See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/269510375

Interpersonal Aggression in Work Groups: Social Influence, Reciprocal, and

Individual Effects

Article  in  The Academy of Management Journal · August 2003

DOI: 10.2307/30040640

CITATIONS READS
311 2,987

2 authors:

Theresa M Glomb Hui Liao

University of Minnesota Twin Cities University of Maryland, College Park
40 PUBLICATIONS   3,432 CITATIONS    56 PUBLICATIONS   7,877 CITATIONS   

SEE PROFILE SEE PROFILE

All content following this page was uploaded by Theresa M Glomb on 18 February 2015.

The user has requested enhancement of the downloaded file.

® Academy of Management journal
2003, Vol. 46, No. 4, 486-496.

INTERPERSONAL AGGRESSION IN WORK GROUPS: SOCIAL

INFLUENCE, RECIPROCAL, AND INDIVIDUAL EFFECTS

THERESA M. GLOMB
University of Minnesota

HUI LIAO
University of Illinois at Urbana-Champaign

This cross-level study of 149 employees from 25 groups demonstrates the impact of
group social context on individual interpersonal aggression. Extending the work of
Robinson and O'Leary-Kelly (1998), results suggest that both being the target of aggres-
sion and the mean level of aggression in a work group (absent the target individual) are
predictors of employees' reports of engaging in aggression. Effects persisted when
individual differences related to aggression, demographics, and situational variables
were controlled. Results suggest individual, reciprocal, and group influences.

In recent years, workplace aggression has gar- 1996; Folger & Baron, 1996; Glomb, 2002; Neuman
nered increasing attention from researchers, practi- & Baron, 1997, 1998; O'Leary-Kelly, Griffin, &
tioners, the media, and the general public. Al- Glew, 1996; Tepper, 2000).
though the focus of the populace has been on Although a knowledge base is accumulating,
extreme violence and physical forms of aggression much of the work on aggressive behaviors and
(such as homicide and physical attacks), accumu- other forms of antisocial behavior at work has been
lating data on incidence rates suggests that work- limited by approaching these behaviors as strictly
place aggression has crept into organizations in a individual-level phenomena, thereby ignoring so-
variety of insidious ways. Employees frequently cial and work group factors that may influence
report being the targets of relatively mild manifes- individual employee behavior. One exception is
tations of aggression, including forms of verbal Robinson and O'Leary-Kelly's 1998 work suggest-
harm such as talking behind others' backs and in- ing the antisocial behavior of work group members
terrupting others when they are speaking; among influences an individual employee's antisocial be-
the least frequently reported forms of aggression are havior. The current work was an attempt to extend
overt acts of physical aggression such as physical the theoretical and empirical work of Robinson and
assault and attack with a weapon (Baron & Neu- O'Leary-Kelly in several ways.
man, 1996, 1998; Glomb, 2002; Greenberg & First, we studied a different set of behaviors,
Barling, 1999; Neuman & Baron, 1998; Northwest- examining interpersonal workplace aggression
ern National Life Employee Benefits Division, rather than antisocial behaviors. In doing so we
1993). Available information on employees' self- examined the extent to which an overlapping but
reports of perpetrating aggression (see Baron, Neu- distinct set of behaviors functions similarly to the
man, & Geddes, 1999; Glomb, 2002) suggests con- antisocial behaviors investigated in the Robinson
sistency with the self-reports of recipients of and O'Leary-Kelly work. Second, we assessed the
aggression; relatively mild aggressive behaviors are influence of individual differences variables rele-
enacted most often. Research on prevalence has vant to aggression. Individual differences anteced-
been complemented by empirical and theoretical ents were not accounted for in Robinson and
work examining the antecedents and outcomes of a O'Leary-Kelly's work; these individual differences
wide array of aggressive behaviors at work (Barling, may drive the link between individual and work
group levels of aggression. Third, we posited an
additional social or dyadic influence on individual
The research reported here was supported by a Grant- employee aggression by examining social exchange
in-Aid of Research, Artistry and Scholarship from the
or reciprocal explanations in addition to the effects
University of Minnesota. We are grateful to the organiza-
posited by Robinson and O'Leary-Kelly. The origin
tion that participated in this research; to Michael Tews
and Roxanne Laczo, who assisted in data collection; and of the positive relationship between group and in-
to Joyce Bono, Andrew Miner, Paul Sackett, and Connie dividual employee aggression could be multiply
Wanberg for their comments on an early version of this determined. In this study, we posited and tested
paper. multiple individual and group influences using

