娀 Academy of Management Journal

2007, Vol. 50, No. 6, 1385–1399.

BUILDING THEORETICAL AND EMPIRICAL BRIDGES ACROSS

LEVELS: MULTILEVEL RESEARCH IN MANAGEMENT
MICHAEL A. HITT
Texas A&M University

PAUL W. BEAMISH
University of Western Ontario

SUSAN E. JACKSON
Rutgers University and GSBA Zurich

JOHN E. MATHIEU
University of Connecticut

Most management problems involve multilevel phenomena, yet most management

research uses a single level of analysis. A micro or a macro lens alone yields incom-
plete understanding at either level. Multilevel research addresses the levels of theory,
measurement, and analysis required to fully examine research questions. This forum
presents multilevel research on bribery, national identity, team boundary spanning,
professional role identity, organizational citizenship, interorganizational exchanges,
and divestitures. To enrich the impact of future management research, we recommend
(1) applying multilevel designs to existing models (2) considering bottom-up effects, (3)
collaborating across disciplines on multidisciplinary topics, and (4) addressing major
real-world problems via multilevel approaches.

Most management research investigates phenom- incomplete understanding of behaviors occurring

ena by examining them at single levels of analysis at either level (Porter, 1996) and a proliferation of
(e.g., individual, group/team, organization, indus- diverse research paradigms. Paradigmatic diversity
try, country, geographic region). As the field of has some advantages, but the benefits of that diver-
management matures, however, researchers are de- sity are more likely to be realized through active
veloping more complex understandings of phe- paradigm competition and synthesis rather than
nomena by using multilevel lenses. Using a multi- parallel evolution and proliferation (cf. Pfeffer,
level lens reveals the richness of social behavior; it 1993; Rousseau, 2000). Multilevel research is one
draws our attention to the context in which behav- way to promote the development of a more expan-
ior occurs and illuminates the multiple conse- sive management paradigm for understanding or-
quences of behavior traversing levels of social or- ganizational systems.
ganization. For management to continue advancing Consider, for example, how multilevel research
as a field in which scholars seek to explain the might improve understanding of organizational
behaviors of individuals, groups, and organiza- performance. The macro approach adopted by stra-
tions, we must expand our theories and empirical tegic management scholars has provided a good
investigations to encompass these multilevel understanding of the links between the strategies
effects. employed by firms and their environments (e.g.,
Management scholars have a long history of rec- particular market characteristics). Yet we still do
ognizing that organizational phenomena unfold not fully appreciate the ways in which strategies
within complex and dynamic systems (e.g., Katz & are formulated inside organizations and, espe-
Kahn, 1978; Scott, 1974), yet our scholarship often cially, how they are implemented. In fact, some
ignores the multilevel dynamics of these social sys- scholars have argued that a poor understanding of
tems (Kozlowski & Klein, 2000). Rather than deal strategy implementation is one reason for divergent
with the complexities of multilevel systems, man- findings in the area of strategic management (Hos-
agement scholarship has become bifurcated into kisson & Hitt, 1990). For example, research has
camps of micro and macro experts who rarely en- shown that the performance effects of adopting a
gage each other in debates or collaboration. Adopt- related diversification strategy are inconsistent.
ing either a micro or a macro stance yields an Differences in the environments of organizations
1385
Copyright of the Academy of Management, all rights reserved. Contents may not be copied, emailed, posted to a listserv, or otherwise transmitted without the copyright holder’s express
written permission. Users may print, download or email articles for individual use only.
1386 Academy of Management Journal December

