Goals and Fuel Consumption:
Goal-Framing as a Way to Achieve a Reduction in Fuel Usage and the Moderating Role of Perceived Occupational Stress
Julia Schlegelmilch (S 1820583)
February 28, 2014
Master Thesis, Human Resource Management Faculty of Economics and Business
University of Groningen
Supervised by: Dr. Jennifer Jordan Secondary evaluator: Tim Vriend
Dedication
This master’s thesis is dedicated to my parents, who always encouraged me to pursue my goals and dreams and have given me the freedom to do so with their endless love and support.
Acknowledgement
I would like to thank my supervisor, dr. Jennifer Jordan, for her helpful remarks, knowledge and engagement during the learning process of writing this master thesis.
Table of Contents
List of Figures 6!
List of Tables 7!
Abstract 8!
Introduction 9!
Conceptual Framework and Hypotheses 10!
Goal-Framing and Performance 11!
Moderating Effect of Stress 13!
Pilot Study 16!
Method 16!
Participants 16!
Procedures 17!
Interview Questions 17!
Results 17!
Job and Professional Expertise 17!
Goals 18!
Stress 19!
Discussion and Conclusion 19!
Main Study 21!
Method 21!
Participants and Design 21!
Procedures 22!
Measures 24!
Analyses 26!
Results 27!
Preliminary Analysis 27!
Manipulation Check 29!
Repeated Measures Analyses Fuel Usage 30!
Behavioral Intentions 33!
General Discussion 34!
Reflection on the Results and Theoretical Implications 34!
Limitations and Future Research 38!
Practical Implications 40!
Methodological Implications 40!
Conclusion 41!
References 43!
Appendix A: Interview Questions 54!
Appendix B: Questionnaires 55!
Appendix C: Goal-Frame Messages 56!
List of Figures
Figure 1. Conceptual Model with Hypothesized Relationships ... 16!
Figure 2. Graphical Display of Interaction (Goal-frame x Stress) for Manipulation Check. .. 30!
List of Tables
Table 1 - Overview of Procedures During the Main Study ... 22!
Table 2 - Overview of Inter-Item Correlations of Stress Measures. ... 28!
Table 3 - Paired-Samples T Tests for All Stress Scale Items. ... 28!
Table 4 - Regression Coefficients, Standard Errors, and R2 for Manipulation Checks. ... 29!
Table 5 - Correlation Matrix of Variables ... 31!
Table 6 – Repeated Measures Analysis of Variance for Grade. ... 32!
Table 7 – Repeated Measures Analysis of Variance for Usage. ... 32!
Table 8 – Repeated Measures Analysis of Variance for Over-revving. ... 33!
Table 9 – Repeated Measures Analysis of Variance for Coasting. ... 33!
Table 10 - Regression Coefficients, Standard Errors, and R2 for Behavioral Intentions. ... 34!
Abstract
Public transportation is a large-scale source of fuel consumption, which is a costly issue for both economic and environmental reasons. The current experiment attempted to achieve consistent fuel conservation behavior through manipulating goal-frame either in terms of progress or commitment (Fishbach & Dhar, 2005). It was conducted in the field with 24 professional bus drivers. I also examined if perceived occupational stress moderated this relationship. The results neither provided support for the goal-framing approach in achieving consistent fuel conservation behavior, nor for the moderating role of stress. I provided possible explanations for the non-significant results, directions for future research as well as theoretical, practical and methodological implications.
Keywords: goal-framing, commitment, progress, behavioral change, bus drivers, occupational stress, fuel conservation, fuel consumption.
Introduction
Public transportation is a large-scale consumer of fuel. The price for diesel fuel has risen by 27.7%, 21.9% for gas respectively, in the Netherlands over the last six years (Centraal Bureau voor de Statistiek, 2013). In view of this development, fuel consumption creates challenges, both environmentally and economically. Besides the fuel used in total, the
“excess fuel” is of particular interest because it is variable. Excess fuel is defined as the amount of fuel used above the level of what is absolutely necessary when driving fuel- efficiently. This thesis examines a potential method for reducing the excess fuel usage of drivers in the public transportation sector. Specifically, I take a goal-framing approach (Fishbach & Dhar, 2005) to examine how to reduce behaviors related to excess fuel
consumption and look at how the person-related variable, occupational stress, moderates the goal-framing effects.
Reducing excess fuel usage is important for economic and environmental reasons.
Firstly, the costs associated with excess fuel are a major entry in a public transportation company’s balance sheet, and therefore an important factor in determining a company’s ability to compete in the marketplace. This is of great concern for Dutch public transportation providers because they have to compete with each other for the permit to be the public transportation provider for a certain number of years. Hence, if excessive fuel usage were reduced, related expenses would be reduced and the ability to compete increased. Secondly, the general awareness of climate change, the necessity of resource conservation, and the emphasis on ‘going green’ has increased during the last decades. In this context, the
detrimental effects of fuel usage for the environment, such as contributing to climate change and destruction of habitats, have received great attention (e.g., Chow, Kopp, & Portney, 2003). Reducing the fuel usage in public transportation contributes to the reduction of society’s carbon footprint, and consequently its negative environmental effects.