486


2003 Glomb and Liao 48 7

data from employees working in 25 health care O'Leary-Kelly addressed have different targets and
homes. could have different correlates. Combining them
may obfuscate the functioning of these relations.
Our focus here is on one set of behaviors, interper-
THEORETICAL BACKGROUND AND sonal aggressive behaviors directed at other indi-
HYPOTHESES viduals with the intent of inflicting harm, rather
than on behaviors directed at an organization or
Conceptualizing Workplace Aggression
purposeful attempts to impede performance. De-
Our conceptualization of workplace aggression is lineation of the determinants of interpersonal
similar to that of Neuman and Baron, who defined aggression is important to further researchers'
workplace aggression as "efforts by individuals to understanding of the phenomenon and to guide
harm others with whom they work, or have prevention efforts. We leave investigation of aggres-
worked, or the organizations in which they are sion directed toward organizations for future
presently, or were previously, employed" (1998: research.
395); however, we focus on interpersonal forms of
aggression rather than on aggression directed at an Potential Influences on Individual Aggression
organization. Following an approach similar to that
taken by previous aggression researchers (Baron & Robinson and O'Leary-Kelly used three theoreti-
Neuman, 1998; Folger & Baron, 1996; Neuman & cal perspectives to explain the potential influence
Baron, 1998), we tried to capture the behavioral of groups on individual antisocial behaviors: social
construct space by incorporating the assessment of information processing theory (Salancik & Pfeffer,
low-level, less extreme aggressive behaviors (such 1978), social learning theory (Bandura, 1977), and
as yelling at another person, talking behind another the attraction-selection-attrition perspective (ASA;
person's back, and withholding needed resources Schneider, 1975). We propose two additional de-
from another) in addition to assessment of the ex- terminants of individual interpersonal aggression:
treme behaviors usually associated with workplace reciprocal aggression and aggression-related indi-
violence (such as physical assault and threats). vidual differences variables.
There exists a degree of overlap between the be-
haviors in our workplace aggression construct and Social Information Processing, Social Learning
those in other constructs, including organizational
Theory, and the ASA Model
retaliatory behaviors (Skarlicki & Folger, 1997), re-
venge responses (Bies, Tripp, & Kramer, 1997), Social information processing, social learning
workplace deviance (Robinson & Bennett, 1995), theory, and the attraction-selection-attrition model
incivility (Andersson & Pearson, 1999), and emo- all support the proposition that work group aggres-
tional abuse at work (Keashly & Jagatic, 2003). Most sion and individual employee aggression should be
relevant for the current study is the distinction positively related, although they imply slightly dif-
between our conceptualization of interpersonal ferent mechanisms as driving the relationship. The
workplace aggression and Robinson and O'Leary- social information processing approach (Salancik &
Kelly's conceptualization of antisocial behavior, Pfeffer, 1978) suggests that individual employees
which they measured using nine items, such as use information about values, norms, expectations,
"damaged property belonging to my employer," and behavior outcome contingencies gathered from
"criticized people at work," "started an argument others in their social environment to guide behav-
with someone at work," and "did work badly, in- ior. For example, employees may receive informa-
correctly or slowly on purpose" (Robinson & tion about the outcomes of different aggressive be-
O'Leary-Kelly, 1998: 663). Although the authors haviors; a heated disagreement may be tolerated,
provided evidence of the construct validity of the wheraas verbally abusive behavior may be grounds
antisocial behavior construct, the component items for discipline. Under a social learning framework
assess several different forms of antisocial behav- (Bandura, 1977), individual behavior is influenced
ior. Like Robinson and Bennett's construct of pro- by role models for behavior: an individual em-
duction deviance, Robinson and O'Leary-Kelly's ployee acquires a repertoire of aggressive behaviors
construct includes behaviors directed at organiza- by observing others' behaviors and their conse-
tions. Robinson and O'Leary-Kelly's construct also quences. According to the attraction-selection-
includes behaviors directed at work group mem- attrition model (Schneider, 1975), individuals with
bers. We view the second type of behaviors as more characteristics similar to those of other members of
rightly defined as interpersonal workplace aggres- a work group are attracted to, selected into, and
sion. The two types of aggression Robinson and retained by the group; individuals with aggressive


48 8 Academy of Management Journal August

tendencies would be more likely to be attracted havior, he or she is likely to become the target of
to, selected by, and remain in work groups with aggression from the person against whom he or she
similarly aggressive members, resulting in a work aggressed; a dyadic process is suggested. However,
group with relatively homogeneous aggressive despite the notion that being the target of aggres-
tendencies. sion may elicit a like or a more serious response,
We feel the same processes used to explain the there is little empirical evidence accounting for the
group effects of antisocial behavior on individual influence of reciprocal processes in research on
behavior can also apply to the behavioral construct workplace aggression. Andersson and Pearson
of interpersonal aggression. Thus, we propose the (1999) provided a theoretical explication of the "in-
following hypothesis, which is analogous to a hy- civility spiral" in which a tit-for-tat pattern creates
pothesis put forth by Robinson and O'Leary-Kelly: an escalatory series of uncivil behaviors. Glomb's
(2002) empirical examination of the patterning of
Hypothesis 1. The level of aggression of a focal aggressive behaviors within a specific incident of
individual in a work group will be positively workplace aggression also suggests an escalatory
related to the aggregate level of aggression of pattern.
all other group members. Ideas of reciprocal aggression have their under-
These group-level influences operate on individ- pinnings in social exchange theory, which suggests
ual aggression by providing shared information social exchange relationships develop between two
about behavioral norms and expectations, exposure parties through a series of mutual exchanges that
to aggressive role models, and homogeneous ag- yield reciprocal activities from each party (Blau,
gressive tendencies within a group. Individuals ag- 1964); for example, an employee may reciprocate
gress largely because aggression is present in their unfair or adverse treatment from work group mem-
social context. A parallel idea has been discussed bers. A reciprocal or social exchange explanation is
in the sexual harassment literature as "ambient sex- not proposed as an explicit influence in the Robin-
ual harassment" (Glomb et al., 1997), a concept son and O'Leary-Kelly work, although it is not at
based on work by Hackman (1992), in which he odds with group-level explanations; aggression be-
proposed that characteristics of a work environ- gets aggression among members of a work group.
ment may be conceived of as either ambient stimuli We explicitly tested the influence of the reciprocal
that pervade the group setting and are potentially explanation using the following hypothesis:
available to all group members or discretionary Hypothesis 2. Employees' reports of being the
stimuli that are transmitted to individuals differen- targets of aggression will be positively related
tially. The operation of work group aggression as an to reports of engaging in aggression.
ambient stimulus would be largely consistent with
explanations offered by Robinson and O'Leary- Individual Differences Influences
Kelly: work group aggression influences individu-
als' aggression as a function of their being members As Barling noted, "Aggression is a highly stable
of a work group pervaded by aggression. An addi- behavior" (1996: 34). Extensive empirical and the-
tional effect might result from discretionary stim- oretical work suggests that individual characteris-
uli, present for an individual in the form of being a tics are related to the likelihood of engaging in
direct target of aggression from another work group aggression; characteristics identified as related to
member. There is an important distinction between this likelihood include a hostile attribution bias, in
being in a work group where people behave which individuals perpetually attribute hostile in-
aggressively (ambient) and being in a work group tentions to others (Neuman & Baron, 1998); the type
where people behave aggressively toward you (dis- A behavior pattern (Baron et al., 1999; Neuman &
cretionary). These discretionary stimuli in a work Baron, 1998); state anger (Chen & Spector, 1992);
group may be captured in the notion of reciprocal lack of self monitoring (Neuman & Baron, 1998);
aggression. and a personal history of aggression (Greenberg &
Barling, 1999). This work suggests individual dif-
ferences account for variance in engaging in aggres-
Reciprocal Aggression and Social Exchange
sion; thus, it may be that employees' aggressive
Theory characteristics drive a work group effect. Although
Bandura (1973) suggested that the most reliable Robinson and O'Leary-Kelly included demo-
way of eliciting aggression is by assaulting or graphic variables in their analysis, they stated the
threatening another. Researchers have long sug- following: "Ideally we would have more directly
gested that aggression is reciprocal in nature, in measured and controlled for antisocial disposi-
that if one individual engages in an aggressive be- tions" (1998: 670). By including individual differ-


