Painting with the Same Brush? Surveying Unethical Behavior in the Workplace Using Self-Reports and Observer-Reports

Jones (1991) himself noted that this definition is relativistic. Tenbrunsel and Smith-Crowe (2008, p. 550) have discussed the fundamental problem in business ethics research of providing definitions of unethical decisions and behavior that “in and of themselves truly explicate the content of what is ethical”. For the purposes of the current paper, we use Jones’ definition, while we acknowledge the need for and importance of developing definitions of unethical behavior based on substantive philosophical grounds.

The response alternative “not applicable in my direct work environment” (“n.a.” in what follows) was treated as equivalent to “never” responses in the following analysis. This treatment is based on the analysis of the frequency with which respondents chose the response alternative “n.a.”. While respondents could have chosen the answer “n.a” instead of skipping the question, in which case selecting “n.a.” would be similar to not answering the question, there are a number of UB items in Kaptein’s (2008a) scale that are quite specific to interactions with customers, suppliers, or regulatory authorities. Engaging and/or observing such behavior is thus not possible in all job roles or environments, such that some respondents can be expected to select the “n.a.” responses for such behaviors. Consistent with this reasoning, examination of the frequency of “n.a.” responses across all 37 UB items showed that the number of “n.a.” responses per item varied widely across the items, from 1 in self-reports (1 for observer-reports) for breaching employee privacy to 178 (178) for doing business with third parties that may be involved in money laundering or are prohibited under international trade restrictions or embargoes. Comparison of the average number of “n.a.” responses in groups of items related to specific stakeholders as defined in Kaptein’s scale similarly revealed that for both self- and observer-reports, this number was highest for UB in relation to suppliers (158 for observer- and 160 for self-reports), followed by UB in relation to customers (134 and 135), and lowest for UB towards employees (8 and 8). This data supports the idea that respondents chose the “n.a.” response when, in their job role and/or their direct work environment, a behavior could not be observed or engaged in because their tasks and interactions do not provide the possibility of this behavior. Finally, treating “n.a.” responses as missing values, and calculating participation rates subsequently in % of non-missing responses would result in inflated participation rates. Similarly, inflated means would result if “n.a.” responses were excluded from analysis, rather than treated as “never” responses.

Tests using robust measures of location that rely on trimming of data or removal of outliers (Wilcox and Keselman 2003) do not seem appropriate for the data in this study, as the non-normality of the data does not result from a situation “where the underlying population is normally distributed, but a certain portion of observations deviates for any reason” (Wiedermann and Alexandrowicz 2011), but rather from a situation where the behavior examined does not follow a normal distribution in the population.

The ratio in means for each item was calculated as \(\frac{{{\text{mean}}\,({\text{observer}} {\text{-}} {\text{reports}})}}{{{\text{mean }}({\text{self}} {\text{-}} {\text{reports}})}}\) i.e., the ratio of the mean of observer-reports to the mean self-reports. Alternatively, a mean ratio of observer-reports to self-reports could be calculated as \({\text{mean}}\left( \frac{\text{observer-report}}{\text{self-report}}\right)\) i.e., the mean of the ratios of observer- to self-report across the respondents.

In the results section, we have reported the average ratio in means of observer- to self-reports calculated by taking the average of the ratios of the means across the 37 items. An alternative for determining the average ratio of observer-reports to self-reports across all items is to calculate the mean scale score of observer-reports across the 37 items for each respondent (the sample mean of this sale score is 1.18) and the mean scale score of self-reports across the 37 items for each respondent (the sample mean of this scale score is 1.06), and to take the ratio of these two sample means. This alternative method results in an average ratio of observer- to self-reports of 1.11, compared to 1.10 using the other method. The alternative method is more comparable to how the ratios can be calculated based on the scale score reported in the studies cited above.

Bailey et al. (2012) provide a useful distinction of different types of virtual work. In their terminology, we focus on virtual teams in our discussion.

The study by the Ethics Resource Center (2013) does not provide reasons as to why those spending 30 % or more of their working time connected to social networks (active social networkers) observe more UB. On the one hand, it is not entirely clear from that study how “observing” was defined in the questionnaire, i.e., whether indirect knowledge also is included in this concept. On the other hand, the study reports that active social networkers are more likely to be managers, are younger, and more likely to be male than other employees. In line with the argument presented in the following paragraph of the main text, because more of the active social networkers are in management positions that involve monitoring and checking of work by others, active social networkers may on average observe more UB.