Nader (1991) established that fuel consumption is the outcome of specific features of driving behavior, thus the drivers can exert influence on fuel consumption. For example, behaviors like breaking hard, accelerating fast, and coasting all individually contribute to excessive fuel consumption. Within the corporate domain (meaning, not referring to individual consumer drivers), a first step towards reducing excess fuel usage is recognizing the importance of an eco-friendly driving style (Stillwater & Kurani, 2013) and making fuel conservation a company-wide goal. A second step is researching ways to achieve this goal and in this thesis, I focus on the second step. Specifically, I examine if it is possible to encourage a more fuel-efficient driving style by altering the way the drivers view the goal of fuel conservation. Moreover, I examine the boundary conditions of the proposed relationship by investigating if perceived occupational stress influences it. I focus on occupational stress because the bus drivers face many demands in their job, which has been found to be stressful (Bartone, 1989; Carrère, Evans, Palsane, & Rivas, 1991; Kühlmann, 1990; Tse, Flin, &
Mearns, 2006). While stress evidently affects performance (Jex, 1998; Matthews & Wells, 1996; McGrath, 1976), there is a gap in the literature with regard to its influence on goals (Elliot, Thrash, & Murayama, 2011; Gaudreau, Carraro, & Miranda, 2012), which makes it important to invest research time in it.
Conceptual Framework and Hypotheses
Previous research regarding behavioral change and fuel conservation focused on several different approaches to achieve a reduction in fuel usage. Duarte, Gonçalves and Farias (2013) used visual and acoustic feedback to alert the driver in real-time of undesirable events, for example, hard acceleration or excessive speed, based on on-board sensors. Af Wåhlberg (af Wahlberg, 2002) used training sessions and van der van der Voort (2001) used simulation sessions. All of these approaches have been successful in achieving varying degrees of fuel conservation. However, the focus of these studies was either technical- or
training-oriented, and lacked an arguably necessary psychological component, such as environmental self-identity (van der Werff, Steg, & Keizer, 2013) or social influence (Abrahamse & Steg, 2013). The present study will approach behavioral change from a goal- framing perspective. Thereby, I will add to the existing literature by offering a solution to achieve fuel conservation behavior (combining goal-framing and behavioral change) and filling the present psychological gap in this field. Furthermore, the proposed influence of stress on the effects of goal-framing will provide an understanding of the boundary conditions, thereby providing a nuance to this potential relationship.
Goal-Framing and Performance
Goals can be defined as, “concrete cognitive representations that serve a directional function” (Elliot & Thrash, 2001, p. 143) via the route of motivation, which is the process that initiates and guides goal-oriented behaviors (Locke & Latham, 1990). People’s behavior, though, is realistically assumed to be driven by multiple goals (A. N. Dalton & Spiller, 2012;
Lindenberg & Steg, 2007), which are related to different parts of their identities. With regard to bus drivers, these goals can range from striving to be a good bus driver, which targets the individual’s job role identity, to being good employee, which targets the individual’s identity as a member of the organization.1
However, not only the goal itself, but also the way goals are framed influence subsequent behavior because a goal-frame is “the way in which people process information and act upon it” (Lindenberg & Steg, 2007, p. 118). Previous research found that framing goals in terms of avoidance versus approach (Roney & Lehman, 2008) or extrinsic versus intrinsic (Vansteenkiste, Timmermans, Lens, Soenens, & Van den Broeck, 2008) was influential to people’s behaviors related to approaching these goals. Fishbach and Dhar
1 In order to get a better understanding of drivers’ actual goals, I conducted a Pilot Study to examine their goals.
More details on this are contained in the Methods Section.
(2005) framed goals in terms of commitment and progress, which they argued have differential influences on one’s motivation to pursue a focal goal. The progress focus emerged from variety-seeking research (e.g., Fishbach, Ratner, & Zhang, 2011; Ratner, Kahn, & Kahneman, 1999), which suggested that engaging in actions related to the goal was interpreted as a justification for an individual to pursue other goals instead of the initial, focal goal (Fishbach, Dhar, & Zhang, 2006). The underlying reasoning for this was that with a progress focus, one engaged in balancing: the dynamic to disengage from a goal in which one had already invested, resulting in decreased motivation and the pursuit of a goal in which one has not yet invested (Koo & Fishbach, 2008). On the other hand, the commitment focus was rooted in behavioral consistency research (e.g., Laran & Janiszewski, 2009) and suggested that engaging in goal-related actions was interpreted in terms of increased motivation towards more goal-consistent actions (Fishbach et al., 2006). Here, the underlying reasoning for consistency was that with a commitment focus, one engaged in highlighting: the tendency to act consistently with previous actions because of the initial investment (Koo & Fishbach, 2008), which signaled high commitment and in turn led to increased motivation towards the goal. Previous research by Fishbach and Dhar (2005) regarding the differential effects of the foci, goal commitment and progress, found supporting evidence. For example, they provided participants with information on goals (relevance was established in a pilot study; Fishbach &
Dhar, 2005, Study 3). After imagining the pursuit of these goals, the participants in the respective framing condition were asked to evaluate either their level of commitment or their level of progress (based on the experimental condition to which they were assigned). They found that people were more likely to choose a subsequent goal-incongruent action in the progress focus group, than in the commitment framing group (Fishbach & Dhar, 2005).
Additionally, they found that people who focused on goal progress did not need to make actual progress toward the goal; perceived progress had the same effect of reducing people’s
subsequent behavior in support of the goal due to overoptimistic evaluations (Fishbach &
Dhar, 2005, Study 4).
Therefore, I hypothesize that regarding the goal of fuel conservation, a bus driver with a progress focus will be more motivated to engage in goal-incongruent actions, such as braking hard or driving recklessly, thereby increasing fuel consumption. On the contrary, I expect that a bus driver with a commitment focus will be more motivated to engage in goal- congruent actions, such as coasting, thereby decreasing fuel consumption.
