Sample & design. Study 1 was a two-wave two-source survey conducted using an online panel of full-time workers from various workplaces. Measures assessing all constructs relevant to hypotheses were administered. Data was collected twice from focal participants, with a four-week lag, and also from a supervisor at the time of the second data collection. Supervisor emails were collected from focal participants using a snowball style sampling procedure (for similar sampling procedures, see e.g., Grant & Mayer, 2009; Moore, Detert, Treviño, Baker, & Mayer, 2012).
To ensure that participants did not complete all three surveys themselves,
timestamps and IP addresses were checked. Additionally, participants were informed that these checks would be made, and that payment would not be issued if supervisor surveys
are filled out by the focal employee. As a further measure, participants were asked to provide work-hosted emails for their supervisor if possible.
For this study, a statistical power level of .80 for mediation hypotheses required a sample of approximately 160 observations (assuming small to moderate indirect effect sizes; Fritz & MacKinnon, 2007). For structural equation modeling (SEM), achieving a statistical power level of .80 would require a sample of at least 245 (Preacher & Coffman, 2006). The multi-wave design and sampling procedure utilized make the ultimate sample size somewhat difficult to predict based on the initial number of participants recruited.
Thus, I estimated an initial sample size which could result in an workable number of final observations if response rates were not ideal – N = 350 (60% Time 2 response rate would result in N = 210, potentially enough for SEM; 30% Time 2 response rate would result in N = 105, which may allow power to find hypothesized effects, but would cast doubt about the generalizability of null effects).
Ultimately, 405 usable Time 1 responses and 146 Time 2 responses (36%
response rate). Twenty-seven were dropped from analyses due to failing attention checks, straightlining through responses, and completing the survey at excessive speeds, resulting in a sample size of N = 119 for focal participants. A supervisor response rate of 48% led to a sample of N = 58 with matched supervisor responses.
Procedure. The focal sample completed a survey measuring independent variables (including mediators) at Time 1. Three weeks later, at Time 2, participants completed another survey measuring moderator variables (ethical leadership & bottom-line mentality) and dependent variables. At Time 1, participants were asked to volunteer
a supervisor and/or a coworker’s email address – at Time 2, the supervisors were
contacted and asked to complete a survey measuring their moral identity and bottom-line mentality, as well as measures of the dependent variables for the focal employee.
Participants were paid $5 for completing Survey 1, another $5 for completing Survey 2, and another $5 if their supervisor completed the survey ($15 total).
Supervisors/coworkers were compensated $5 for completing their survey.
Time 1 measures (all measures included in Section VIII. Appendix)
Self-construal. Gudykunst and Lee’s (2003) 12-item self-construal scale was
used. This is a short form of the 28-item measure developed by Gudykunst, Matsumoto, Ting‐Toomey, Nishida, Kim, and Heyman (1996). The measure developed by Singelis (1994) is more widely used, but has been shown to display inadequate levels of internal reliability (Cross et al., 2011).
Anticipated guilt & gratitude. Anticipated guilt and gratitude were measured for
the hypothetical situations of a) “if you were to break social and moral norms to benefit yourself at work,” b) “if you were to break social and moral norms to benefit yourself and others at work,” c) “if you were to help another person or group of people at work”; while anticipated guilt only was measured for d) “if you were to miss an opportunity to help another person or group of people at work.” Four items for anticipated guilt and three items for anticipated gratitude (adapted from Grant & Wrzesniewski, 2010) were used for each respective hypothetical, for a total of 25 items.
Perspective taking. The four-item scale used by Grant and Berry (2011) was utilized.
Moral disengagement. The 16-item measure developed and validated by Moore
and colleagues (2012) was used.
Time 1 controls
Moral identity. Moral identity has been shown to relate negatively to unethical
behavior. The five-item measure of the internalization dimension of moral identity developed by Aquino and Reed (2002) was used.
Self-concern & other-orientation. Self-concern and other-orientation are closely
related to self-construal, and sometimes used in lieu of self-construal in the applied psychology literature. The six-item measure (three for self-concern & three for other-orientation) utilized by De Dreu and Nauta (2009) was used.
Prosocial motivation. Prosocial motivation relates positively to OCB (Rioux &
Penner, 2001), and has been used as an operationalization of other-orientation (Grant &
Wrzesniewski, 2010). The four-item measure used by Grant (2008) was used.
Positive reciprocity. Positive reciprocity has been shown to relate positively to
unethical pro-organizational behavior (Umphress, Bingham, & Mitchell, 2010). A five-item version of the 10-five-item scale developed by Eisenberger, Lynch, Aselage, & Rohdieck (2004) was used. The scale was adapted by dropping the five items from the original scale which displayed factor loadings of less than .50 in the Eisenberger et al. (2004) study.
Big five personality traits. Agreeableness, extroversion, and conscientiousness were measured using items from Donnellan, Oswald, Baird, and Lucas’ (2006) 20-item Mini-IPIP scale. This measure will consist of 12 items (four for each construct).
Agreeableness and extroversion may arguably be related to self-construal, while
conscientiousness may relate both to OCB (Chiaburu, Oh, Berry, & Gardner, 2011) and pro-self unethical behavior (e.g. Greenbaum et al., 2012).
Trait affectivity. An 8-item short form of the Positive and Negative Affect
Schedule (PANAS; Watson, Clark, & Tellegen, 1988; Zellars, Tepper, & Duffy, 2002) was utilized. Participants will respond regarding the extent to which they generally feel 8 discrete emotions.
Self-regulatory depletion. Self-regulatory depletion has been shown to relate
positively to unethical behavior (Gino, Schweitzer, Mead, & Ariely, 2011). The 10-item state self-control scale developed by Ciarocco, Twenge, Muraven, and Tice (2007) was used.
Envy. Envy is likely to be relevant specifically to social undermining (Duffy,
Scott, Shaw, Tepper, & Aquino, 2012). The five-item scale from Vecchio (1995; 1999) was utilized.
Sense of power. Participants will respond to the eight-item measure developed by
Anderson and Galinsky (2006), with questions anchored on a work context. Power at work will reflect participants’ level of autonomy, while also being relevant to the effects of (un)ethical leadership (Pitesa & Thau, 2013).
Private self-focus. Participants will respond to the three-item private self-focus measure developed by Govern and Marsch (2001). Private self-focus may be relevant to the effects of individual differences on the outcomes of concerned (Pitesa & Thau, 2013).
Demographics. I plan to measure participants’ gender, age, race, education level,
nationality, and tenure.