blowing the whistle on data fudging: a controlled field experiment1

Blowing the Whistle on Data Fudging: A Controlled Field Experiment’

MARCIA P. MICEL? JANELLE B. DOZIER Management and Human Resources

Ohw State Unioersity Deprtment of Management

Uniuersity of Houston-Uniwrsity Park

JANET I? NEAR Department of Management

Indinno Uniuersity

This experiment examined determinants of “whistle blowing,” the disclosure of organizational wrongdoing. Previous research, limited largely to analyses of cross-sectional, self-reported data, has not established cameffect relations, and dispositional variables generally have not been examined. In this study, 295 students witnessed apparent wrongdoing by a research assistant and were given an opportunity to report it to a university authority. Results indicated that men were more likely than women to report the wrongdoing. Whistle blowing also was a function of lower, rather than higher, levels of moral judgment development. More whistle blowing occurred when there were more, rather than fewer, observers of wrongdoing. Implications for research and practice are descried.

Whistle blowing, the ”disclosure by organization members of illegal, immoral or illegitimate practices under the control of their employers, to persons or organizations who may be able to effect action” (Near & Miceli, 1985, p. 6), has attracted much attention. Whistle blowers may threaten the authority structure and functioning of organizations (Weinstein, 19791, and they may disrupt member relations in groups (Jensen, 1987). At the same time, whistle blowers may also enable organization leaders to correct harmful practices (Graham, 1983).

Interest in whistle blowing has been heightened by widely publicized individual cases and other factors (Ewing, 1983). However, very little scholarly work on whistle blowing exists (Graham, 1983). Some survey research concerning the predictors of whistle blowing has been conducted (e.g., Graham, 1983; Miceli & Near, 19851, and it has

‘Portions of this paper were presented at a meeting of the Academy of Management. The authors thank the College of Business and the Ohio State University Seed Grant Programs for providing funding for this research, and Melissa Baucus and Linda Ferguson for their comments on earlier versions of this manuscript. They also thank Todd Gadowski, David Homewood, Bill Nemeth, Marjorie Stassen, Sherry Sullivan, Mike Thomas, Doug Wassell, and Rich Wiggins for their research assistance.

’Requests for reprints should be sent to Maraa Miceli, Management and Human Resources, Ohio State University, 356 Hagerty Hall, 1775 College Road, Columbus, OH 43210-1399.

271

Journal of Applied Social Psychology, 1991,21,4, pp. 271-295. Copyright Q 1991 by V.H. Winston 8 Son, Inc. All rights reserved.

272 MICELI, DOZIER, AND NEAR

shown that characteristics of the wrongdoing observed and jobs held by whistle blowers differ from those of other organizational members. However, such studies are limited in two major ways. First, they generally have not considered the impact of dispositional variables on whistle blowing decisions, yet organization members’ characteristics may affect their reactions to organizational wrongdoing (Farrell & Petersen, 1982; Kolarska & Aldrich, 1980; Perrucci, Anderson, Schendel, & Trachtman, 1980). Secondly, the findings do not show whether cause- effect relationships exist. Respondents may rationalize their actions or failures to act because attitudes may sometimes follow from, rather than direct, behavior (e.g., Salancik & Pfeffer, 1978). This problem seems especially critical where the observation of wrongdoing is concerned, because one may feel compelled to describe one’s behavior in a socially desirable or morally correct-but inaccurate-manner.

Controlled field experimentation is useful in such cases, because dispositional characteristics can be measured, and situational conditions can be manipulated independently. Further, the researcher need not rely upon self-reports as measures of both the independent and dependent variables. The present study attempts to address both of these limitations in investigating whether certain individual characteristics and situational conditions influence whistle blowing.

Factors Influencing Whistle Blowing

There is no widely acknowledged, comprehensive theory of whistle blowing; however, various authors have suggested that existing theories in the psychological literature are of value in conceptualizing the factors that may lead to whistle blowing (Near & Miceli, 1987). In the present study we rely primarily on social learning theory (e.g., Bandura, 1977), and on research concerning prosocial behavior (e.g., Latan6 & Darley, 1970) in developing hypotheses. Each of these approaches is described below.

Threatened Retaliation

Observers of wrongdoing may be concerned with potential retaliation that may result from reporting wrongdoing (Farrell & Petersen, 1982). Bandura’s (1977) research provides one theoretical framework for understanding this concern. It is possible to use other frameworks, such as reinforcement theory (e.g., Skinner, 1953), but only if reference is made to stimulus control; we refer here to expected consequences rather

WHISTLE-BLOWING 273

than to the effects of experienced retaliation. We rely on Bandura simply because more readers may be familiar with his propositions than with stimulus control propositions, but the predicted relationships among variables would be the same.

Bandura proposed that if individuals' desire and ability to manipu- late the environment, or "self-efficacy," is essential to their well-being, then individuals will be motivated to perform acts that demonstrate their self-efficacy. "A risk of feared consequences" and "situational uncertainty'' can influence efficacy expectations (Bandura, 1977, p. 83). In the case of whistle blowing, where wrongdoers suggest that they would react unfavorably to whistle blowing, observers of wrongdoing may be discouraged from acting because such consequences would be punish- ing ("a risk of feared consequences") or because it is unclear to them what the appropriate response should be ("situational uncertainty"). Consistent with this idea, previous research (cited by Trevino, 1986) shows that subjects were more likely to behave unethically if it appeared that ethical behavior would be punished. Thus, even if blowing the whistle is viewed as ethical behavior, retaliatory threats should reduce whistle blowing.

Although the reasoning is straightforward, not one previous study of whistle blowing supports it. The comprehensiveness or severity of perceived actual or threatened retaliation was unrelated to intended future whistle blowing at the individual level (e.g., Near & Jensen, 1983; Near & Miceli, 1986). In another study (Miceli & Near, 1985), a retaliatory climate was associated with a greater propensity to blow the whistle to parties external to the organization. However, in these survey-based studies, uncontrolled variables may have operated to obscure the relationships, and perceptions may have been distorted. For example, some whistle blowers unknowingly may have exaggerated perceived threats or harassment. It is also possible that, in all the cited studies, restriction of range may have attenuated the observed relationship. Self-selection by respondents (following a mailed request for participants) may have excluded many respondents who experienced no retaliation, or those who did not want to relive a traumatic experience. In some studies, sur- veys never reached those who left their organizations, perhaps after being fired or experiencing other extreme forms of retaliation. In the present study, none of these problems could occur. Thus, we predict:

H1: Subjects will be more likely to blow the whistle where no retaliation is threatened than where retaliation is threatened.