Hypothesis 1: Framing previous driving behavior in terms of commitment will
subsequently lead to consistent fuel conservation behavior, whereas framing previous driving behavior in terms of progress will subsequently lead to less consistent fuel conservation behavior.
Moderating Effect of Stress
In order to increase the informative value of the study, it is important to explore the boundary conditions of the first hypothesis. A commonly researched factor in the work context is occupational stress and it has been established that the job of a bus driver is a stressful one (Bartone, 1989; Meijman & Kompier, 1998; Tse et al., 2006). Occupational stress is defined in the current study as a mismatch of job demands with the individual’s capacities (Jamal, 2005; McGrath, 1976). Hence, job demands are partly the cause of
perceived stress (Lang, Thomas, Bliese, & Adler, 2007; Rubino, Perry, Milam, Spitzmueller,
& Zapf, 2012) and the demands of a bus driver are, amongst others, adhering to a schedule (time demand), taking the traffic situation and regulations (safety demand) as well as the passengers (passenger demand) into account (Meijman & Kompier, 1998).
It was suggested that the influence of stress on performance is positive, or U-shaped, but most studies have found a negative influence (e.g., Gilboa, Shirom, & Fried, 2005; Van Dyne, Jehn, & Cummings, 2002; Wu, 2011). Negative stress (in contrast to eustress or good
stress2; Selye, 1974) has a negative influence on one’s performance (Jex, 1998) and evidence is far spread, for example, for self-appraised performance (Wu, 2011) or performance on cognitive processing tasks (Lawrie, 2000). The impairment with regard to the latter is due to
“lapses in […] concentration” (Chen & Kao, 2013, p. 105), which have been found to lead to aberrant driving behaviors, which in turn increase fuel consumption (Lan & Kuo, 2004).
Therefore, I expect that drivers’ perceived occupational stress will impair their performance.
Hypothesis 2: Stress has a negative influence on fuel conservation behavior.
The literature is scarce with regard to the specific influence of occupational stress on the relationship of goal-framing (and goals, in general) to performance. This is reflected by Gaudreau, Carraro, and Miranda’s (2012) note that this relationship remains “relatively underexplored in psychological science” (p. 512). In agreement, Elliot, Thrash and
Murayama (2011) stated, “aside from a generic acknowledgement, […] a general recognition […], theorists have allocated little attention to goals” in the stress literature (p. 644).
However, this gap in the literature is disconcerting in view of the negative influence occupational stress can have on workers (Gilboa, Shirom, Fried, & Cooper, 2008; Van Dyne et al., 2002; Wu, 2011), and the importance of goals as a directional function for behavior (Elliot & Thrash, 2001; Locke & Latham, 1990). Due to the lack of research for the described relationship, I will draw on the literature of a close brethren of goals, namely motivation (Koo & Fishbach, 2008, 2012; Locke & Latham, 2002; Orehek, Bessarabova, Chen, &
Kruglanski, 2011), as motivation is the mechanism via which goals serve their directional function (Elliot & Thrash, 2001).
Wells and Matthews (1996) stated that occupational stress affects motivation, and subsequently performance. Similarly, the model by McGrath (1976) suggested that occupational stress reduced employees’ willingness to put effort (motivation) in the job,
2 Although I do not have the space to discuss these topics in detail, eustress or “good stress” is defined as stress
that “stimulates and enhances performance” (Johnston & Lee, 2012, p. 34).
which led employees to engage in behaviors that decreased job performance. Furthermore, role overload, an antecedent of occupational stress (Kahn, Wolfe, Quinn, & Snoek, 1964), was likely to reduce motivation because it was difficult for employees to overcome, even when exhibiting extra effort (Clarke, 2012). Also, role overload was found to decrease the motivational benefits of self-efficacy (Brown, Jones, & Leigh, 2005). And finally, with regard to the specific driving context, Matthews and Desmond (2002, p. 681) concluded after two experiments that “interventions should be geared towards enhancing […] motivation.”
Hence, the above-described literature provides supporting evidence for the negative effect of stress on motivation, and therefore also goal-framing, which I propose in turn will affect the relationship with performance. Thus, in the current study, I seek to extend the literature on stress and motivation by examining how stress interacts with goal-framing.
Since the two goal-frames are predicted to have differential effects on performance, I hypothesize that stress will interact with each of the two goal-frames. Specifically, given stress’ role in affecting motivation, stress should affect goal-framing negatively. I expect that that a driver with a commitment frame who experiences occupational stress will be less motivated to engage in goal-consistent actions due to reduced motivation towards the focal goal, and therefore engage in more goal-inconsistent actions such as braking hard. Similarly, a driver with a progress frame who experiences occupational stress will be even less
motivated to engage in goal-consistent actions because of reduced motivation towards to focal goal, hence engage in more goal-inconsistent actions such as accelerating fast and over- revving.
Thus, stress is a viable variable to add to the current study because not only has it been shown to affect a variable closely tied to goals, but it is prevalent amongst bus drivers (Bartone, 1989; Meijman & Kompier, 1998; Tse et al., 2006). I propose stress will interact with one’s goal-framing to affect actual behavior. Specifically, I hypothesize the following:
Hypothesis 3: Perceived stress moderates the relationship between goal-frame and fuel conservation behavior, such that it exacerbates the effect of a progress frame, but dampens the effect of a commitment frame.