pursuing such a strategy are one possible explana- THE EVOLUTION OF MULTILEVEL RESEARCH
tion for the inconsistent findings, and looking for
With the benefit of hindsight, it is clear that the
such environmental explanations is a natural path
field of management began a slow evolutionary
for macro scholars to follow. It is less common for
change almost three decades ago. Until then, it was
those with expertise in the field of strategy to in-
commonplace for research dealing with individual-
vestigate the behavioral dynamics of strategy im-
level phenomena to be considered “micro,” while
plementation processes, yet such differences in
research that dealt with organizations was consid-
strategy implementation are likely to account for
ered “macro.” Usually, this distinction was quite
much of the observed variation in the effectiveness
easily made. Some work also addressed phenom-
of related diversification strategies (e.g., see ena at the level of groups and organizational sub-
Hoskisson & Hitt, 1990). Developing a more com- units, but it was less prevalent and had no distinct
plete understanding of strategy implementation is label. More importantly, few scholars made an ef-
likely to require research that recognizes the impor- fort to integrate theory or conduct research that
tance of leadership behaviors, relationships be- crossed these levels of inquiry or analysis. In effect,
tween managers and employees, performance mea- two or three lines of inquiry were progressing
surement and monitoring, incentive programs, and simultaneously.
so on. In other words, the relationship between The disciplinary heritages of scholars working in
firm strategies and firm performance is too complex these areas reinforced their differences. The micro
to be explained in only macro terms— understand- approach was rooted in psychology and focused on
ing it is likely to require the integration of insights understanding the thoughts, feelings, and actions of
from macro and micro scholars collaborating on individuals. The macro approach was rooted in soci-
joint research projects. ology and economics; it focused on understanding
Conversely, research conducted at the micro organizations and markets. The middle ground was
level too often ignores social dynamics arising at rooted in social psychology and closely related fields
higher levels of analysis. Research on discrimina- such as communications. In short, the field of man-
tion and diversity illustrates this point particularly agement was—and to a large extent, still is—fractured
well. Much of the work in this area has developed into specialized subfields, as evidenced by subsec-
out of an interest in understanding and, hopefully, tions of textbooks, specialized “niche” journals, and a
reducing the negative consequences of the preju- proliferation of divisions and interest groups in the
dices and biases that pervade daily interactions. Academy of Management.
Research on the interpersonal dynamics of preju- Discontent within the field of management began
dice and bias is firmly grounded in psychological to build (cf. Pfeffer, 1993) as the breadth of accept-
theories, which are used to explain the cognitive able theoretical approaches, substantive variables,
and emotional processes associated with stereotyp- research designs, and statistical analyses narrowed
ing and intergroup relations (e.g., see Jackson, May, within each of these domains. There was little
& Whitney, 1995). However, the social composition cross-pollination across specialties. The bound-
of organizations and work groups, as well as the aries and limitations of specialized viewpoints re-
larger cultural context, appear to influence these garding complex organizational phenomena be-
individual-level processes. Thus, improving the came increasingly evident. Anomalous facts that
sense of fairness perceived by individuals in a did not fit accepted subdiscipline doctrines began
workplace requires improving understanding of the to accumulate but were often dismissed as the by-
context within which employees interact on a one- products of flawed research.
to-one basis (e.g., see Jackson, Joshi, & Erhardt, However, around 1980, the first serious attempts
2003; Triandis, 1995; Tsui & Gutek, 2000). to break the paradigm of specialization and to forge
This Special Research Forum on Building integration emerged. The field appeared ready for a
Bridges across Levels was organized to encourage, paradigm shift. In an innovative yet often over-
promote, and support high-quality multilevel re- looked book, Developing an Interdisciplinary Sci-
search. We begin by providing a brief summary of ence of Organizations, Roberts, Hulin, and Rous-
the evolution of multilevel research. We then de- seau (1978) called for the integration of the
scribe the research published in this SRF. Finally, different disciplines that studied organizations. In-
we look to the future and provide a few suggestions deed, 30 years later, it is precisely this spirit that
that we hope will encourage readers to embark on motivated the framework for this special research
multilevel investigations that yield major advances forum of the Academy of Management Journal.
in our knowledge and contribute to improvements Roberts et al. (1978) argued that the organiza-
in management practice. tional outcomes of interest to management scholars
2007 Hitt, Beamish, Jackson, and Mathieu 1387