Locus of Control

Previous authors have provided evidence that whistle blowing generally can be considered to be a type of prosocial behavior (Brief & Motowidlo, 1986, Dozier & Miceli, 1985; Graham, 1983; Miceli & Near, 1988), in which the whistle blower attempts to benefit other parties believed to be harmed by the wrongdoing. This suggests that predictors of other types of prosocial behavior may also predict whistle blowing.

Research shows that individuals with an internal locus of control (Rotter, 1966) are more inclined than those with an external locus of control to engage in prosocial behavior, such as helping and crime reporting (see Spector, 1982, for a review). Internals may believe that they control the events that affect them or have a stronger “action orientation” (Spec- tor, 1982), or they may feel more personally competent in general than do externals. In previous studies, felt personal competence to deal with an emergency was prominent in distinguishing intervenors (e.g., Shot- land & Heinold, 1985) and other evidence (cited by Spector, 1982) suggests that internals outperform externals in organizations, which may lead to felt competence. Similarly, research on whistle blowing suggests that whistle blowers tend to be high performers (Miceli & Near, 1988); competent individuals may exert greater minority influence over other organization members, enabling them to avoid retaliation or to bring about cessation of wrongdoing (Near & Miceli, 1987).

These findings imply that the more “internal” the locus of control, the more likely that organization members will blow the whistle, because (a) whistle blowing is a form of prosocial behavior and previous research shows that locus of control is related to prosocial behavior; (b) felt competence, which previous research shows may be related to locus of control, may also be related to one’s expected efficacy in bringing about cessation of wrongdoing or in avoiding retaliation.

However, other research suggests that locus of control will interact with situational conditions to influence whistle blowing. Therefore, it may be that the effects of locus of control will be more pronounced under certain situational conditions than in others. Specifically, research suggests that competent-feeling individuals may underestimate the risk involved in intervening in wrongdoing, believing in their own capacity to meet the threat, while persons with feelings of lower competence may inflate the risk involved (Shotland & Goodstein, 1984). Consequently, “high competents” may be less sensitive to threatened retaliation. Similarly, internals are less conforming and compliant than externals

WHISTLE-BLOWING 275

and they may be less influenced than externals by influence attempts (such as threatened retaliation), especially where the pressure seems strong (Spector, 1982). Also, internals may experience greater psychological reactance (Brehm, 1972) in response to influence attempts (Biondo & MacDonald, 1971). Internals will be more likely to take personal responsibility for the consequences of ethical behavior and to be less affected by external forces (e.g., attempts to dissuade the observer from blowing the whistle) than by personal ethical standards (Trevino, 1986). Since no known research demonstrates that internals have different ethical standards than do externals-nor would any theoretical perspective suggest this-we predict:

H 2 Internals will be more likely than externals to blow the whistle, particularly in situations where conditions are not supportive of whistle blowing (i.e., retaliation is threatened).

Level $Moral Judgment Deuefupmenf

Another dispositional characteristic previously found to be associated with prosocial behavior is the level of moral judgment development (Kohlberg, 1969; Rushton, 1984). Individuals who reason at “higher,” or more complex, moral levels are more likely to engage in prosocial behavior, particularly where situational conditions support prosocial intervention (Staub, 1974). Moral judgment development may affect one’s assessment of personal responsibility for intervening, a critical step in the decision process used by bystanders (Latan6 & Darley, 1970) and, presumably, by whistle blowers (Dozier & Miceli, 1985).

This reasoning suggests that generally, whistle blowers will have higher levels of moral judgment than will observers of wrongdoing who do not blow the whistle. Consistent with this proposition, Brabeck (1984) demonstrated that ”highly moral” subjects were more likely to point out a possibly unintentional error in a manuscript in response to a specific request by the manuscript writer. However, this study is not definitive. Subjects may have viewed error identification to be purely ”helpful,” because the public, the organization, and the ”error-maker”-who may not have been viewed as a wrongdoer-would be helped. In most instances of successful whistle blowing, however, the public would be helped, but the wrongdoer (and perhaps, the organization) would likely be hurt. Situational conditions, which in Brabeck’s study seemed supportive of whistle blowing but which were not manipulated, would be particularly


important in settings that evoke moral conflict-or in those which call for behavior that is not entirely altruistic (Brief & Motowidlo, 1986). If the situation did not suggest that blowing the whistle would be viewed as clearly helpful, moral judgment development may be less closely related to whistle blowing, because the hypothesized underlying psychological process is that helping and not hurting is caused by higher moral development. Therefore, we predict that:

H3: Subjects whose moral judgment is more highly developed will be more likely to blow the whistle than will subjects whose moral judgment is less highly developed, particularly where conditions support whistle blowing (ie., no threatened retaliation).

Gender

Most of the cases of whistle blowing reported in the media have involved male whistle blowers; this implies that women are less likely to blow the whistle than are men, a finding reported by Miceli and Near (1988). Similarly, Gelfand, Hartmann, Walder, and Page (1973) found that women were significantly less likely to report shoplifters than were men.

If it is true that men and women may differ in their propensity to blow the whistle, it is not clear why. Men may be more likcly to be established members of professional groups where whistle blowing is en- couraged as part of a code of ethics (e.g., engineers, physicians). Thus, they may feel more morally responsible for correcting perceived wrongdoing. But in Brabeck's (1984) study, all whistle blowers were female (though the effects of gender were not statistically significant) and they had higher levels of moral judgment development than did non-whistle blowers (which was significant). Thus, moral development rather than gender may have accounted for those findings. A second hypothesis linking gender and whistle blowing stems from the observation that men in general may have higher levels of self-esteem or initiative (Eagly, 1983). These characteristics may cause prosocial behavior (Rushton, 1984) and whistle blowing (Kolarska & Aldrich, 1980). A third hypothesis is that men may tend to occupy positions where more serious wrongdoing is more likely to be observed, because many high status occupations are sex segregated (England, 1979).