Figure 1. Conceptual Model with Hypothesized Relationships
I conducted an experiment in the field to test these hypotheses. Before doing so, I conducted a Pilot Study in order to explore the bus drivers’ general goals and attitudes toward the job, as well as the focal goal of fuel conservation. I considered understanding these components necessary to framing my experimental manipulations in the Main Study, gaining an understanding of the population, as well as conveying my sincere interest in the bus drivers, as they would be participants in the Main Study.
I focused my data collection on a large public transportation company in the
Netherlands. This organization reported to spend around 10.7 million Euros per year solely on fuel, which equals 12% of the annual expenses. Thus, the question of methods for reducing fuel consumption was particularly cogent. The company provides public
transportation on regional and urban routes in several provinces of the Netherlands with its fleet of 460 buses, of which almost three quarters belong to the standard city configuration (12m). The total number of employees was 1070 at the time of the data collection.
Pilot Study Method Participants
Progress
Commitment
Fuel consumption
Perceived stress H1 (-/+)
H3 (-)
H2 (-) Goal frame
In total three male Dutch bus drivers were interviewed. These people’s ages ranged from 43 to 55 years with tenures at the organization ranging between 5 and 25 years. These three people were selected based on the recommendation of their team leader, who regarded them to be representative of the opinions of the bus drivers at the particular location.
Participation was voluntary.
Procedures
The interviews were conducted with each driver separately in a meeting room in the building where the drivers worked at a time that suited them best, thus before or after their shift. Each interview was preceded by a brief explanation, where I informed the participant that the interview would take about 20 minutes and that the purpose was to explore their work and attitudes. Finally, the interviewees were asked if they agreed that the interview be
recorded; all agreed. Afterwards, each interviewee could choose between a piece of fruit or cookies as compensation.
Interview Questions
The interview was semi-structured and consisted of ten questions (Appendix A) concerning three topics: the bus driver’s daily job and expertise (“Is your expertise important to you?”), goals, in general (“What are your goals as a bus driver?”) and regarding fuel conservation (“Do you think fuel conservation is important?”). Finally, some questions concerned the experience of occupational stress (“Do you think your job is stressful?”).
Results Job and Professional Expertise
The first question about their working day elicited similar descriptions of work routines and mentioned various duties, such as checking for “changes in the route” or a
“sufficient number of available tickets”. Regarding the effects of the irregular work schedule, two interviewees stated that it “did not bother” them and one stated that it was “sometimes
inconvenient”. When encouraged to elaborate, two of them said that they had “been a bus driver for a long time” and the other interviewee “did not know how to elaborate more”. The next question regarding the reasons for becoming a bus driver revealed that the career paths of the interviewees differed, though they shared that initially none of them planned on being a bus driver. With regard to the final question about professional expertise, all stated that their expertise was important to them.
Goals
When asked about the characteristics of a good bus driver, all interviewees mentioned the following characteristics: “Friendliness towards the passenger” and “behaving
responsibly”. One interviewee also stated “being on time” as a characteristic. In addition, two of them said that they “fulfilled” these characteristics. With regard to the second question (about their goals as a bus drivers), two interviewees indicated, “not to understand” the question, thus goal was substituted by “being professionally important to you”. All
interviewees mentioned that a “sense of responsibility for the bus and the passengers” as well as the “relationship with the colleagues” were important to them on the job. Two
interviewees mentioned that “fulfilling the duties of a bus driver” was important to them.
With regard to the fourth question, if they can think of undesirable tasks they have to
perform, responses such as “not really” or “I cannot think of any at the moment” were given.
In turn, everyone was willing to share their views about fuel-efficient driving at the company (final question). All interviewees stated to have a “positive” attitude towards a training session they had had about three months prior (and unrelated) to the interview. One driver stated that the focus on fuel-efficient driving was “interesting” and “even discussed” amongst the bus drivers in the beginning, but that it had “lost its appeal with time”. Two interviewees also described it as “just another idea of a company”. One interviewee stated that it was difficult to incorporate the fuel-efficient driving style because the “traffic and route
circumstances did not always allow” it. One interviewee would recommend driving fuel efficiently to colleagues.
Stress
All three indicated explicitly that they were neither stressed about their jobs, nor perceived stress while working. One mentioned to have “backaches”.
Discussion and Conclusion
The interviews resulted in two important conclusions, which are discussed in the following paragraphs. First, the concept of goals is not prevalent, as is evident in the
observation that the question about goals had to be rephrased. It is rather that the bus drivers deem it important to be a good bus driver (expertise), which differs from the initial
assumption that they aim to be a good employee as they identified themselves as (proud) bus drivers, not as employees. The latter may be due to the fact that while the employer changes every number of years, the group of bus drivers stays, in turn leading to, on average, long tenures. Similarly, it may have played a role that the interviewees did not aim to become a bus driver at the beginning of their career path. It also became apparent that expertise was important, which seems to be tied the profession rather than the job or the company. This indicated that the organizational commitment may be less prevalent, but being a bus driver (or belonging to the group of bus drivers) seems to be part of a social identity following social identity theory (SIT) (Tajfel, 1982). SIT proposes that a common threat (changing employers) and a common history (long tenures) are contributing factors (Ashforth & Mael, 1989). A social identity also fosters self-esteem (Hogg & Turner, 1985), hence the
identification as proud bus drivers.
Secondly, I conclude that fuel-efficient driving is not the top priority of the drivers.