could be better understood if they broke out of their FIGURE 1

limited disciplinary, level-specific mind-sets. The Multilevel Nesting Arrangement
authors chronicled earlier work showing that mac-
roeconomic indicators could explain approxi-
mately 70 percent of organizational or country-
wide employee turnover rates. Other research
conducted during the same time period showed
that about 15–20 percent of the variance in individ-
ual-level turnover was predictable using individual
job attitudes. Roberts et al. (1978) concluded that
the different disciplines were trying to explain
largely the same criterion variance from different
perspectives. They asserted that the inherent pro-
cesses that predict criterion variance within each
level-specific perspective are probably related to
one another across levels and perspectives. They
also advanced a framework for integrating work
across disciplines, highlighted challenges associ-
ated with the aggregation of data, discussed how
the temporal nature of organizational phenomena
plays an important role in scholars’ thinking, and
outlined conditions for multidisciplinary theory.
Similar revelations have emerged from debates The overall logic is that individuals are nested in
and empirical investigations aimed at explaining work groups, which in turn are nested in larger
firm performance by examining industry-level fac- organizational units, such as departments or strate-
tors and the actions taken by a firm (see recent gic business units (SBUs), which are nested in na-
research by Bou and Satorra [2007], Hawawini, tional or multinational organizations. Furthermore,
Subramanian, and Verdin [2003], and Ruefli and organizations are nested in networks of interorgan-
Wiggins [2003]). Industrial organization econo- izational relationships (e.g., strategic alliances),
mists (e.g., Porter, 1980) emphasized the over- which in turn are nested in overall performance
whelming importance of industry structure in environments. Although the exact number and
determining firm outcomes, and strategic manage- nature of layers are likely to vary from one investi-
ment scholars focused on the importance of firm gation to another, the nesting arrangement has cer-
resources and strategies applying those resources tain implications for organizational theory and
in achieving a competitive advantage (Barney, research.
1991; Hoskisson, Hitt, Ireland, & Harrison, 2008). Whenever research traverses levels of analysis, it
The debate continues, but many scholars now in- becomes more complex, and scholars must be vig-
tegrate the two perspectives, suggesting that indus- ilant about carefully articulating the theoretical
try structure and firm resources both play impor- bases of their work. Rousseau (1985) provided a
tant roles in determining the appropriate strategies useful framework to guide such efforts. She argued
for firms to employ. Thus, the determination of that scholars need to simultaneously consider the
firm strategy is a multilevel problem. levels of theory, measurement, and analysis for the
Both theoretical and analytical advances in mul- constructs included in their investigations. Level of
tilevel research were achieved in the 1980s. These theory refers to the focal level to which generaliza-
advances were built on historical foundations that tions are meant to apply. “Level of measurement
evolved over long periods of time in other disci- refers to the unit to which the data are directly
plines—most notably, sociology, education, and attached . . . [whereas] the level of analysis is the
psychology—that had confronted related chal- unit to which data are assigned for hypothesis test-
lenges. The assumption underlying the multilevel ing and statistical analysis” (Rousseau, 1985: 4).
approach is that many outcomes of interest are the Critically, these three facets must be aligned to
result of a confluence of influences emanating from minimize level-related confounds, or what are of-
different levels of analysis (House, Rousseau & ten referred to as “fallacies of the wrong level.”
Thomas-Hunt, 1995; Rousseau, 1985). Arregle, Hebert, and Beamish (2006) presented an
The central theme of multilevel thinking is that example of these problems and their consequences
organizational entities reside in nested arrange- in international business research, as well as a
ments. Figure 1 depicts an example of such nesting. model for how they could be managed.
1388 Academy of Management Journal December