2003 Glomb and Liao 489

ences variables, one can determine the group and Controlling for Perceptions of Job and
dyadic impact on aggression over and above the Organizational Features
effects of individual differences. We propose that
We controlled for employee perceptions of job-
reactions to anger, negative affectivity, and self-
monitoring are related to engaging in aggression. and organization-related factors so that we could be
Reactions to anger. Trait anger, "the disposition more confident about capturing variance accounted
for by social influences over and above that ac-
to perceive a wide range of situations as annoying
counted for by contextual factors that may result in
or frustrating, and the tendency to respond to such
situations with more frequent elevations in state aggression (and may be shared by group members).
anger" (Spielberger, 1991: 1), is reflected in indi- Guided by previous research, we included percep-
viduals' reactions to anger. We assessed these rel- tions of organizational justice, stress, and work sat-

Theoretical and anecdotal evidence for a rela-

atively stable individual differences in the manner isfaction as controls in our analyses.

tionship between organizational justice and aggres-

in which individuals react to anger in general by
sion (Barling, 1996; Folger & Baron, 1996; Neuman
assessing whether angry feelings are generally sup-
& Baron, 1998) is only somewhat supported by
pressed or outwardly expressed toward other peo-
ple or toward objects. empirical research, which suggests mixed findings.
Empirical results typically show a restricted set of
Negative affectivity. Negative affectivity is con-
relations, often suggesting links between a subset
ceptualized here as a stable, dispositional trait re-
of the justice components (that is, procedural, dis-
sulting in a tendency to view things negatively
(Watson, Clark, & Tellegen, 1988). Negative affect, tributive, interactional justice) and aggression or
as a negative emotional state, has been associated suggesting links only for particular targets of
aggression (Baron et al., 1999; Folger et al., 1998;
with increased aggression (Berkowitz, 1993). If a
Greenberg & Barling, 1999).
transitory negative affective state is related to ag-
gression, then, by extension, those who possess a
Research has suggested a relationship between
more permanent negative disposition experience
stress and aggression (Chen & Spector, 1992; North-
negative affect more often, increasing the likeli- western National Life Employee Benefits Division,
hood of the individual's engaging in aggression. 1993). However, the causal direction of the rela-
Self-monitoring. Self-monitoring refers to a pro-
tionship remains ambiguous; data suggest that
cess of "self-observation and self-control guided by
stress can be both a cause and an outcome of ag-
situational cues to social appropriateness" (Snyder, gression (Northwestern National Life Employee
1974: 526). People who score high on measures of Benefits Division, 1993). We used stress as a con-
self-monitoring ("high self-monitors") regulate
trol to account for this potential influence on em-
their expressive self-presentation and are respon-
ployees' aggressive behavior.
sive to social and situational cues about desired or
Like stress, job satisfaction has been proposed as
appropriate behavior (Snyder & Gangestad, 1986).
both an antecedent and an outcome of workplace
"Low self-monitors" do not regulate their expres- aggression (Chen & Spector, 1992; Glomb, 2002;
sive self-presentation and instead base responses
Tepper, 2000). Here, we controlled for an individ-
on internal states, attitudes, and feelings (Snyder &
ual's attitudes about work, given the evidence that
Gangestad, 1986). Neuman and Baron (1998) sug-
satisfaction with one's work is more closely related
gested that low self-monitors are more likely to to overall assessments of satisfaction than are other
behave provocatively in conflict situations and
facets of job attitudes.
more likely to misinterpret others' motives because
Note that these job characteristics could be con-
they have low social sensitivity; empirical results
sidered ambient stimuli that pervade a group set-
have supported this proposition. We propose the ting and are potentially shared phenomena. How-
following regarding the influence of individual
ever, we employed these measures at an individual
differences:
level of analysis because they are perceptions of job
features that are likely to exist and operate differ-
Hypothesis 3. Individual differences will be re- ently for different group members.
lated to employees' reports of engaging in ag-
gression; the frequency with which angry feel-
ings are expressed and negative affectivity will METHODS
be positively related to reports of engaging in
Participants and Procedures
aggression, and the frequency with which an-
gry feelings are suppressed and self-monitor- Two-hundred-seventeen employees of an organi-
ing will be negatively related to reports of en- zation that operates assisted-living group health
gaging in aggression. care homes for individuals with disabilities com-