Each of these hypotheses assigns causation to factors that can be

WHISTLE-BLOWING 277

controlled experimentally or statistically. The first two focus on moral reasoning and initiative. In the present study, moral judgment development and locus of control, which has been used to represent ”initiative” (Near & Miceli, 1985), are measured and their independent effects are assessed. The third process concerns organizational position and status, which are controlled in this study. All participants have the same opportunity to observe and report wrongdoing. Therefore, we predict that:

H4: Women will be less likely to blow the whistle than will men, but this effect will disappear when the effects of the dispositional variables are controlled.

Related Issues

Finally, research on the diffusion of responsibility effect (Latane) & Darley, 1970) suggests that the presence of others affects prosocial behavior; the ”diffusion of responsibility” hypothesis holds that observers feel less responsibility for action when other observers are present. Be- cause this effect was not of central interest to us in this study, the manipulation of the group size was not undertaken. However, we are able to examine the general research question that larger group sizes should produce fewer incidents of whistle blowing than do smaller group sizes, by treating varying group sizes as a covariate.

A second variable examined here is conformity to the request for wrongdoing. Many whistle blowers are placed in situations where they are either asked to participate in wrongdoing or in the cover-up of that wrongdoing; such was the case in this study as well. A second dependent variable of potential interest then was whether subjects would ac- quiesce to a call for wrongdoing, as described below. While conformity is not a direct interest of this research (i.e., we developed no specific hypotheses on this variable), it is clear that conformity might be nega- tively related to whistle blowing, and we conducted exploratory analyses on this variable for that reason.

Method

Subjects

Subjects were 155 male and 140 female students enrolled in a junior- level course in organizational behavior and management at a large midwestern university. They received course credit for participating.

278 MICELI, DOZIER, A N D NEAR

Training and Assignment of Research Confederates

To accommodate students’ schedules, many sessions-each of which required two research confederates, the “Research Assistant” (RA) and the ”(Name of Faculty Department) Committee Representative” ( C R b were administered. Five paid confederates were hired following their response to an advertisement in the student newspaper. To reduce potential extraneous variance, (a) all confederates were white male undergraduates whose age and appearance were typical of male subjects’ and (b) two training and three refresher sessions, in which the RA and CR roles were rehearsed, were held. Details of these procedures are omitted because of space limitations, but they are available from the authors. Various pairings of confederates occurred across the experimental conditions.

Design Considera f ions

The research design, procedure, and measures selected were strongly influenced by concerns about minimizing experimental demand and potential for suspicion. The procedure was extensively pretested, modi- fied, and again pretested; results are omitted because of space limitations but they are available from the authors. In particular, decisions concerned the trade-offs of asking numerous exploratory questions of subjects in order to gain insight into the psychological processes that may have occurred during the study versus the threats to the validity of the design that might be posed by the intrusions. We believed that interrupting the study to ask numerous questions about subject reactions would jeopardize the internal validity of the design, because such inter- ruptions would artificially call attention to activities that might otherwise be reacted to differently. Certain questions might invoke norms (e.g., that whistle blowing was morally correct and expected) that would not be invoked in an actual organizational setting or by a less intrusive design. Or they might cause a large number of subjects to become suspi- cious (e.g., if they were asked immediately after the manipulated wrongdoing but before an opportunity to report it, ”what are you now thinking about what just happened?”).

Rather than interrupting subjects to ask immediately about their reactions, we could have asked questions at the end of the study about subjects’ earlier thoughts, i.e., retrospective accounts. But this procedure would also yield ambiguous results, because (a> subjects’ reports of

WHISTLE-BLOWING 279

reactions to early events might be distorted by intervening events, and (b) in any study involving moral actions, subjects might resort to self- protective, socially acceptable justification of their actions. As an cx- ample of the basis for this concern, the findings of a field experiment involving reactions to confederate shoplifters, mentioned earlier, are illuminating (Gelfand et al., 1973). Gelfand et al. asked shoppers some questions after staging a shoplifting incident, secretly recording whether the shopper witnessed the incident, and allowing time for them to report it to cashiers. They reported that “of the customers who failed to report, 57% justified their behavior with one of the following reasons: The shoplifter left too quickly, no clerk was readily available, or the store was too crowded. Surprisingly, 19% of the nonreportcrs indicated during the interview that they would always report” (p. 281). These findings raise questions about whether the ”reasons” given in response to post hoc questions are truly indicative of individuals’ perceptions and motivations in such situations.

We therefore believed that it was critical to ask as few questions as possible of the experimental subjects and to focus our attention on ob- servable behavior, that is, whether subjects blew the whistle when given the opportunity to do so. However, because so little is known about whistle blowing and subjects’ perceptions were potentially valuable, we posed hypothetical questions to a holdout group of 207 other subjects drawn from the same population. We described the events in the experiment and asked the holdout subjects their reactions. Relevant responses are included below. Clearly, this solution is not perfect in that reactions of the holdout group may not be the same as those of the other subjects and may still be influenced by self-enhancement effects, but it is the most conservative design-in terms of undermining internal validity- of the options available to us.

Procedure

Subjects completed the dispositional instruments during regular class time. Several weeks later, they were scheduled to participate in the remaining part of the “managerial decision-making study.” Before each experimental session, the room was prepared for the bogus “job applicant selection task.” A folder containing ”applications for em- ployment,” computer-scored rating sheets, a rating form and “job description,” and a ”note pad,” was placed at each desk, which was physically separated from other desks. After all subjects were seated, the


RAdistributed written instructions and asked subjects to follow along as he read aloud:

Today you will be assuming the role of a manager who must determine whether some applicants are qualified for a particular job. Then, the student representative from the (Department) Commit- tee will ask you to complete their questionnaire. This procedure is followed in (Department) experiments because the experiment requirement is a new one.

The RA then explained how to perform the selection task, and then asked if there were any questions. After all questions were answered, he put the instructions aside and made the following request:

Well, that’s all of the ”official” instructions. I have another request to make, We have been doing this study for some time, and unfortunately, the results are not coming out the way we hypothesized. It’s going to be difficult to publish the results, the way things are going. You are in Condition 4, and we‘ve hypothesized that you will say that Applicant 1 is less qualified than Applicants 2 and 3. So, could you indicate on your computer sheets that 2 and 3 are more qualified? This will mean that we will be more likely to publish the study.