We should note that a multilevel theory is not a employs a large number of part-time or temporary
substitute for a multilevel design. For example, re- workers, should they be considered organizational
searchers drawing on Hambrick and Mason’s members? If two organizations are bitter rivals in
(1984) upper echelons theory have argued that be- most markets but have recently launched a joint
havioral integration, and thereby organizational venture, where do they fall in the nested arrange-
performance, is driven by characteristics of a firm, ment shown in Figure 1?
its top management team (TMT), and its CEO. Subdisciplines have confronted the challenge of
Clearly, upper echelons theory is multilevel in na- defining focal units of theory and analysis. For
ture in that it incorporates features of individuals, example, Hackman submitted that work groups are
groups, and organizations. However, tests of this “intact social systems, complete with boundaries,
theory typically have been conducted at a single interdependence among members, and differenti-
level (Cannella & Holcomb, 2005). For example, ated member roles” (1990: 4). Hackman further
Simsek, Veiga, Lubatkin, and Dino (2005) tested specified that work groups (1) have one or more
propositions about behavioral integration derived shared purposes, (2) have tasks to perform, (3) op-
from upper echelons theory using a sample of sev- erate in an organizational context, and (4) have
eral hundred organizations, but their design used a consequential transactions with entities outside the
single-level model tested at the organizational level group boundary. Such definitions are models of
of analysis. Because there is only one CEO and one clarity and precision and thereby are both powerful
TMT per organization, they were unable to use the and helpful. But even with clear definitions to
traditional nesting arrangement requiring that mul- guide them, researchers are likely to experience
tiple lower-level entities be nested within an upper- some ambiguity when specifying focal units. For
level entity. Usually, testing a multilevel theory instance, how much agreement is needed among a
necessitates a multilevel design, but not always. group’s members for an observer to conclude they
Thus, although the focus on behavioral integration have a “shared” purpose? How stable must a
represented an advance beyond much upper eche- group’s membership be for it to be considered “in-
lons research, the contribution was constrained by tact?” How much interdependence, of what type,
the single-level approach that has been common in among what proportion of members, for what peri-
management research. ods of time, is sufficient to justify a group’s defini-
tion as a work group? In the end, the focal units of
a theory are often difficult to define and identify.
Level of Theory
Nevertheless, the adequacy of a multilevel theory
A key attribute of the level of theory is the notion rests squarely on how well this is done.
of focal unit. Focal units are entities about which After a focal unit for generalization is identified,
one wishes to make generalizations (e.g., individu- a multilevel theory can be developed. That is, pre-
als, groups, organizations, etc.). The distinction be- dictions can be made about how entities are related
tween individuals and collectives is relatively easy to each other, and through what processes. Figure 2
to make, but it can be more challenging to identify depicts a simplified two-level arrangement that
the precise boundary where one collective ends shows possible relationships among entities in a
and another begins (e.g., groups), as well as the multileveltheory.Therelationshipsinvolvinglower-
point at which one has moved beyond one level of level constructs are shown using lower-case letters
analysis (e.g., SBU) and into another (e.g., organi- (x 3 y). The relationships involving upper-level
zation). Such distinctions are even more difficult in variables are shown using upper-case letters (X
the age of team-based organizations, communities 3 Y). Figure 2 depicts in greater detail the relation-
of practice, networks, strategic alliances, virtual or- ships suggested by the nested rings shown in Fig-
ganizations, and multinational enterprises. ure 1. The two levels shown in Figure 2 could
Before examining relationships between entities, correspond to individuals in groups, SBUs, organ-
within or across levels of analysis, one must first be izations, or many other combinations of nested en-
able to determine that such entities exist. Research- tities. The prevailing logic in management research
ers often rely on formal designations, such as as- is that the larger context within which lower-level
signment to a work team, full-time employment in processes are nested generally exerts a stronger
an organization, SIC code, or country location. downward influence, and the lower-level variables
Such conventions are not always satisfactory, how- generally exert a weaker upward influence. How-
ever. The boundaries around a standing work team ever, a multilevel perspective does not preclude the
may be easy to specify, but defining the boundaries possibility of upward and reciprocal influences
of ad hoc task forces and communities of practice is (Griffin, 1997; Kozlowski & Klein, 2000). We return
much more difficult. Similarly, if an organization to this point later.
2007 Hitt, Beamish, Jackson, and Mathieu 1389

FIGURE 2
Two-Level Relationships

Level of Measurement characteristics, functional backgrounds, and so

forth as indicators of group-level diversity.
Level of measurement refers to the level of the
Compilation is an aggregation principle in which
entities from which data are derived. Whenever the
measures collected from lower-level entities are
level of measurement used in research differs from
combined in complex and nonlinear ways to yield
the level of analysis (discussed below), some justi-
a gestalt, or whole, that is not reducible to its con-
fication for aggregating lower-level data is needed.
stituent parts (Kozlowski & Klein, 2000). For exam-
For example, one might have individuals answer ple, Ellis, Bell, Ployhart, Hollenbeck, and Ilgen
survey items about their own work attitudes and (2005) advanced a model of team competencies in
how they behave in their group. In that case, the which the characteristics of some members (e.g.,
work attitudes would be naturally aligned with in- the most central person for information flow) con-
dividual constructs at the individual level of anal- tribute more to an overall index of team abilities
ysis. However, if one wished to use the individual- than do the characteristics of other (e.g., more pe-
level data to test hypotheses about group-level ripheral) members.
phenomena, the level of measurement (individual) In summary, when the level of measurement is
and the level of theory and analysis (group) would different from the level of analysis, researchers
be different. Before the individual-level data are must use an underlying theory to guide their ap-
used to measure group-level phenomena, psycho- proach for aggregating lower-level data to measure
metric evidence of the suitability of using the data higher-level constructs. The nature of the higher-
in this way should be demonstrated. Depending on level construct and the processes that researchers
the nature of the constructs in question, different believe relate the lower-level data to the higher-
types of validity evidence should be provided (see level construct will determine the type of evidence
Chan, 1998; Chen, Mathieu, & Bliese, 2004). More needed to justify the aggregation of scores.
generally, however, when conducting multilevel
investigations, researchers need to be explicit about
how data collected at one level of analysis are re-
lated to constructs at a higher level of analysis. Two Level of Analysis
aggregation principles for identifying the relation- Except under extremely limited circumstances,
ship between lower-level data and higher-level the level at which data are analyzed to test hypoth-
constructs are composition and compilation. eses should be aligned with the level of theory for
Composition refers to the use of simple descrip- the constructs involved. To the extent that the
tive statistics to represent the processes that asso- levels of theory and analysis are misaligned, mis-
ciate lower-level data with higher-level constructs. specifications of various forms will arise. As noted
For example, one might use the average of individ- earlier, Rousseau (1985) referred to such misspeci-
uals’ competencies to represent organizational hu- fications as “fallacies of the wrong level.”
man capital. Alternatively, researchers might use Historically, every researcher has faced a deci-
measures of the variance in members’ demographic sion as to which level of analysis to adopt. Choos-
1390 Academy of Management Journal December