490 Academy of Management Journal August

pleted paper-and-pencil surveys via on-site admin- and convergent validity with alternative measures
istration in conjunction with monthly staff meet- of aggression (e.g., Baron & Neuman, 1996; Neuman
ings. (Twelve employees who were present at these & Baron, 1998), and factor analyses supported a
staff meetings declined to participate, primarily be- unidimensional solution. The scale consists of two
cause they were engaged in work tasks.) Participa- separate 20-item subscales, one in which a respon-
tion in the study was voluntary, and participants dent indicates whether he or she has engaged in the
were ensured confidentiality. Thirty-eight percent aggressive behavior described (AES-engaged in)
of the sample's members were employed full-time, and one in which the respondent indicates whether
90 percent were Caucasian, and 77 percent were he or she has been the target of the aggressive
women; the average tenure was 22 months, and the behavior (AES-target). Respondents are asked to
average age was 24 years. report how often each of the 20 behaviors (for in-
To enable work group analyses, we grouped re- stance, yelling or raising your voice, swearing at
spondents on the basis of the health care home in another person, damaging another person's prop-
which they were employed. Employees within a erty, spreading rumors) occurred using a five-point
home worked closely with one another to provide scale (1, "never," to 5, "once a week or more"). In
effective and coordinated care for the residents. the current study, we removed three items (physi-
Homes were dispersed throughout a metropolitan cal assault, sabotaging another's work, and with-
region and were relatively decentralized, character- holding resources needed to do one's job) when
istics allowing for readily identifiable work groups. scoring the scale because of a lack of variance; all
Thirty participants did not provide home informa- employees responded they had never engaged in or
tion and were eliminated from analyses. Eight work been the target of these behaviors.
groups had data from fewer than three employees Scores on the AES were used in the analyses at
and were eliminated from analyses because mean the work group level in two ways. First, we deter-
work group scores could not be computed reliably. mined an individual's work group aggression score
For the remaining respondents, we imputed miss- by computing the average AES-engaged in subscale
ing values and rounded them to the nearest re- score for all employees in her or his work group,
sponse option using the two-way imputation excluding the focal individual's score. This proce-
method (Bernaards & Sijtsma, 2000) applied on a dure provides an estimate of work group scores,
scale-by-scale basis; no more than two missing val- uncontaminated by the problem of common
ues were imputed for scales with more than nine method variance inherent in using self-report, sin-
items and no more than one missing value was gle-source data. Second, individuals' AES-target
imputed for scales with fewer than nine items. Af- scores were used to test for reciprocal or social
ter "listwise" deletion of individuals with incom- exchange effects.
plete information, the final sample consisted of 149 Individual differences antecedents. Negative
participants. Work group size ranged from 3 to 12 affectivity was assessed via the Positive Affect Neg-
members, with a mean of 6. T-tests of sample ative Affect Schedule (PANAS; Watson et al.,
means for the 149 usable cases and 68 dropped 1988). Respondents indicate the degree to which
cases revealed that age was the only variable with a each of a series of ten descriptors of negative emo-
statistically significant difference; the dropped tions (such as "distressed," "hostile," and "irrita-
respondents were 3.84 years older on average ble") describes how they generally feel on a five-
(p < .05). point scale (1, "very slightly or not at all," 5,
"extremely").
Self-monitoring was assessed via items from the
Measures Self-Monitoring Scale (Snyder, 1974; Snyder &
Workplace aggression. The Aggressive Experi- Gangestad, 1986). The original 18-item scale was
ences Scale (AES), which was developed by the shortened to 9 items after consultation with the
first author (Glomb, 1998) assesses the frequency original author. Respondents indicated whether or
with which respondents engage in and are the tar- not each of the 9 items described them ("yes" or
gets of aggressive behaviors at work. Development "no"); example items are "I have trouble changing
of the scale relied on interview data, on Buss's my behavior to suit different people and different
(1961) framework classifying behavioral forms of situations" and "I find it hard to imitate the behav-
aggression (as direct and indirect, active and pas- ior of other people."
sive, and verbal and physical), and on the work of To assess anger expression, we used two eight-
Baron and his colleagues (e.g., Folger & Baron, item scales from the State-Trait Anger Inventory
1996; Neuman & Baron, 1997). Scale validity evi- (STAXI; Spielberger, 1991) to assess the frequency
dence suggested considerable conceptual overlap with which respondents generally suppressed or


2003 Glomb and Liao 491

expressed angry feelings toward other people or Work satisfaction was assessed using the work
objects in the environment. The suppressing anger scale of the job Descriptive Index (JDI; Smith, Ken-
subscale is called anger-in, and the expressing an- dall, & Hulin [1969]; revised by Roznowski [1989]).
ger subscale is called anger-out. Using a four-point Using item response theory parameters obtained
scale, respondents indicated how often they engage from prior research, we shortened the 18-item scale
in anger-in behaviors (such as withdrawing from to 9 items for the current research. Participants
people) or anger-out behaviors (such as arguing responded "yes," "no," or "?" to adjectives describ-
with others) when they are angry or furious (1, ing work characteristics.
"almost never," to 4, "almost always").
Perceptions of job and organizational features.
The Stress Diagnostic Survey (Matteson & Ivance-
vich, 1982) was used to assess the degree of stress RESULTS
respondents experienced from organizational, Descriptive statistics, scale reliabilities, and in-
group-related, and job-related stressors. There were tercorrelations are reported in Table 1. In general,
ten items on organizational stress (examples: struc- the bivariate correlations reflect expected relations
ture, change, and leadership); nine items on group- and provide confidence that the measures func-
related stress (examples: norm pressures, lack of tioned properly at the individual level of analysis.
group cohesiveness, and inadequate group sup- Results for the regression analysis of employee
port); and ten items on job stressors (examples: role aggression are presented in Table 2. Since individ-
conflict and ambiguity, work overload, time pres- uals in a group are exposed to similar stimuli, the
sures, and working conditions). Respondents indi- assumption of independent observations that un-
cated the frequency with which each condition was derlies traditional ordinary least squares (OLS)
a source of stress using a five-point scale (1, "never regression analysis was violated in the current
a source of stress," to 5, "always a source of stress"). sample. As a result, traditional OLS regressions
We adapted an organizational justice scale from generated biased estimates of the standard errors
Moorman (1991) to assess general perceptions of and invalid test statistics. Therefore, we estimated
distributive, procedural, and interpersonal justice. a variance-covariance matrix that allows interde-
The ten modified items asked about the organiza- pendent error terms within groups but indepen-
tion in general (example: "At this organization, su- dence across groups, as well as accounting for het-
pervisors consider employees' viewpoints") rather erogeneous errors across groups (Rogers, 1993). To
than about an employee's specific experience construct this matrix, the conventional variance-
("Your supervisor considered your viewpoint"). covariance matrix is weighted by using contribu-
Response options were "yes," "?," and "no." tions (to the score function) of each group, instead