When this request to “fudge” the data was posed hypothetically to the holdout subjects from the same population, 76.3% of them said they considered it to be ”very wrong” and another 13% described the request as ”wrong.” Thus, the ”fudging” request served as the ”organizational wrongdoing.” I t should also be emphasized that the subjects were actual members of the organization in which wrongdoing appeared to be oc- curring, and thus this study represents a controlled field (as opposed to a laboratory) experiment, as organizational members were put in a position to react to apparent wrongdoing in an organization to which all had belonged for at least three years. Evidence from the holdout subjects confirmed that the consequences of the wrongdoing were harmful and were of concern to them. More than 97% identified at least one party that was harmed by the wrongdoing (most commonly ”the in- tegrity of the University”). When asked about the consequences of not reporting the wrongdoing (e.g., ”biased findings would result”), 82.1 % of the holdout subjects identified at least one consequence that was

WHISTLE-BLOWING 281

"somewhat important" to them; 65.7% said that at least one consequence was "impor tan t."

After making the request, the RA then administered one of the two levels of Threatened Retaliation (described later). As the subjects performed the task, the RA did his own schoolwork in the back of the experienced room.

After all materials were completed (about 30 minutes later) the RA collected them, and left the room, saying, "in a minute, the (Department) Committee Representative will be coming in to give you their questionnaire. Thanks a lot." After he left, a second confederate entered and in- troduced himself as the "student representative" on the committee, which was described as having the responsibility for getting student input on the experiment participation requirement. He asked subjects to complete the questionnaire anonymously, and to leave their materials on the desks. As described later, this questionnaire provided an opportunity for each subject to blow the whistle. After completion, he went back to the hall and returned with the RA. Both men then explained that they were confederates and debriefed the subjects concerning the true purpose of the study. They answered all questions raised by the subjects. Then, subjects were asked to put their names or social security numbers on the questionnaires so that the dispositional measures, task materials, and questionnaires could be matched. All but one subject (excluded from the analysis) agreed to this. Subjects were asked to refrain from discussing the study with other students. Subjects were asked to indicate whether they had heard anything about the study from other subjects who had participated previously. Data from the two subjects who indicated that they had heard about the study were excluded from the analysis.

Manipulations and Measures

Threatened retaliation. In the "threat" condition, the RA made an additional statement following his committing wrongdoing. He said, "Don't tell anyone about this or we will have to do the experiment over and have to get students to participate all over again." Presumably, students would not want to participate a second time as the statement implied. In the "no threat" condition, the RA made no additional statement. Ob- viously, the severity of the Threatened Retaliation manipulation was limited by ethnical concerns and by the University Human Subjects Committee guidelines, which protect subjects from experimental stress.

282 MICELI. DOZIER, AND NEAR

Thus, it was impossible to use more powerful, direct, or potentially realistic threats (e.g., "or I will see to it that your grade is lowered," "or I will physically harm you," "or you'll be sorry," etc.).

Locus of control (LOC). Subjects' scores on the 29-item forced-choice format of Rotter's (1966) Locus of Control scale were calculated (six filler items were deleted in scoring). This format is relatively free from social desirability bias and it measures a unitary construct (Ashkanasy, 1985). High scores indicated an external locus of control. The median was com- puted and subjects were grouped as above the median or below the median. Data from subjects with missing item scores were excluded from the analysis.

Level of morn1 judgment development ("Percent P"). Subjects completed the short version of Rest's (1979) Defining Issues Test (DIT). The DIT, which measures six levels of moral reasoning identified by Kohlberg (1969), has been widely used and has acceptable psychometric proper- ties (Rest, 1979). The short DIT presents subjects with three moral dilemmas and asks them to rate and rank the importance of 12 statements about each dilemma in deciding how the actor should resolve it. Several scores are generated by the DIT. The Percept P score reflects subjects' use of upper levels of moral judgment development, i.e., prin- cipal moral considerations (Stages 5 and 61, relative to lower levels (Stages 2, 3, and 4). High scores indicate high levels of moral reasoning (Rest, 1979). The M score identifies subjects who select lofty but nonsen- sical statements as being important in making a decision. Subjects who answer items inconsistently or indiscriminantly are also identified. As Rest (1979) recommended, only subjects with high "consistency" scores or low M scores were included. The median Percent P score was com- puted and subjects' scores were dichotomized as above the median or below the median.

Gender. Subjects reported their gender (coded "1" if male and "2" if female) on the bogus application review task materials.

Dqendenf Variables

Whistle blowing. Following several bogus items on the Committee Questionnaire, another item asked, "during the study, were you asked to do anything you considered objectionable?" This question was carefully worded broadly (as opposed to a more direct question such as "did the RA ask you to fudge the data?") so that (a) only those subjects who found this request objectionable-a defining characteristic of whistle

WHISTLE-BLOWING 283

blowing-would report it; and (b) suspicion would not be aroused (a real questionnaire independent of any given study would not have anticipated a fudging request). Subjects could check "yes," "no," or "can't remember." Subjects who checked "yes" were asked to describe the objectionable request. Two trained raters, who were unaware of the subject's dispositional scores or experimental condition, read each subject's answer and assigned a "2" to "whistle blowers"-subjects who answered "yes" and described the RA's "fudging" request. All other answers were scored as "l.!' (e.g., "yes-I thought the room was too hot"; "no"). The interrater reliability coefficient was .98. Reexamination of the two responses on which there was disagreement revealed a clear recording error on the part of one of the raters-as opposed to ambiguity of subjects' responses or differences in rater judgement-so the appropriate corrections were made and the two cases were retained.

Conformity. In addition, we collected data to determine the extent to which (a) subjects conformed to the request to "fudge" the data and (b) whistle blowing was related to conformity to this request. Thus, conformity served as a second dependent variable, though it was not of central focus as was whistle blowing. Conformity scores were calculated as follows. Subjects indicated on 10-point scales (from 0 to 9) the extent to which each of three applicants were qualified for a "job." It will be recalled that one applicant was actually rated by subjects not in this experiment to be substantially more highly qualified than the other two applicants, but subjects were asked to rate the third applicant lower than the other two. We subtracted the rating of Applicant 1, the more qualified applicant, from the mean rating for Applicants 2 and 3. Thus, potential scores ranged from -9 to 9, with an observed range of -8 to 5; lower (including negative) scores represented less conformity, while higher scores represented greater conformity to the "fudging" request.

