
43
Appendix L
Complete Overview of Demographics in Survey Sample Results 2
Demographic Subcategories N (total 344) % Cumulative %
Age 18-30 98 28,5 28,5 31-40 113 32,8 61,3 41-50 56 16,3 77,6 51-60 38 11,1 88,7 61-70 33 9,6 98,3 71-75 6 1,7 100 Gender Male 150 43,6 43,6 Female 193 56,1 99,7 Other 1 0,3 100 Education Level
Less than high school
2 ,6 ,6
High school diploma 74 21,5 22,1
52