TABLE 1
Descriptive Statistics, Alpha Coefficients, and Correlation'
Variable Mean s.d. 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
b
1. Tenure 21.94 26.37
2. Genderb 0.77 0.43 -.16
3. Age 23.62 7.46 . 34 . 00
4. Negative affectivity 18.83 5.12 -.02 -.04 -.16 (.85)
5. Anger-out 13.32 3.07 . 00 . 02 -.19 . 34 (.72)
6. Anger-in 15.28 3.85 . 05 -.07 -.04 . 45 . 16 (.77)
7. Self-monitoring 5.05 2.43 -.11 -.09 -.31 . 17 . 36 -.06 (.74)
8. Job stress 11.65 3.81 . 25 -.04 . 24 . 06 -.01 . 05 -.07 (.79)
9. Work group stress 9.43 3.56 . 19 -.11 . 22 . 14 -.03 . 04 -.14 . 42 (.89)
10. Organizational stress 14.54 4.94 . 39 -.17 . 20 . 00 -.04 . 03 -.17 . 58 . 46 (.87)
11. Organizational injustice 15.39 4.97 . 33 -.11 .07 -.05 -.03 . 07 -.11 . 32 . 27 .60 (.88)
12. Work satisfaction 21.38 6.29 -.08 . 21 -.05 -.01 . 15 . 04 . 01 -.08 -.35 -.26 -.26 (.85)
13. Target of aggression 23.51 6.18 . 20 -.08 . 18 . 02 . 12 . 05 . 11 . 40 . 20 .43 . 36 -.07 (.86)
14. Work group aggression' 21.14 2.80 . 12 -.11 .01 . 04 . 10 . 09 . 04 . 04 . 21 . 15 . 20 -.12 . 19
15. Engaging in aggression 21.36 5.25 . 29 -.17 .02 . 21 . 32 . 13 . 10 . 26 . 20 . 27 . 21 -.02 . 52 . 36 (.80)

a n = 149. Alpha coefficients appear on the diagonal in parentheses. Correlations greater than .16 are (in absolute value) significant at

p < .05; those greater than .21 are significant at p < .01.
b
Tenure = months. Female = 1, male = 0.
c Significance designations do not apply to the work group aggression variable because of nonindependence in the data; the point
estimate is accurate.


492 Academy of Management Journal August

TABLE 2
Results of Hierarchical Regression Analysis on Individual Aggressive Behaviors'
Ordinary
Step 2: Individual Step 3: Reciprocal Step 4: Social Least Squares
Variableb Step 1: Controls Differences Effects Influencesc Results

Tenured 0.04** 0.04** 0.04** 0.04*** 0.04**

Genders -1.55 -1.46 -1.45 -1.28 -1.31
Age -0.09* -0.02 -0.08 -0.07 -0.07
Work satisfaction 0.09 0.04 0.03 0.04 0.04
Organizational injustice 0.06 0.07 -0.02 -0.06 -0.05
Job stress 0.20 0.18 0.05 0.10 0.09
Work group stress 0.16 0.11 0.12 0.04 0.06
Organizational stress 0.05 0.07 -0.04 -0.04 -0.04
Negative affectivity 0.09 0.11 0.12 0.12
Anger-out 0.47** 0.39** 0.36** 0.36**
Anger-in 0.01 -0.01 -0.04 -0.03
Self-monitoring 0.06 -0.15 -0.16 -0.16
Target of aggression 0.40** 0.36*** 0.37**
Work group aggression 0.51* 0.43*

R2 . 16 . 26 . 41 . 46 . 46
AR2 . 09 . 15 . 05 . 05
F 9.53** 9.09** 15.65*** 81.69*** 47.36***
df 8, 24 12, 24 13, 24 14, 24 14, 24

a n = 149. Entries are unstandardized regression coefficients. In all regressions, we adjusted the variance and covariance matrixes using
the cluster option in STATA 6.0 to account for interdependent errors within groups and heterogeneous errors across groups. Adjusted R2
and standardized coefficients are invalid in regressions with the cluster option; therefore, only regular R 2 and unstandardized coefficients
are reported.
b All regressions included an intercept, which was omitted from this table.
Estimated using a two-stage least-squares (2SLS) regression.
d
Tenure = months. Female = 1, male = 0.
*p<.05
**p<.01
***p<.001
Two-tailed tests.