Results

Because subjects did not fill all slots in each session, group sizes ranged from 2 to 23, independent of the experimental condition. There- fore, group size was treated as a covariate. Two analyses were performed to test the hypotheses. First, because discriminant analysis is particularly appropriate with dichotomous dependent variables, a hierarchical discriminant analysis was performed, with whistle blowing serving as the group (dependent) variable. The covariate was entered first in the analysis, followed by (a) the main effects; (b) the two-way


Table 1

Percenfages of Whistle Blowers as a Function of the Independent Variables"

Low threatened High threatened Dispositional retaliation retaliation

variables Men Women Men Women

Internal locus of control Low moral judgment development 45% 31% 42% 31%

External locus of control

High moral judgment development 35% 24% 47% 6%

aN = 295; ns range from 13 to 20.

High moral judgment development 25% 30% 20% 20%

Low moral j u d p e n t development 40% 33% 47% 20%

interactions; (c) the three-way interactions; and (d) the four-way interaction. In this analysis, the dispositional variables were not dichotomized, because discriminant analysis uses continuous or dichotomous variables to predict a categorical dependent variable, in this case, whistle blowing.

Second, because it allows for planned contrasts, analysis of co- variance was used (a) to test Hypothesis 1, which predicted a main effect for threatened retaliation and (b) to examine the general research question involving group size. ANCOVA with planned contrasts were used to test H2 and H3, which posited interactions between threatened retaliation and each of the two dispositional variables, respectively. A saturated model was tested. A separate ANCOVA partialling the effects of the dispositional variables in order to measure gender effects in the absence of the dispositional effects was performed. For this analysis, the median split of the dispositional scores was not used; instead, the interval data were retained. Similar analyses using conformity rather than whistle blowing as the variable of interest were employed. Discriminant analysis was not used to examine the effects of the variables on conformity because conformity was an interval variable.

Unadjusted cell means and standard deviations are reported in Table 1. The mean locus of control score was 10.47 (standard deviation = 4.09); the mean Percent P score was 36.0 (standard deviation = 16.86).

WHISTLE-BLOWING 285

Prelimina y Analyses

Approximately 38.1% of the mean and 24.3% of the women (31.5% of the total) blew the whistle. A comparison of these figures with those of other studies provides evidence of the likely meaningfulness of the study results. These figures are similar to percentages of whistle blowers using internal channels in prior field studies (e.g., slightly less than a third of observers of wrongdoing reported it to internal parties, Miceli & Near, 1985) and Brabeck's (1984) lab study (32% of those exposed to wrongdoing reported it). The findings are also consistent with previous field research showing that women are slightly less likely to blow the whistle (Miceli & Near, 1988). The figures are strikingly similar to those reported in the field study of shoplifters (Gelfand et al., 1973); 38% of men reported the shoplifter and 19% of the women did.

As expected, the ANCOVA revealed that group size was an important determinant of whistle blowing, F (1,277) = 20.03, p < 0.001. Unex- pectedly, however, subjects were more likely to blow the whistle when there were more, rather than fezuer, other observers present.

The fudging request asked subjects to rate one applicant below the other two, who were in fact clearly less qualified than the first. The vast majority of the subjects (75.3%) did not conform to the request; they rated the highly qualified applicant equal to or higher than the mean scores given to the other applicants (mean conformity score = -1.97, sd = 2.71). There was no correlation between conformity and whistle blowing, r(279) = 0.052; p > .192). The results of the analysis of variance with conformity as the dependent variable are reported following the results involving the tests of the hypotheses.

Tests of the Hypotheses

The results of the hierarchical discriminant analysis were virtually identical to the ANCOVA testing the first three hypotheses. The only significant standardized coefficients that emerged were those for group size (351, moral judgment development (-.31), and subject gender (-.53). None of the interaction effects were significant. Wilks' lambda was .90, eigenvalue = .109, with a chi-square = 30.11, p < .001.

Likewise, the ANCOVA revealed that no interaction effects were significant. Below we discuss the results pertinent to each hypothesis in turn. Hypothesis 1 predicted that threatened retaliation would reduce whistle blowing, but no main or interaction effect involving threatened

286 MICELI. DOZIER, AND NEAR

retaliation was significant at .05 or less. Hypothesis 2 predicted that in- ternaIs would be more likely than externals to blow the whistle, particularly in situations where conditions are not supportive of whistle blowing (i.e., retaliation is threatened). There was no significant main effect for locus of control, and planned contrasts revealed no support for this hypothesis, t(62.1) = -0.54; p < .59. Further, no interaction effects involving locus of control approached significance. As anticipated, level of moral development was related to whistle blowing, but not in the direction predicted by Hypothesis 3; subjects who had high moral development scores and who were not threatened were not more likely than low "Percent I"' subjects to blow the whistle, f(49.6) = 0.45; p < .66. However, a main effect for level of moral development was observed; as Table 1 shows, subjects with lower moral development scores were more likely to blow the whistle than were subjects with higher scores, F(1, 277) = 4.76; p <: 0.03.

As predicted in Hypothesis 4, a significant gender effect was observed, F(1, 277) = 8.23, p < 0.004. But, contrary to our prediction that any effects associated with subject gender would disappear when the impact of dispositional variables was held constant, it could not be accounted for by dispositional effects. As shown in Table I, men were more likely to blow the whistle than were women. When the effects of the dispositional scores were controlled in a second ANCOVA, the gender effects were virtually unchanged, F(1,277) = 7.99, p < .005.

Effects of the Independent Variables and Cuvariate on Conformity

The results of ANCOVA showed that no main effects were significant, nor was the covariate group size related to conformity. Only one significant interaction effect emerged, F(1, 278) = 9.78, p < .002; it was a three-way interaction involving retaliation, locus of control, and gender. For men, conformity was more pronounced when locus of control was internal and no retaliation was threatened and when locus of control was external and retaliation was threatened. For women, just the opposite held true. Conformity was more pronounced when locus of control was external and no retaliation was threatened, and when locus of control was internal and retaliation was threatened.