ing a single level sometimes led to misalignment cant lower-level effects (cf. Cohen, 2005; Moerbeek,
between the levels of theory and analysis—partic- van Breukelen, & Berger, 2000).
ularly when predictors were hypothesized to reside In multilevel studies, the scaling of measures
at different levels than focal units. Fortunately, also has important implications for the inferences
more sophisticated methods for analyzing multi- that are drawn from the data. The decision to center
level data have been developed, such as within and lower-level data within each higher-level unit or
between analysis (WABA; see Dansereau & Yam- around the overall mean affects the interpretation
marino, 2000), and cross-level operator techniques of the RCM intercept and slope parameters (Hof-
(CLOP; see James & Williams, 2000). Although mann & Gavin, 1998). In short, centering decisions
these techniques have bolstered multilevel re- are not simply statistical choices made in a vac-
search throughout the past two decades, recent de- uum; they are predicated on underlying theory
velopments have proven to be more flexible and about the multilevel relationships.
appropriate for multilevel data arrangements. Most Recent work has extended RCM cross-level anal-
notably, random coefficients modeling (RCM) has ysis to include the role of mediators. Both concep-
seemingly become the analysis of choice, as evi- tual and analytic challenges emerge when intro-
denced, at least in part, by the submissions that we ducing mediators that reside at the higher versus
received for this issue. Hofmann (1997) provided a the lower level of analysis, or operate from both
cogent introduction of the technique to organiza- levels (cf. Bauer, Preacher, & Gil, 2006; Mathieu &
tional researchers, and its popularity is quite evi- Taylor, 2007). Finally, we should note that multi-
dent today. However, RCM is not a panacea, nor level versions of structural equation modeling tech-
necessarily appropriate, for all multilevel niques (ML-SEM) have also recently been intro-
investigations. duced (Muthén & Muthén, 2007). Like their single-
RCM can be performed using a wide variety of level counterparts, ML-SEM techniques permit
software packages, the most popular of which researchers to consider the influence of measure-
seems to be hierarchical linear modeling (HLM; ment errors and to assess the relative fits of entire
Raudenbush, Bryk, Cheong, Congdon, & du Toit, structural models. The development of these so-
2004). Easy-to-use software has enabled more wide- phisticated ML-SEM techniques is encouraging, yet
spread use of RCM, but there remain a number of to date their application has been limited to fairly
methodological, design, and analytical issues to be simplistic models that include only a few variables.
resolved. For example, questions about how many Additional work is needed before they can accom-
lower- and upper-level entities should be sampled modate the complexity of modern-day management
in cross-level investigations provide both scientific theories. Nevertheless, they hold great promise for
and practical challenges. In multilevel designs, sta- the future. Moreover, given the ambiguities inher-
tistical power is a complex combination of the ent in multilevel research, the use of qualitative
number of higher-level units and lower-level units methodologies appears to be particularly suitable.
under investigation. Although some software pro- The field of management has come a long way
grams allow researchers to make a priori estimates during the past 30 years. Current research shows
of the power of multilevel designs, these applica- much greater sensitivity to multilevel issues, in-
tions are far more complex than their single-level cluding those associated with theoretical develop-
counterparts (cf. Raudenbush, 1997; Snijders & ment, approaches to measurement, and analytical
Bosker, 1993). techniques, yet many opportunities for further de-
The heterogeneity of multilevel sampling frames velopment remain. The research presented in this
also influences power estimates and the inferences special research forum exemplifies the diversity
that can be drawn from multilevel studies. Con- and richness of multilevel management research.
sider an instance in which one collects work atti- Next, we concisely describe the studies in this SRF.
tude data from a sample of engineers drawn from a
wide variety of organizations. Contrast that sam-
SUMMARY OF ARTICLES IN THE SPECIAL
pling frame to collection of the same work attitude
RESEARCH FORUM
data from a wide variety of employees who work
for subsidiaries of one large corporation. In the first As the foregoing discussion makes clear, multi-
case, the sampling frame is likely to result in more level scholarship is already well underway in the
variance at the higher level— e.g., the organization- field of management. This SRF is intended to move
al level—which substantially favors finding signif- the field forward a bit more rapidly than might
icant higher-level effects. In the second case, the otherwise occur by encouraging such work and
relatively high amount of heterogeneity at the publishing several examples that illustrate the
lower (the individual) level favors finding signifi- many forms such research can take.
2007 Hitt, Beamish, Jackson, and Mathieu 1391