of each individual. Analyses were performed with influence of work group members' aggression on
STATA 6.0. individual aggression. To justify using a group-
As a block, the control variables explained 16 level measure of aggression, we first conducted a
percent of the variance in aggressive behaviors, and one-way analysis of variance (ANOVA) to test be-
tenure and age were significantly related to engag- tween-group variance. Results indicate significant
ing in aggression. Younger employees and employ- between-group differences for aggressive behaviors
ees with longer tenure engaged in more aggressive (F24,124 = 2.70, p < .001).
We then computed with-
behavior. in-group agreement (as rwg) using a uniform null
Individual differences variables explained an ad- distribution (James, Demaree, & Wolf, 1984) for the
ditional 10 percent of the variance in aggressive aggression scale and obtained values ranging from
behavior. As expected, anger-out positively pre- . 94 to . 99, with a median of . 98. Although the use of
dicted aggressive behavior (b = 0.47, p < .01). uniform distribution tends to lead to overestima-
However, anger-in, negative affectivity, and self-
monitoring were not significantly related to em- tion of within-group agreement (James et al., 1984),
ployee aggression. Thus, Hypothesis 3 was only the rwg values we obtained in this sample were well
partially supported. above the conventionally acceptable value of . 70.
Hypothesis 2 predicts a reciprocal effect, which To test Hypothesis 1, we entered the mean level
was supported by the results; being the target of of work group members' aggression, computed ex-
work group members' aggression had a positive cluding the focal individual, as a predictor in the
relationship with engaging in aggressive behavior regression. However, in doing so we encountered
(b = 0.40, p < .001) and accounted for an addi- an endogeneity problem, since work group mem-
tional 15 percent of the variance. bers' aggressive behavior was likely to correlate
The next step in the analysis involved testing the with the error terms in the regression. Unmeasured


2003 Glomb and Liao 49 3

factors (for instance, an ambient group characteris- aggressive behavior exhibited by other members of
tic such as a supervisor's punishment policies) as- an individual's work group is a significant predic-
sociated with higher (or lower) levels of aggression tor of the individual's interpersonal aggressive
for both a focal employee and members of his or her behavior.
work group become included in the focal employ- We also account for the role of individual differ-
ee's error term, creating a correlation between the ences in explaining aggressive behavior. While
predictor and error term. This situation violates the Robinson and O'Leary-Kelly included demo-
OLS assumption that each explanatory variable graphic characteristics as proxies for predisposi-
must be independent of the error term and will tions, our results suggest the individual differences
result in biased and inconsistent estimates of the variable anger-out had a significant relation with
coefficients. Thus, we conducted two-stage least aggression. One can have additional confidence in
squares (2SLS) regression, a procedure that em- the group and social effects after aggression-related
ploys instrumental variables to "replace" endoge- individual differences have been accounted for.
nous variables where they appear as predictors (see Our data also suggest a social exchange or recip-
Johnston & DiNardo, 1997). A valid instrumental rocal process as a determinant of individual aggres-
variable should be: (1) a good proxy for the endog- sion. Results suggest that being the target of aggres-
enous variable (in this case, work group aggression) sion is related to engaging in aggression, thus
and (2) uncorrelated with the error term. We con- providing support for a social exchange or reciproc-
structed four instrumental variables for work group ity effect. This reciprocal effect is also consistent
aggression-work group members' average tenure, with the literature on injustice; employees who
average age, average anger-out scores, and the per- perceive that they have been wronged may retaliate
centage of women in the work group, which ex- through aggression, using it as a mechanism to
plained a significant portion of variance in work restore justice (Aquino, Tripp, & Bies, 2001; Bies et
group aggression (R 2 = .30) and thus offered a rea- al., 1997; Skarlicki & Folger, 1997). Although the
sonably good proxy for this endogenous variable. reciprocal process may be considered to be "re-
We then conducted an overidentification test venge," we did not restrict the behaviors to those
(Johnston & DiNardo, 1997) by regressing the 2SLS having a retaliatory intent (Bies & Tripp, 1996; Bies
residuals on the four instrumental variables; results et al., 1997; Tripp & Bies, 1997). Had we the ability
indicated they were not correlated with the error to identify targets and aggressors in our data, we
terms. Therefore, in the following step, which in- might have been able go beyond speculating that
cluded work group aggression as a predictor, we our reciprocal process is akin to that at work in
conducted 2SLS analyses with cluster adjustment retaliation and revenge. However, our data are
for within-group interdependence, using these four limited, showing only the correlational relation-
instrumental variables. Results estimated by tradi- ship between being the target of and engaging in
tional OLS coupled with cluster adjustment are aggression.
also reported. As evident from Table 2, these alter- Overall, our model explained a large amount (46
native methods yield highly consistent results. percent) of the variance in individual aggression
Consistent with Hypothesis 1, work group ag- and suggests that individual aggression is multiply
gression significantly predicted individual aggres- determined.
sive behavior (b = 0.51, p < .05), explaining 5
percent additional variance in employee aggression Managerial Implications
beyond what was accounted for by the control vari-
ables, individual differences, and by the reciprocal Our results underscore the importance of reduc-
effect. Overall the model explained 46 percent of ing aggression in organizations, given the evidence
the variance in employee aggression. that aggression may be "contagious," a view echoed
by Robinson and O'Leary-Kelly (1998). The mech-
anisms through which a group influences individ-
DISCUSSION
ual behavior are multifold, suggesting the value of a
Our study extends Robinson and O'Leary-Kelly's multipronged approach to preventing and reducing
(1998) work on antisocial behavior by examining a aggression. Managers may take preventative action
different construct-interpersonal aggression-and by altering the social information disseminated by
by including two additional determinants of indi- communicating strong behavior-outcome contin-
vidual behavior: a reciprocal or social exchange gencies (for instance, having and enforcing a zero-
effect and individual differences effects. Consistent tolerance approach, and communicating serious
with Robinson and O'Leary-Kelly's findings about consequences for aggressive employees), eliminat-
antisocial behavior (1998), our data suggest that ing aggressive role models, and intervening when