Congruent responses were given by all interviewees, such as acting upon the sense of
responsibility about the passengers and the bus, as well as fulfilling duties. One could assume
that fuel-efficient driving is implied in “fulfilling duties”, though it is crucial to note that it was not explicitly stated. This is in congruence with the Dutch National Job Guide’s description of a bus driver, which does not state driving fuel efficiently as a competency either (Nationale Beroepengids, 2014). One reason may be that driving fuel efficiently is relatively novel for the bus drivers and not yet accepted as a the norm, which would be necessary to serve as a behavioral standard according to SIT (Christensen, Rothgerber, Wood, & Matz, 2004). Also, previous research has found that one’s self-identity is an
“important predictor of environmental actions” (van der Werff et al., 2013, p. 1258).
Additionally, the goal (fuel-efficient driving) is associated with the employer (referred to
“just another idea of a company"). Finally, very recent research found that people are more likely to exhibit environmentally friendly behaviors, such as fuel-efficient driving, if they act under the perception that an entity, e.g. a company, has a long past, which infers a long future (Hershfield, Bang, & Weber, 2014). However, in the current study the employer changes every few years, which may create the bus drivers’ perception of a short past, hence they are less likely to act environmentally friendly.
Finally, the interviewees stated not to experience stress, which contradicts previous research on stress amongst bus drivers over the years (e.g., Bartone, 1989; Carrère, Evans, Palsane, & Rivas, 1991; Kühlmann, 1990; Querido, Nogueira, Gama, & Orlando, 2012;
Winkleby, Ragland, Fisher, & Syme, 1988). A likely explanation is that the social desirability bias was present, which is the “phenomenon of respondents seeking to present themselves in a favorable manner” (Thompson & Phua, 2005, p. 541) and consequently, the interviewees engaged in impression management, which is the attempt to regulate and control the
information in social interaction (Piwinger & Ebert, 2001), thus trying to appear less stressed than they actually were.
Thus, in conclusion, these responses helped me to determine what the bus drivers
found important, namely being a good bus driver, which was necessary in order to choose what could be used as potential goals in the Main Study. However, in the process, it became clear that I should not use ‘goal’ in the communication during the intervention, as the
interviews gave the impression that the concept may be unintelligible. Finally, it helped me to explore the bus drivers’ attitudes towards fuel-efficient driving, as well as to discover that they were likely to give socially desirable answers when their level of experienced stress was concerned. The latter was of concern, as the Main Study includes stress measures.
Main Study Method
In the Main Study, I manipulated the independent variable goal-frame and measured how this goal-framing affected several indicators of the dependent variable fuel consumption.
Additionally, I investigated how perceived stress affected fuel consumption, as well as moderated the hypothesized relationship between goal-frame and performance.
Participants and Design
The participants were randomly selected from all the bus drivers working on at a particular location of a large public transportation company in the north of the Netherlands.
All 51 drivers at the site were invited to participate however, only 24 agreed to participate (47% response rate).3 Of these 24 bus drivers (8.33% female, Mage = 52.5 years, SD = 8.64), eighteen were full-time employees and six were temporary workers.
A one-way between-subjects design was used in which participants were randomly assigned to one of two goal-frame conditions (Progress vs. Commitment). In view of the small sample size (i.e. n = 24), I resigned from including a control condition. There were 12 participants in each condition and all were told that in exchange for participating, they would
3 There are several reasons assumed for the low participation rate. First, participation in the study was voluntary.
Second, a disturbing event occurred and is assumed that it affected trust and commitment of the bus drivers in the organization.
receive entry to a raffle for two €50 prizes. Participation was voluntary and anonymous.
Procedures
The procedures that took place over the course of ten weeks are described in this paragraph and an overview can be found in Table 1.
A week before the preparation for the study began, a notice was placed on the company’s news board to inform all employees about the upcoming study and that
participation in the study would be appreciated. Then, (after the Pilot Study), all bus drivers received an envelope in their personal locker that included a standard informed consent form, an information sheet about the study, as well as the Stress Questionnaire. Over the course of the next five weeks until the experiment manipulations were introduced, several reminders were distributed in the two most frequently attended places (i.e., the canteen and locker room), which asked the drivers to take part in the study. Furthermore, I was present at the location at least twice a week to foster trust and answer any arising questions. In addition, a box with the printed-on title, “hand-in box”, was deposited in the canteen and the participants were instructed to hand in all documents related to the study using the designated box.
Table 1 - Overview of Procedures During the Main Study
Week No Phase Procedure
1 Preparation Information sheet, Informed consent, Stress Questionnaire
2 – 6 Reminders, Presence at company
6 Individual reminder
7 Experiment Messages (2x)
8 Messages (2x), Questionnaire II, Stress measure
9 Messages (2x), Questionnaire III, Stress measure
10 After Care Debriefing, Raffle
One week before the study, all bus drivers who had indicated their willingness to take part in the experiment by handing in the informed consent and the Stress Questionnaire, received a note in their individual locker. The note reminded drivers that from that point on and for the upcoming three weeks (the duration of the experiment), they were to check their lockers daily, they were not to discuss the experiment with their colleagues, and they were to
hand in all documents related to the study.
During the three weeks of the experiment, each participant received a total of six envelopes, in which the goal-frame messages (progress or commitment, based on their randomly-assigned condition) could be found. The participants had been instructed
beforehand to place these messages in a holder near the steering wheel of the bus during their next shift and to read them multiple times throughout the day to keep the condition-dependent goal-framing salient. In addition, they were instructed to hand in the messages after each day.