In response to the call for papers, we received 99 enced the reconstruction of professional role
manuscripts for possible publication in this Special identities.
Research Forum on Building Bridges across Levels. In “A Multilevel Investigation of Antecedents
The articles selected for publication survived an and Consequences of Team Member Boundary-
intensive review and development process, which Spanning Behavior,” Jennifer Marrone, Paul
screened out a large number of papers. The full set Tesluk, and Jay Carson examine boundary span-
of articles we reviewed included many highly in- ning in the context of 31 MBA student teams work-
teresting research ideas and studies. Surprisingly, ing on consulting projects for periods of about a
however, several of the manuscripts submitted did month. In the settings studied, greater levels of
not fully represent multilevel empirical research. team boundary spanning were associated with
Some of those manuscripts were of sufficiently higher team performance. The multilevel analysis
high quality that they have been accepted and pub- showed that team-level boundary spanning also
lished in regular issues of the Academy of Manage- yielded benefits to individual team members by
ment Journal. We are pleased to present the follow- reducing their role overload. Finally, Marrone et al.
ing papers in this special research forum. Table 1 (2007) demonstrated that team-level boundary
provides a finer-grained description of each study’s spanning could be predicted by examining both
characteristics. The table entries briefly indicate individual and team-level factors.
the research focus, the levels of analysis examined, In “Disentangling the Influences of Leaders’ Re-
and the methodological approach used. lational Embeddedness on Interorganizational Ex-
In “Does Prevalence Mitigate Relevance? The change,” Jeffrey Barden and Will Mitchell examine
Moderating Effect of Group-Level OCB on Em- the roles of the prior exchange experiences of or-
ployee Performance,” authors William Bommer, ganizations and individual leaders in subsequent
Erich Dierdorff, and Robert Rubin examine the interorganizational exchanges. They focused on
multilevel moderating effects of group-level organ- mutual understanding, trust, and commitment to
izational citizenship behavior (OCB). Past research develop a microlevel model of organization-level
has established a clear relationship between indi- relationships. Their study of player trades in Major
vidual-level OCB and individual performance. For League Baseball showed that interorganizational
this study, the Bommer et al. (2007) reasoned that exchanges were influenced more by an organiza-
this individual-level effect should be stronger in tion’s set of ties than by individual leaders’ ties.
work teams for which OCB is distinctive and rela- Barden and Mitchell (2007) found that leaders’ in-
tively rare; in work teams with higher levels of OCB dividual influence on interorganizational ex-
occurring among all members, the relationship be- changes was largely based on the contexts of their
tween individual-level OCB and performance was organizations. As such, the leaders had little inde-
predicted to be weaker. In a study of 100 teams pendent influence; rather, the effects of that influ-
distributed across seven locations of a U.S. ma- ence became visible in the interaction with organ-
chined metals manufacturer, the authors found izational influences.
support for the predicted multilevel moderating In “Deciding to Bribe: A Cross-Level Analysis of
effect. Firm and Home Country Influences on Bribery Ac-
“Interlevel Influences on the Reconstruction of tivity,” Kelly Martin, John Cullen, Jean Johnson,
Professional Role Identity,” by Samia Chreim, and Praveen Parboteeah examine how organiza-
Bernie Williams, and Bob Hinings, describes the tion-level and contextual pressures combine to in-
interplay of institutional, organizational, and per- fluence firms’ supplying bribes to public officials.
sonal factors that together shaped the reconstruc- Using anomie theory as a theoretical foundation,
tion of the professional role identities of eight phy- the authors proposed that social institutions, na-
sicians in a medical setting. Using a longitudinal tional culture, and perceived financial and compet-
case research design, Chreim and her colleagues itive pressures all play important roles in determin-
(2007) gathered qualitative data by observing meet- ing the likelihood of bribery. Using archival data
ings, conducting multiple interviews, and compil- involving 3,769 firms and 38 institutional and cul-
ing written material. Their analysis tracked the tural contexts, Martin and her coauthors (2007)
changes in professional role identities that oc- found support for anomic conditions at both the
curred among the physicians as they went from firm and societal levels. The results indicate that
working as independent practitioners to working as the propensity to bribe is greater in achievement-
a multidisciplinary team of collaborators providing oriented, individualistic societies than in collectiv-
integrated medical services. The authors’ analysis ist societies. The results also indicate that firms in
revealed the dynamics through which institutional achievement-oriented environments, in which de-
structure and individual agency collectively influ- cision makers perceive high levels of competitive
 TABLE 1
Description of Studies in the Special Research Forum
Title/Authors Topic Levels of Analysis Literature Sample Methodology Analytical Technique