Discussion

Five major findings emerged from this study. Each is considered in turn, below.

WHISTLE-BLOWING 287

Rates of Whistle blowing

The first finding concerns the process of self-correction in the scientific literature (Tangney, 1987), completely apart from the specific hypotheses concerning whistle blowing. Reminiscent of studies of bystander intervention and of field studies of whistle blowing, most of the persons who observed apparent wrongdoing committed by a research assistant did not report it. This was true even though, as the data showed, the activity they observed was judged to be clearly wrongful by other parties who had no opportunity to report wrongdoing and hence no motivation to distort their perceptions to be consistent with their inaction. Data showed that this was true even though the wrongdoing indirectly harmed the subjects through harming the reputation of their organization (the university) and even though subjects were not required to identify themselves. This finding thus corroborates those of Tangney (1987), who found that scientists were reluctant to report suspected research fraud and described how this reluctance undermines the process of self-correction in the scientific literature. These studies suggest disconcertingly that if scientists and subjects are unlikely to report concerns, wrongdoing and its consequences may continue. Thus, i t would be helpful to find ways to encourage valid reporting of questionable activity. Further attention should be devoted to this important topic.

Number of Observers

Second, observers of wrongdoing were more likely to blow the whistle when more, rather than fewer, other observers were present. This finding is similar to that reported in a recent field study (Miceli & Near, 1988)-that whistle blowing occurred more frequently in larger groups; it suggests that the field result is not attributable to self-report bias. In- terestingly, this finding appears to be in opposition to the diffusion of responsibility effect identified in other studies of prosocial intervention (e.g., Latane & Darley, 1970). While one might argue that the present subjects were influenced differently because the presence of additional others increases the possibility that at least one other person would report the wrongdoing, thus implying that the nonreporting subject is a moral failure, this speculation is questionable for at least three reasons. First, the same reasoning could be used to suggest that bystanders would be more likely to report crime when more others are present. But this is


opposite to the effect that has been observed in bystander intervention studies. Second, the present subjects were anonymous to the wrongdoer and to the complaint recipient at the time of reporting, and the reporting opportunity was designed so that the subject could not be aware of whether others had reported the wrongdoing, precisely to minimize the social impact. Therefore, no one could learn who had failed to report the wrongdoing and the nonreporter would be unlikely to fear that some- one could identify and humiliate the nonreporter. Third, unlike reporting obvious criminal assault to the police, which most persons would agree is morally correct, as noted earlier, whistle blowing frequently may raise moral conflicts. Different parties may view it to be appropriate under different circumstances. Thus, a nonreporting subject could believe that others might not condemn his or her refusal to be a “fink.” It is therefore unlikely that the fear of potential embarrassment accounted for the present results.

The group size finding also is consistent with Latane’s (1983) social impact theory, which views social influence to be a function of the number of sources of influence (and other factors). Unfortunately, this theory docs not specify the direction of influence; that is, alone, it is of limited value in predicting reactions to whistle blowing because it does not at- tempt to explain the conditions under which others may influence the observer of wrongdoing to report or to refrain from reporting. However, we can speculate that if (a) the Research Assistant is viewed as an influence agent discouraging whistle blowing and (b) the other members of the group were influence agents potentially supporting whistle blowing, then the other members’ influence would be less powerful in smaller rather than larger groups, because the number of sources is smaller. Or perhaps observers of wrongdoing fear that their reports would be less easily corroborated when few observers are present, that the foundation for their accusations would be viewed as “your word against his.” A third possibility may be that i t would be easier for the RA to identify and retaliate against subjects in smaller groups. More research is needed to clarify why the presence of others appears to facilitate rather than inhibit whistle blowing-even when it is anonymous and observers cannot know how others have responded to the wrongdoing.

Gender

A third important finding in this study was that men were more

WHISTLE-BLOWING 289

likely to blow the whistle than were women, even when levels of moral development and locus of control were held constant. Future research should explore the generalizability of this finding, and determine the reasons why men were more likely to act. One obvious possibility is that observer gender may interact with the gender of the wrongdoer and with the gender of the complaint recipient. Observers may be more comfortable reporting wrongdoing to members of the same gender, particularly if the wrongdoer is of the same gender. In the present study, in which other variables were the primary focus, the gender of the wrongdoer and that of the complaint recipient were held constant to reduce extraneous variance. The wrongdoer was male, and the complaint recipient-represented by a male student member of a faculty committee in a college in which nearly 90% of the faculty were male-was probably perceived as male. This may have depressed women’s, and increased men’s, propensity to report wrongdoing. On the other hand, Gelfand et al.’s (1973) findings suggest that this hypothesis will not be confirmed in future studies of gender and whistle blowing. In that study, all wrongdoers were women (confederate shoplifters) and although the authors did not report the gender com- position of the group of complaint recipients (cashiers at a convenience store), it is likely that at least some of them were women. Yet, as described earlier, the propensity of women and men to report wrongdoing was remarkably similar to the present results.

A related possibility is raised by a recent study of the use of workplace power strategies (Mainiero, 1986). Mainiero found that, when confronted with a frustrating workplace situation on which they depended, women were more likely to use an ”acquiescence strategy” in which the ”low-power individual accepts the power imbalance and decides that nothing else can be done in the situation” (p. 643). Mainiero identified two perspectives that predicted this finding: (a) that struc- tural segregation (into less powerful organizational positions) has caused women to be less likely to exert power forcefully, and (b) that early socialization shapes men’s and women’s behavior differently; men are perceived as and rewarded for using “direct aggressive strategies,” whereas women are expected to remain powerless. Both the structuralist perspective and the socialization perspective suggest that women feel less capable of bringing about the change in a male- dominated organization. Thus, future research should explore whether the gender and other characteristics of the wrongdoers, perceived vic- tims, and complaint recipients affect whistle blowing directly, or


whether these characteristics interact with the characteristics of the observer.