49 4 Academy of Management Journal August

aggressive behavior is likely to be reciprocated or to which relationships are more fluid and different
escalate. Results suggesting the influence of indi- teams, departments, and so forth may intermingle.
vidual differences provide an additional aggression However, such factors as fluctuations in schedules,
reduction strategy, namely, selection processes roles and tasks assigned within the homes, and the
whereby aggression-prone individuals could be level of care required by residents may be sources
identified and screened out. However, selection ef- of variability in the amounts of exposure the mem-
forts may have problems (including faking, adverse bers of a work group had with one another. Future
i mpact, and negative applicant reactions) and tar- research might use alternative methodologies, in
get only one category of precursors of aggression, a particular network analysis, to determine the pat-
category that may address a small amount of vari- tern of interactions within a work group.
ance (10 percent in this study). Training in conflict In the current study, we were unable to deter-
management techniques and self-management mine the sources and targets of aggression; we did
strategies may prove beneficial in their own right not know who was aggressing against whom. Ad-
and also as a signal; allocating resources to training ditional consideration of the dynamics of work-
provides social information signaling the impor- place aggression in a work group may reveal that
tance of aggression prevention. Given that the ex- the experience of aggression may be unduly dis-
planations for aggression are dynamic, the solu- persed to specific individuals, as in instances of
tions are likely to be dynamic as well and will work workplace bullying (Keashly & Jagatic, 2003). Con-
collectively over time to reduce aggression. sideration of workplace bullying may inform the
relationship between work group and individual
aggression; if bullying occurs, we might expect to
Limitations and Future Directions
see group members more likely to engage in aggres-
Our analyses were based on a sample of health sion toward easy targets. Future research might il-
care workers present at staff meetings, a character- luminate this potential work group effect.
istic that limits the generalizability of results. How-
ever, one could argue that this sample produced
Conclusion
attenuated results since health care workers are
expected to be caring and empathetic to those they This study demonstrates that even after individ-
care for and, perhaps by extension, to one another. ual predispositions are accounted for, dyadic and
Additionally, consistency with Robinson and workgroup factors influence individual member
O'Leary-Kelly's study and previous research pro- behaviors. To the best of our knowledge, this was
vides confidence that the current results are not the first study to combine individual and social
sample-specific. characteristics in explaining interpersonal aggres-
The cross-sectional design makes causal infer- sive behavior. Research on workplace aggression
ences impossible. Future research should examine should acknowledge the importance of individual
how these dynamic effects unfold over time via differences and social and group factors on individ-
longitudinal designs. Many of the effects are by ual behavior.
their nature dynamic and require longitudinal
methods to assess the operation of the role model- REFERENCES
ing, reciprocity, and attraction-selection-attrition
processes. Andersson, L. M., & Pearson, C. M. 1999. Tit for tat? The
These data are not immune to the possibility of spiraling effect of incivility in the workplace. Acad-
emy of Management Review, 24: 452-471.
Aquino, K., Tripp, T. M., & Bies, R. J. 2001. How employ-
common method bias-always a concern when

ees respond to personal offense: The effects of blame

single-source, self-report data are used. However, a

attribution, victim status, and offender status on re-

key predictor of work group aggression did not
venge and reconciliation in the workplace. Journal
include information from the focal individual. In
addition, the differential relations are counter to
of Applied Psychology, 86: 52-59.
what might be generated by common method bias.
Further, many of the relationships are consistent Bandura, A. 1977. Social learning theory. Englewood
Cliffs, NJ: Prentice-Hall.
Barling, J. 1996. The prediction, experience, and conse-
with prior empirical and theoretical work on

quences of workplace violence. In E. Q. Bulatao & G. R.

aggression.
Exploration of work group effects is limited by
the accuracy with which one can define work VandenBos (Eds.), Violence on the job: Identifying
groups. In this study, groups were defined by the risks and developing solutions: 29-49. Washington,
homes in which employees worked, making them DC: American Psychological Association.
better defined than groups in organizations in Baron, R. A., & Neuman, J. H. 1996. Workplace violence


2003 Glomb and Liao 495

and workplace aggression: Evidence on their relative ents and consequences. Organizational Behavior
frequency and potential causes. Aggressive Behav- and Human Decision Processes, 71: 309-328.
ior, 22: 161-173.
Greenberg, L., & Barling, J. 1999. Predicting employee
Baron, R. A., & Neuman, J. H. 1998. Workplace aggres- aggression against coworkers, subordinates and su-
sion-The iceberg beneath the tip of workplace vio- pervisors: The roles of person behaviors and per-
lence: Evidence on its forms, frequency, and targets. ceived workplace factors. Journal of Organizational
Public Administration Quarterly, 21: 446-464. Behavior, 20: 897-913.

Baron, R. A., Neuman, J. H., & Geddes, D. 1999. Social Hackman, J. R. 1992. Group influences on individuals in
and personal determinants of workplace aggression: organizations. In M. D. Dunnette & L. M. Hough (Eds.),
Evidence of the impact of perceived injustice and the Handbook of industrial and organizational psychol-
Type A behavior pattern. Aggressive Behavior, 25: ogy. Palo Alto, CA: Consulting Psychologists Press.
281-296. James, L. R., Demaree, R. G., & Wolf, G. 1993. R wg : an
Berkowitz, L. 1993. Aggression: Its causes, conse- assessment of within-group interrater agreement.
Journal of Applied Psychology, 78: 306-309.
quences, and control. Philadelphia: Temple Univer-
sity Press. Johnston, J., & DiNardo, J. 1997. Econometric methods.
New York: McGraw-Hill.
Bernaards, C. A., & Sijtsma, K. 2000. Influence of impu-
tation and EM methods on factor analysis when item Keashly, L., & Jagatic, K. 2003. By any other name: Amer-
nonresponse in questionnaire data is nonignorable. ican perspectives on workplace bullying. In S. Ein-
Multivariate Behavioral Research, 35: 321-364. arsen, H. Hoel, D. Zapf, & C. Cooper (Eds.), Bullying
and emotional abuse in the workplace: Interna-
Bies, R. J., Tripp, T. M., & Kramer, R. M. 1997. At the
tional research and practice perspectives: 31-61.
breaking point: Cognitive and social dynamics of
London: Taylor & Francis.
revenge in organizations. In R. A. Giacalone & J.
Greenberg (Eds.), Antisocial behavior in organiza- Matteson, M. T., & Ivancevich, J. M. 1982. Managing job
tions: 18-36. Thousand Oaks, CA: Sage. stress and health. New York: Macmillan.