1 “A Multilevel Investigation Antecedents and Individual; group Boundary spanning; MBA students Three surveys Hierarchical linear
of Antecedents and consequences of cross-functional implemented at modeling
Consequences of Team boundary-span- teams three different
Member Boundary- ning activities points in time
Spanning Behavior,” by
Jennifer Ann Marrone,
Paul E. Tesluk, and Jay
B. Carson

2 “Deciding to Bribe: A Multiple forces Organization; country Institutional norms; Third-party surveys such Archival data Hierarchical linear
Cross-Level Analysis of affecting bribery anomie theory as the World Business modeling
Firm and Home Country Environment Survey,
Influences on Bribery the Global Leadership
Activity,” by Kelly D. and Organizational
Martin, John B. Cullen, Behavior Effectiveness
Jean L. Johnson, and K. Study, and the United
Praveen Parboteeah Nations Human
Development Report

3 “Does Prevalence Mitigate OCB as a Individual; group Organizational Works groups and Multilocation, Hierarchical linear
Relevance? The multilevel citizenship individual employees multilevel modeling
Moderating Effect of construct behavior surveys
Group-Level OCB on
Employee Performance,”
by William H. Bommer,
Erich C. Dierdorff, and
Robert S. Rubin

4 “Interlevel Influences on Role reconstruction Individual; organization; Role identity and A Canadian health clinic Single-case study Textual analysis
the Reconstruction of as a result of institutional professional (nonparticipant
Professional Role multilevel forces environment identity observation,
Identity,” by Samia interviews,
Chreim, Bernard E. archival data)
Williams, and C. R.
Hinings

5 “Prospect Theory, Forces affecting Individual; organization Prospect theory; Purposeful U.S.-based Archival data Event history analysis
Behavioral Theory, and risk-seeking behavioral firms that had divested
the Threat-Rigidity behavior of theory; the threat- a division identified
Thesis: Combinative organizations rigidity thesis within the database
Effects on Organiza- Mergers and Corporate
tional Decisions to Transactions
Divest Formerly
Acquired Units,” by
Katsuhiko Shimizu

6 “When Does National Existence of Individual; country; Integrative social Managers from the Experimental Test for equal means;
Identity Matter? hypernorms and global contracts theory United States and surveys repeated-measures
Convergence and universal ethics Russia analysis of
Divergence in covariance
International Business
Ethics,” by Wendy
Bailey and Andrew
Spicer

7 “Disentangling the Role of individual Individual; organization Social capital; Player trades between Archival data Survival analysis
Influences of Leaders’ leaders’ and organ- relational Major League Baseball (Cox regression)
Relational Embed- izational ties in embeddedness organizations
dedness on Interorganiza- interorganization-
tional Exchange,” by al exchange
Jeffrey Barden and Will
Mitchell
2007 Hitt, Beamish, Jackson, and Mathieu 1393