I did this in order to induce a sense of responsibility. Again, reminders to take the messages with them on the shift were distributed in the locker room and the canteen. At the end of the second week, the Questionnaire II (including questions about the most prominent word from the messages and days the printout was checked), and the Stress Questionnaire were included in the envelope. At the end of the third week, the Questionnaire III (including questions of the manipulation check, behavioral intentions and demographics), and the Stress Questionnaire were included in the envelope. The envelopes were distributed based on a variable-interval schedule (identical for both conditions). The variable-interval schedule ensured that the participants received two envelopes per week with about two to three days in between. I did this in order to prevent the participants from predicting when the next envelope would arrive (Gray, 2007).
To guarantee the participants’ anonymity, I did not identify the participants by their names. Instead, I used pre-existing numbers, hence referred to as ‘FuelSave® numbers’, which were already used by the company to record the data about drivers’ fuel consumption.
To deliver the envelopes to the personal lockers of the participants, I chose an employee at the location who was trusted by the drivers but neither belonged to the bus drivers, in general, nor to the participant group specifically. I provided this employee with a list of the
participants’ FuelSave® numbers to enable this person to match these with the corresponding
locker numbers and distribute all documents to the participants. Since the employee did not have access to the data of the experiment, the anonymity of the participants was maintained.
One week after the study, the debriefing was distributed to the participants via their lockers. Also, a raffle for two gift certificates, each worth €50, was held in the canteen. The lottery was held as a reward for participating in the study. In order to gain access to it, each participant had to hand in all three questionnaires. The prizes were distributed in the same way as the other communication (through the company employee) in order to ensure the participants’ anonymity.
Measures
Stress questionnaire. This standardized measure of perceived occupational stress by Pearlin and Schooler (1978) had been previously used on bus drivers (Carrère et al., 1991). It was translated from English to Dutch for the use in the current study (Appendix B). It was preceded by a short statement asking the participants to think about their job, working conditions and colleagues, which was adapted from the original interview version by Pearlin and Schooler (1978). Then the following question was stated: “When you think of your day to day job, how … do you feel?”. The participants were instructed to fill in the blank (indicated by dots) mentally with seven different adjectives from a list, namely “bothered or upset”,
“unhappy”, “worried”, “frustrated”, “tense”, “contended” and “relaxed”. The participants were then asked to indicate, using a 7-point Likert scale, ranging from 1 (not at all) to 7 (very much), to what extent each of the adjectives applied to them.
Goal-frame messages. The twelve messages consisted of six pairs of messages and each pair included a message for the Commitment and Progress condition (Appendix C). The messages concerned what the drivers found important (as suggested by the Pilot Study), namely being a good bus driver, and the company goal, driving fuel-efficiently, and were formulated in past tense (adapted from Fishbach & Dhar, 2005). Between conditions the
messages differed regarding wording, such that they were formulated either in terms of progress and “progressing toward” (“By driving efficiently, I have made progress toward being a better bus driver.”) or commitment and “being committed to” (“By driving efficiently, I have shown my commitment to being a better bus driver.”), the focal goal of saving fuel.
Participants received one of the two types of messages based on the condition to which they were randomly assigned. The participants were instructed to put these messages on a
particular clip in the bus near their steering wheel, where they would see it during the shift.
The messages were written in the font Georgia (size 14) and printed on a yellow, roughly A5 size piece of paper in landscape layout.
Questionnaire II. It entailed three questions in total (Appendix B). One asked the participants in an open-ended question to write down the two most prominent words they remembered from the messages in order to verify that the participant had actually paid attention while reading them. Second, the participants were asked to indicate if they had looked at the printout of the FuelSave® results and thirdly, on which days they had done so.
This was intended as a subtle reminder to look at these printouts, which represented the initial action from which the subsequent action choice should be inferred (adapted from Fishbach &
Dhar, 2005, p. 374)
Questionnaire III. It included a total of eight questions (Appendix B), of which the first three were the identical to those in Questionnaire II. Then, two questions regarding the focal goal of fuel conservation were posed, of which one was framed in terms of commitment (“How committed are you to driving fuel-efficiently?”) and one in terms of progress (“Did you make progress as a fuel-efficient driver?”) (adapted Fishbach & Dhar, 2005). These were intended to serve as a manipulation check, such that a higher score on the question framed in terms of progress or commitment, should indicate a participant’s affiliation to the progress or commitment condition. In addition, two questions inquired about the participant’s intentions
to engage in a goal-incongruent activity of accelerating fast and a goal-congruent activity of recommending fuel-efficient driving to colleagues. The answers to the last four questions had to be indicated on a 7-point Likert scale, ranging from 1 (not at all) to 7 (very much). The final question asked the participant to indicate the highest obtained level of education.
Fuel usage measures. The company used a fuel management system called ‘Shell FuelSave®, which recorded fuel usage data per driver. The fuel consumption was measured over the course of four weeks, resulting in a pre-intervention score and three scores during the intervention for each dependent variable. The weekly individual driver data included Grade (%), Usage (km/l), Over-revving (m/km), and Coasting (m/km). Grade was a composite score that was comprised of equal parts of eight driving behavior indicators and Usage referred to how many kilometers were driven per liter of fuel. Over-revving referred to driving with an incorrectly selected gear and Coasting referred to when the drivers uses the momentum of the vehicle to carry on. Over-revving was the only variable in which a higher amount was considered less valuable. All of the above were used as separate measures for the dependent variable.