Moral Judgment Development

The fourth major finding in the present study was that whistle blowing was associated with lower rather than higher levels of moral judgment development. It should be noted that scorers in the upper half of the distribution were not necessarily ”principled moral reasoners,” whose scores are above 50. However, as Rest (1979, p. 5.4) noted, there are not many persons who score so highly, and our subjects scored very similarly to the norms reported by Rest. Nevertheless, this finding was surprising, because it appeared to be contrary both to findings of a previous study (Brabeck, 1984) and to predictions based on a prosocial perspective; that is, moral judgment development is thought to increase the likelihood that an immoral event is viewed as wrongful and/or that one has a moral obligation to intervene to stop wrongdoing where the situation suggests that whistle blowing is appropriate.

However, the present findings can be reconciled with prior empirical research and theory. As noted earlier, Brabeck’s “whistle blowers” were asked by the ”wrongdoer” to find manuscript errors. Little moral conflict would be generated by such a request. In the present study, and probably in most “real world” cases of whistle blowing, more moral conflict would be evoked. The wrongdoer obviously acted wrongfully, and reporting was done not at the wrongdoer’s request, but to a powerful third party who could punish the wrongdoer. Clearly the wrongdoer would not appreciate reports of his wrongdoing. Further, research shows that norms against ”finking” are well established in childhood (Dozier & Miceli, 1985). Consequently, the observer may wonder if the ends justify the means. Where moral conflict exists, highly moral subjects may be less likely than “low moral” subjects to define whistle blowing as an appropriate soIu tion.

Other issues pertaining to moral judgment development should be considered. In the present study, the costs and benefits of action versus inaction may have been weighed differently by persons with differing moral development scores. Finally, the issue of social desirability must be taken into account. Individuals who scored highly on the moral judgment development scale may have been more sensitive than low scorers to social approval and may have been reluctant to blow the whistle because whistle blowing may violate social norms. Future researchers

WHISTLE-BLOWING 291

may want to identify conditions under which moral judgment development will either increase or decrease whistle blowing and explore the reasons why.

Conformity

As was the case here, whistle blowing often occurs in the context of ”crimes of obedience’’ (Hamilton, 1986; Kelman & Hamilton, 1989) or ”administrative obedience” (Meeus & Raaijmakers, 1986). These para- digms are based directly or indirectly on the older conformity literature, following Milgram’s (1974) pioneering work. That is, whistle blowers are often asked to participate in either the wrongdoing itself or the en- suing coverup (Kelman & Hamilton, 1989). Whistle blowing and conformity may thus be related actions, depending on the particular incident of wrongdoing. Under this assumption, those conditions that seem to encourage individuals to engage in crimes of obedience might also cause them not to blow the whistle. According to Kelman and Hamilton (1989) three social processes-authorization, routinization, and dehumanization-encourage crimes of obedience:

Through authorization, the situation becomes so defined that the individual is absolved of the responsibility to make personal moral choices. Through routinization, the action becomes so organized that there is no opportunity for raising moral questions. Through dehumanization, the actors’ attitudes toward the target and toward themselves become so structured that it is neither neces- sary nor possible for them to view the relationship in moral terms (p. 16).

Of these three processes, probably routinization was the only one involved in the conformity of subjects involved here. Given the relatively low rate of conformity (compared to, e.g., Meeus & Raaijmakers, 1986) we might specdate that the presence of only one of the three possible social processes prevented subjects from conformity to the extent that they might otherwise have done. Whether these same three social processes might also discourage whistle blowing was not directly tested here; it is of interest to note the relatively low correlation between conformity and whistle blowing found in these data, however. Such results suggest that whistle blowing and conformity are distinct phenomena and should be investigated as such, unless further evidence would

292 MlCELl DOZIER, AND NEAR

indicate otherwise. Clearly, additional research is needed to consider this issue.

Additional Suggestions for Future Research

Obviously, each of the issues considered above should be examined in additional research. Future research also should address the role of threatened and actual retaliation. Detrimental effects of threatened retaliation did not emerge in this study. Perhaps the threat of retaliation was not sufficiently powerful to induce reactions. However, the findings were the same as in field survey studies in which very severe forms of retaliation have been threatened and experienced. As noted earlier, threatened retaliation and retaliation have been examined in several studies of whistle blowing under diverse conditions, and not one of them has shown that threats discourage whistle blowing or that retaliation reduces the propensity to blow the whistle again.

The consistency of these findings suggests that (a) the ”artificiality” of the present setting probably does not account for the present finding of no relationship between threats and whistle blowing, and (b) quite simply, retaliation may not play a role in whistle blowing decisions. However, i t is also conceivable that a number of variables interact with threatened retaliation to cause whistle blowing. For example, if a problem is intolerable and no alternatives to whistle blowing exist, then one may risk serious retaliation and blow the whistle (Dozier & Miceli, 1985). But if a problem is not viewed as important, or if a less risky means of resolution is available, then threatened retaliation may ”chill” whistle blowing. Thus, more research is needed to determine the role of retaliation.

Given that only a few variables could be examined in the present study, it may be useful to speculate as to other variables that might be examined in future research. In particular, variables pertaining to the position held by the whistle blower may be important. For example, the criticality of the position may determine the potential whistle blower’s influence; individuals occupying more critical positions may be more likely to blow the whistle than may other organization members. Conceivably, many variables such as experience, perceived competence, and others’ liking of the observer of wrongdoer may also enhance perceived influence. Further, the literature on personal control (eg., Green- berger & Strasser, 1986) suggests that individuals’ desire for control over their work environments, and the discrepancy between this desire and

WHISTLE-BLOWING 293

the control they believe they possess, is an important determinant of several organizational behaviors. Because whistle blowing may in some cases represent the assertion of control in the workplace, personal control may be related to whistle blowing. In addition, researchers could consider whether dispositional factors that would lead individuals to desire more control beyond their own immediate work environments would be related to whistle blowing.

Clearly, a great deal remains to be learned about whistle blowing in organizations. It is hoped that this study will stimulate future research both in the laboratory and in field settings.

References

Ashkanasy, N.M. (1985). Rotter's internal-external scale: Confirmatory factor analysis and correlation with social desirability for altema- tive scale formats. Journal of Personality and Social Psychology, 48, 5, 1328-1241.

Bandura, A. (1977). Social learning theory. Englewood Cliffs, NJ: Prentice Hall.