Blau, P. M. 1964. Exchange and power in social life. Moorman, R. H. 1991. Relationship between organizational
New York: Wiley. j ustice and organizational citizenship behaviors: Do
fairness perceptions influence employee citizenship?
Buss, A. H. 1961. The psychology of aggression. New
Journal of Applied Psychology, 76: 845-855.
York: Wiley.
Neuman, J. H., & Baron, R. A. 1997. Aggression in the
Chen, P. Y., & Spector, P. E. 1992. Relationships of work workplace. In R. A. Giacalone & J. Greenberg (Eds.),
stressors with aggression, withdrawal, theft and sub- Antisocial behavior in organizations: 37-67. Thou-
stance use: An exploratory study. Journal of Occupa- sand Oaks, CA: Sage.
tional & Organizational Psychology, 65: 177-184.
Neuman, J. H., & Baron, R. A. 1998. Workplace violence
Folger, R., & Baron, R. A. 1996. Violence and hostility at and workplace aggression: Evidence concerning spe-
work: A model of reactions to perceived injustice. In cific forms, potential causes, and preferred targets.
E. Q. Bulatao & G. R. VandenBos (Eds.), Violence on Journal of Management, 24: 391-419.
the job: Identifying risks and developing solutions:
Northwestern National Life Employee Benefits Division.
51-85. Washington, DC: American Psychological
1993. Fear and violence in the workplace: A sur-
Association.
vey documenting the experience of American
Folger, R., Robinson, S. L., Dietz, J., McLean Parks, J., & workers. Minneapolis: Northwestern National Life
Baron, R. A. 1998. When colleagues become violent: Employee Benefits Division.
Employee threats and assaults as a function of soci- O'Leary-Kelly, A. M., Griffin, R. W., & Glew, D. J. 1996.
etal violence and organizational justice. Academy Organization-motivated aggression: A research
of Management Best Papers Proceedings: OB, framework. Academy of Management Review, 21:
A1-A7. 225-253.
Glomb, T. M. 2002. Workplace anger and aggression: Robinson, S. L., & Bennett, R. J. 1995. A typology of
Informing conceptual models with data from spe- deviant workplace behaviors: A multidimensional
cific encounters. Journal of Occupational Health scaling study. Academy of Management Journal,
Psychology, 7: 20-36. 38:555-572.
Glomb, T. M. 1998. Workplace aggression: Anteced- Robinson, S. L., & O'Leary-Kelly, A. M. 1998. Monkey
ents, Behavioral components, and consequences. see, monkey do: The influence of work groups on the
Unpublished doctoral dissertation, University of Il- antisocial behavior of employees. Academy of Man-
linois at Urbana-Champaign. agement Journal, 41: 658-672.
Glomb, T. M., Richman, W. L., Hulin, C. L., Drasgow, F., Rogers, W. H. 1993. sg17: Regression standard errors in
Schneider, K. T., & Fitzgerald, L. F. 1997. Ambient clustered samples. Stata Technical Bulletin, 13:
sexual harassment: An integrated model of anteced- 19-23.


49 6 Academy of Management Journal August

Roznowski, M. 1989. An examination of the measure- Watson, D., Clark, L. A., & Tellegen, A. 1988. Develop-
ment properties of the job Descriptive Index with ment and validation of brief measures of positive
experimental items. Journal of Applied Psychology, and negative affect: The PANAS scales. Journal of
74: 805-814. Personality and Social Psychology, 54: 1063-1070.
Salancik, G. J., & Pfeffer, J. 1978. A social information
processing approach to job attitudes and task design.
Administrative Science Quarterly, 23: 224-253.
Schneider, B. 1975. Organizational climates: An essay.
Personnel Psychology, 40: 437-454.
Theresa M. Glomb ([email protected] ) is an assis-

Skarlicki, D. P., & Folger, R. 1997. Retaliation in the

tant professor in the Department of Human Resources

workplace: The roles of distributive, procedural, and

and Industrial Relations at the University of Minnesota.

interactional justice. Journal of Applied Psychol-

She received her Ph.D. in industrial/organizational psy-
chology from the University of Illinois in 1998. Her
ogy, 82: 434-443. current research interests are in the areas of anger
Smith, P. C., Kendall, L., & Hulin, C. L. 1969. The mea- and aggressive behaviors in organizations, emotional
surement of satisfaction in work and retirement. expression in organizations, sexual harassment, and job
Chicago: Rand McNally. attitudes and behaviors.
Snyder, M. 1974. Self-monitoring of expressive behavior. Hui Liao is an assistant professor at the School of Man-
Journal of Personality and Social Psychology, 30: agement and Labor Relations at Rutgers University. She
526-537. was an assistant professor at the Institute of Labor and
Snyder, M., & Gangestad, S. 1986. On the nature of self-
Industrial Relations at the University of Illinois at Ur-
monitoring: Matters of assessment, matters of valid-
bana-Champaign when this study was completed. She
ity. Journal of Personality and Social Psychology,
received her Ph.D. in human resources and industrial
51:125-139.
relations from the University of Minnesota. Her current
research interests include service performance, em-
Spielberger, C. D. 1991. State-Trait Anger Expression ployee behavior at work, workplace diversity, work
Inventory (professional manual). Odessa, FL: Psy- group dynamics, and organizational restructuring.
chological Assessment Resources.
Tepper, B. J. 2000. Consequences of abusive supervision.
Academy of Management journal, 43:178-190.

View publication stats