Control variables. The company provided me with the participant’s age, gender, tenure, and score on a driver risk assessment. This data was only identifiable through FuelSave® numbers, thus free of any information revealing the identity of the driver.
Analyses
All analyses were executed with IBM SPSS Statistics version 22 for Mac. First, a missing values analysis was conducted to determine if measures needed to be taken, such as mean substitution or exclusion of cases and/or variables. Then, a reliability analysis of the stress scales was performed to determine whether they could be combined into one overall mean stress score. Paired-samples t tests were performed on the means of the items as well as the overall mean of the Stress Questionnaires. Finally, regressions were conducted for the
Manipulation Checks and Intentions, and repeated measures analysis was applied to test the hypotheses.
Results Preliminary Analysis
The missing values analysis showed that 15.39% of the values in the complete dataset were incomplete. The Little’s t test, χ² (911) = 322.00, p = 1.00) was not significant, therefore the data could be assumed to be missing completely at random, which indicated that the non- response data could be viewed as ignorable and the estimates were not likely to be biased (Stuart, Azur, Frangakis, & Leaf, 2009). Variables with more than 30% missing values stemmed exclusively from the Stress Questionnaire and Questionnaire III of the third week.
Thus, I decided to delete all Stress variables (n = 7) of the third week because two stress measures would still remain. Also, mean substitution was performed on Tenure and
remaining stress variables, resulting in a reduced percentage of incomplete values of 10.95%.
Reliability Stress Scale
The correlation matrix (Table 2) showed significant high correlations mainly between constructs within each questionnaire, rather than constructs between questionnaires.
Significant medium correlations were found for Items 1 to 5 (bothered or upset, unhappy, worried, frustrated, tense) of Stress Questionnaire I (SQ I) with Items 6 (contended) and 7 (relaxed), the originally reverse scored items of Stress Questionnaire II (SQ II). Paired- samples t tests showed that the overall mean of SQ II was significantly higher than the overall mean of SQ I, t(23) = -2.09, p = .05. Hence, participants reported a higher level of perceived stress when they filled out SQ II. Also, the means of Items 1, 2 and 3 were significantly higher in SQ II than in SQ I (Table 3). This can be attributed to the relatively low standard deviations for Item 1 to 3, which are about half for SQ I as compared to SQ II.
The initial separate reliability analyses (with the recoded items 6 and 7) of the second Stress
ng Head: GOAL FRAME, FUEL CONSERVATION AND STRESS28 e 2 - Overview of Inter-Item Correlations of Stress Measures. Correlations (n = 24) MeanSD1 2 3 4 5 6 7 8 9 10111213thered or upset 1.13.45 happy 1.22.59 .89 **rried1.39.77 .74 **.78 **ustrated 1.571.21.75 **.87 **.89 **nse 1.701.12.77 **.70 **.70 **.70 **ntended2.391.66.40 .44 *.49 *.45 *.41 *laxed 2.351.71.34 .36 .36 .35 .52 **.89 **
thered or upset 2.111.53.23 .16 .06 .10 .27 -.05 .03 happy 1.951.49.07 .00 -.11 -.09 .14 -.11 -.02 .98 **rried2.211.40.02 -.06 -.10 -.11 .02 -.08 -.02 .91 **.92 **ustrated 2.051.49.44 *.38 .20 .24 .30 .08 .04 .90 **.87 **.79 **nse 2.261.53.46 *.33 .09 .13 .27 .02 .01 .83 **.81 **.81 **.93 **ntended2.651.68.40 .50 *.44 *.60 **.44 *.11 .03 .54 **.45 *.35 .69 **.53 **laxed 2.111.53.43 *.47 *.32 .56 **.43 *.15 .10 .53 **.45 *.38 .70 **.61 **.92 **
*p < .05. **p < .01. Item 6 and 7 for both scales were recoded to make all items unidirectional. e 3 - Paired-Samples T Tests for All Stress Scale Items. SQ I – SQ II (n = 24) 95% CI bleMeanSDt LowerUpper thered or upset -.97 1.49-3.20 **-1.66-.34 happy -.73 1.60-2.24 *-1.41-.05 rried-.82 1.66-2.42 *-1.52-.12 ustrated -.49 1.68-1.42-1.20.22 nse -.57 1.63-1.70-1.25.12 ntended-.26 2.22-.57 -1.20.68 laxed -.44 2.28-.95 -1.40.52 all mean -.61 1.43-2.09 *-1.21-.01 *p < .05. **p < .01.
measure produced a Cronbach’s alphas of α = .94 (all seven items), the first Stress measure a Cronbach’s alpha of α = .86 respectively. Hence, each scale could be summarized in a mean stress score. A reliability analysis of both mean stress scores produced a high reliability, α = .88, indicating that combining both mean scores would be appropriate. However, neither the correlations between the questionnaires (Table 2), nor the Intraclass Correlation Coefficient (ICC), ICC = .86, F(23, 312) = 7.263, p < .01 and 95% CI (.77, .93), supported this. The high ICC indicated that the variability within a stress measure was minimized, while the variability between the stress measures was maximized. Therefore, I decided to retain the scale with the higher α, namely the second Stress measure, hence referred to as Stress. In anticipation of the repeated measures analyses and regressions, as well as to avoid multicollinearity, I centered Stress (Ajken & West, 1991).
Manipulation Check
The regression of Goal-frame, Stress, and their interaction on the Manipulation Check Progress did not yield significant results (Table 4).