Biondo, J., & MacDonald, A.P., Jr. (1971). Internal-external locus of control and response to influence attempts. Journal of Personality, 39, 4074 19.

Brabeck, M.M. (1984). Ethical characteristics of whistle-blowers. Journal of Research in Personality, 18/41-53.

Brehm, J.W. (1972). Responses to loss of freedom: A the0 y of psychological reactance. Morristown, NJ: General Learning Press.

Brief, A., & Motowidlo, S. (1986). Prosocial organizational behaviors. Academy of Management Review, 11,710-725.

Dozier, J.B., & Miceli, M.P. (1985). Potential predictors of whistle-blowing: A prosocial behavior perspective. Academy of Management Review, 10,823-836.

Eagly, A.H. (1983, September). Gender and social influence: A social psychological analysis. American Psychologist, pp. 971-981.

England, P. (1979). Women and occupational prestige: Acase of vacuous sex equality. Signs, 5,252-265.

Ewing, D.W. (1983). Do it m y w a y - o r you'refired! New York: Wiley. Farrell, D., & Petersen, J.C. (1982). Patterns of political behavior in

organizations. Academy of Management Review, 7,403412. Gelfand, D.M., Hartmann, D.P., Walder, P., & Page, B. (1973). Who

reports shoplifters? A field-experimental study. Journal of Personality and Social Psychology, 25,276-285.


Graham, J.W. (1983). Principled organizational dissent. Unpublished doc- toral dissertation, Northwestern University.

Greenberger, D.B., & Strasser, S. (1986). The development and application of a model of personal control in organizations. Academy of Management Review, 11,164-177.

Hamilton, V.L.. (1986). Chains of command: Responsibility attribution in hierarchies. journal of Applied Social Psychology, 16,118-138.

Jensen, J.V. (1987). Ethical tension points in whistleblowing. Journal of Business Ethics, 6,321-328.

Kelman, H.C.. & Hamilton, V.L. (1989). Crimes of obedience: Toward a social psychology of authority and responsibility. New Haven, CT Yale University Press.

Kohlberg, L. (1969). Stage and sequence: The cognitive developmental approach to socialization. In D. Goslin (Ed.), Handbook of socialization theory and research (pp. 347480). Chicago: Rand McNally.

Kolarska, L., & Aldrich, H. (1980). Exit, voice, and silence: Consumers’ and managers’ responses to organizational decline. Organizational Studies, 1,41-58.

Latank, B. (1983). The psychology of social impact. American Psychologist,

Latan6, €3. & Dariey, J.M. (1970). The unresponsive bystander: W h y doesn’t he help? New York: Appleton-Century-Crofts.

Mainiero, L.A. (1986). Coping with powerlessness: The relationship of gender and job dependency to empowerment-strategy usage. Ad- ministrative Science Quarterly, 31,633-653.

Meeus, W.H.J., & Raaijmakers, Q.A.W. (1986). Administrative obedience: Carrying out orders to use psychological-administrative violence. European journal of Social Psychology, 16,311-324.

Miceli, M.P., dr Near, J.P, (1985). Characteristics of organizational climate and perceived wrongdoing associated with whistle-blowing decisions. Personnef Psychofogy, 38,525-544.

Miceli, M.P., & Near, J.P. (1988). Individual and situational correlates of whistle-blowing. Personnel Psychology, 41,267-282.

Milgram, S. (1974). Obedience to authority. New York: Harper & Row. Near, J.P., & Jensen, T.C. (1983). The whistle-blowing process: Re-

taliation and perceived effectiveness. Working and Occupations, 10, 3-8.

Near, J.P., & Miceli, M.P. (1985). Organizational dissidence: The case of whistle-blowing. Journal of Business Ethics, 4,l-16.

Near, J.P., & Miceli, M.P. (1 986). Retaliation against whistle-blowers: Predictors and effects. lournaf of Applied Psychology, 71,137-145.

36,343-356.

WHISTLE-BLOWING 295

Near, J.P., & Miceli, M.P. (1987). Whistle-blowers in organizations: Dissi- dents or reformers? In Barry M. Staw and L.L. Cummings (Eds.) Research in organizational behavior: Vol. 9 (pp. 321-368). Greenwich, CT: JAIPress.

Perrucci, R.M., Anderson, R.M., Schendel, D.E., & Trachtman, L.E. (1980). Whistle-blowing: Professionals’ resistance to organizational authority. Social Problems, 28,149-164.

Rest, J. (1979). Developmenf in judging moral issues. Minneapolis: Univer- sity of Minnesota Press.

Rotter, J.B. (1966). Generalized expectancies for internal and external control of reinforcement. Psychological Monographs, 80,l-28.

Rushton, J.P. (1984). The altruistic personality: Evidence from laboratory, naturalistic, and self-report perspectives. In E. Staub, D. Bar-Tal, J. Karylowski, & J. Reykowski (Eds.), Developmenf and main- tenance of prosocial behavior (pp. 271-290). New York: Plenum Press.

Salancik, G., & Pfeffer, J. (1978). A social information processing approach to job attitudes and task design. Administrative Science Quarterly, 23,224-253.

Shotland, R.L., & Goodstein, L.I. (1984). The role of bystanders in crime control. journal of Social Issues, 40,9-26.

Shotland, R.L., & Heinold, W.D. (1985). Bystander response to arterial bleeding: Helping skills, the decision-making process, and differen- tiating the helping response. Journal of Personality and Social Psychol- om, 49,347-356.

Skinner, B.F. (1953). Science and human behavior. New York: Free Press. Spector, P.E. (1982). Behavior in organizations as a function of em-

Staub, E. (1974). Positive social behavior and morality: Social and personal in-

Tangney, J.P. (1987, August 6). Fraud will ou t -o r will it? New Scienfisf ,

Trevino, L.K. (1986). Ethical decision making in organizations: A person-situation interactionist model. Academy of Management Review,

ployee’s locus of control. Psychological Bullefin, 91,482487.

fluences (Vol. 1). New York: Academic Press.

115,62-63.

11,601-607. Weinstein, D. (1979). Bureaucratic opposition. New York Pergamon.

blowing the whistle on data fudging: a controlled field experiment1

Documents