Table 4 - Regression Coefficients, Standard Errors, and R2 for Manipulation Checks.
Commitment (n = 14) Progress (n = 14)
B SE B B SE B
Constant 5.67 .18 5.36 .28
Goal-frame -.34† .18 -.27 .28
Stress -.49* .14 -.36 .23
Goal-frame x Stress -.45* .14 -.34 .23
R2 .59 .24
R2 change .40* .17
Note. *p = .05 **p = .01 (two-tailed).
In a regression of Goal-frame, Stress, and their interaction on the Manipulation Check Commitment, the model explained 59% of the variance and significant effects were found for Goal-frame, B = -.34, SE = .18, t(10) = -1.91, p = .09, and Stress, B = -.49, SE = .14, t(10) = - 3.42, p = .01, as well as the interaction, B = -.45, SE = .14, t(10) = -3.13, p = .01. These latter two effects were unexpected and suggested that stress had a slightly negative influence on the reported commitment of participants in the commitment condition and a slightly positive
influence on the reported commitment for participants in the progress condition.
Figure 2. Graphical Display of Interaction (Goal-frame x Stress) for Manipulation Check.
As Figure 2 illustrates, both commitment and progress focus resulted in a lower evaluation of commitment under high stress, than under low stress, though the difference was slightly greater for the participants with a commitment focus, hence the interaction effect.
Furthermore, the open-ended Question (Questionnaire II and III) about the most prominent words they remembered from the goal-frame messages, provided additional results. In 20.08% the participants remembered either the word “Commitment” or “Progress”, for 30.50% either no answer or “I don’t know” was given, and in the remaining 49.42%, the participants remembered the overall goal of “fuel conservation”.
Repeated Measures Analyses Fuel Usage
Initially, I planned to conduct a single repeated measures MANOVA to test my hypotheses. However, the small sample size forced me to conduct separate repeated measures ANOVA’s. The fuel consumption measures were entered as dependent variables (DV), Goal- frame as a between-subjects factor, Week as the within-subjects factor (repeated measure) and Stress as a covariate. The control variables (Age, Tenure, Score Risk Assessment) were not included due to the small sample size. A correlation matrix (Table 5) showed that the composite score Grade was expectedly correlated with its contributing variables Usage, positively, and Over-revving, negatively, in all weeks, but Week 2. In addition, Tenure and
5,5 5,55 5,6 5,65 5,7 5,75 5,8 5,85 5,9
Manipulation Check Commitment
Progress Commitment
Low Stress High Stress
ng Head: GOAL FRAME, FUEL CONSERVATION AND STRESS31
e 5 - Correlation Matrix of Variables
Correlations (n = 17) MeanSD1 2 3 4 5 6 7 8 9 101112131415161718192021e 1.536.04nder.94 .24 .40 ore Risk Assessment 2.9422.05-.20 .24 nure-.38 8.78.38 .05 .02 al-frame -.06 1.03.23 -.27 .38 .13 ress.00 1.50.07 .14 .17 -.38 -.22 0 (Pre-scores)ade 83.3611.06.00 -.06 -.21 -.31 .04 -.05 age 3.96.36 .07 .04 -.03 .02 .02 -.12 .76 **er-revving 29.1746.69-.16 .12 .38 .19 -.11 .19 -.78 **-.48 asting 82.6233.12-.35 .32 .30 -.50 *-.46 .45 .11 -.04 .35 1ade 82.277.23.16 -.29 -.36 .43 .21 .23 .39 -.04 -.28 .08 age 3.98.27 .31 .06 -.38 -.19 .01 .43 -.06 -.17 -.11 -.22 .54 *er-revving 31.7131.38.04 .12 -.17 .06 .00 -.12 .19 .21 -.21 -.19 -.29 .02 asting 83.2536.70-.26 .32 .29 -.44 -.36 .43 .10 -.11 .34 .96 **.16 -.19 -.11 2 ade 85.418.98-.24 -.13 .08 -.00 -.08 -.25 .44 .32 -.05 .34 .26 -.42 -.22 .33 age 4.12.39 -.17 -.09 .10 .35 -.14 -.34 .31 .58 *-.01 .06 -.24 -.59 **-.14 -.05 .69 **er-revving 25.6729.54-.07 -.18 -.38 -.09 -.07 .17 -.12 -.02 .02 -.22 -.14 .26 .36 -.28 -.66 **-.23 asting 89.1347.32-.28 .28 .30 -.41 -.38 .37 .37 .24 .08 .86 **.01 -.43 -.21 .83 **.44 .25 -.31 3ade 84.5711.10.17 .16 -.09 -.24 .04 .01 .88 **.73 **-.68 **.04 .25 .01 .26 .04 .30 .25 -.12 .31 age 4.09.40 -.18 -.07 -.13 -.24 -.13 -.11 .80 **.79 **-.53 * .12 .23 -.11 -.07 .06 .44 .47 -.14 .40 .79 **er-revving 28.4644.26-.27 .01 .00 .08 -.27 .01 -.46 -.48 .67 **.43 -.02 -.18 -.18 .44 .14 .12 .10 .17 -.53 *-.33 asting 85.7540.79-.17 .31 .19 -.42 -.43 .39 .36 .24 .11 .87 **.14 -.29 -.24 .86 **.41 .20 -.26 .96 **.33 .45 .27 *p = .05 **p = .01 (two-tailed). Age, Tenure, Score Risk Assessment and Stress were centered. Missing values were excluded listwise.