intensity, exhibit an even greater likelihood of en- ages are only estimates because one could use sev-
gaging in bribing behavior. eral different criteria for identifying articles as
“Prospect Theory, Behavioral Theory, and the multilevel. Regardless of their approximate nature,
Threat-Rigidity Thesis: Combinative Effects on Or- however, these percentages indicate that multilevel
ganizational Decisions to Divest Formerly Ac- considerations are more likely to be reflected in our
quired Units,” by Katsuhiko Shimizu, extends conceptual models than in our empirical research.
prospect theory, with its notion of risk seeking and The opportunities for new multilevel research
risk avoidance, to the organizational level of anal- are many, and they arise out of both theoretical and
ysis. In addition to examining the main relation- methodological considerations. For example, there
ship between unit performance and choice to di- are opportunities to examine whether particular
vest, Shimuzu (2007) used behavioral theory and phenomena appear at multiple levels of analysis.
the threat-rigidity thesis to develop a set of moder- One such research question is, Do patterns of con-
ating factors. Longitudinal archival data on 68 U.S. flict found within teams also appear in the patterns
firms that divested 68 units from 1988 through of interactions among rival firms? Research that
1998 were used in hypothesis testing. Event history incorporates more than two levels of analysis also
analysis showed that the main effect was indeed represents an opportunity for improving under-
nonlinear and that differing levels of ambiguity, standing of organizational life. For example, in a
lack of improvement in unit performance, resource study of gender-based pay differentials, Joshi, Liao,
availability, divestiture experience, and relative and Jackson (2006) demonstrated that pay differen-
unit size produced significant interaction effects. tials reflected the interplay of individual, team, and
In “When Does National Identity Matter? Conver- subunit influences. Certainly, research on the rela-
gence and Divergence in International Business tionship between lower-level phenomena, such as
Ethics,” Wendy Bailey and Andrew Spicer examine employee attitudes and behaviors, and organiza-
differences between and among members of na- tional phenomena, such as firm-level outcomes
tional groups. Focusing on convergence as well as (e.g., financial performance), is also needed (e.g.,
divergence of ethical beliefs, the authors built on Schneider, Hanges, Smith, & Salvaggio, 2003). Also
integrative social contracts theory to predict when potentially valuable is research that investigates
such beliefs would be similar rather than different. the role of interfirm relationships within networks
Using an experimental design, each with six survey or between alliance partners in the development of
scenarios, Bailey and Spicer (2007) compared the new products, market performance, and so on
conditional effects of hypernorms and local norms (Almeida & Phene, 2004; Yamin & Otto, 2004). We
on national and expatriate differences using a sam- also need more research to help us understand the
ple of managers from both Russia and the United differences in the effects of national cultures and
States. Statistical results showed the similarity in institutional arrangements on the dynamics of firm
ethical attitudes between the two country groups strategies and interfirm behaviors such as subsid-
when hypernorms are present in their situations, as iary and home office relationships (Gamble, 2003;
well as a similarity between Russian managers and Hitt, Dacin, Levitas, Arregle, & Borza, 2000;
U.S. expatriates working within Russia, thereby Makino, Isobe, & Chan, 2004).
lending support to the hypothesized notion of Undoubtedly, a variety of methodologies will
convergence. prove useful in the conduct of such research, in-
cluding triangulation approaches. Intensive case
studies and other forms of qualitative research may
THE REST OF THE STORY: OPPORTUNITIES
be especially useful for understanding multilevel
AND VALUE IN MULTILEVEL RESEARCH
phenomena such as organizational changes and re-
It is apparent that multilevel thinking has been actions to crises (e.g., Dutton & Dukerich, 1991).
increasing in importance among management Similarly, a variety of theoretical perspectives will
scholars. A review of the articles published in the likely prove useful. New insights might be gained
Academy of Management Journal during a recent by innovative research that applies microlevel the-
12-month period (August 2006 –July 2007) revealed ories to phenomena usually examined at the macro
that approximately 25 percent of the studies level; likewise, applying macrolevel theories to mi-
adopted some type of multilevel perspective. By crolevel phenomena may yield new insights. For
comparison, an examination of articles published example, the resource-based view of the firm might
in the Academy of Management Review (AMR) dur- be applied to improve our understanding of some
ing the same period revealed that discussions of aspects of individual and team-level behaviors (see
multilevel phenomena were significant in approx- the work of Barney [1992]).
imately 50 percent of the articles. These percent- Going forward, we hope that this forum catalyzes
1394 Academy of Management Journal December

an increase in multilevel empirical research to to develop a model for understanding the behavior
match the increase in multilevel conceptual work. of activist groups as they attempt to bring about
Toward that end, we offer the following sugges- institution-level change. In another example, Hoet-
tions for the future. ker and Agarwal (2007) investigated the effects on
knowledge diffusion within an industry of firm
exits from that industry. In the special research
Test Existing Models
forum articles presented in this issue of AMJ, the
Our first suggestion is quite simple: Conduct em- upward influences of individuals on groups, organ-
pirical studies to test the conceptual models that izations, and institutions were sometimes acknowl-
have already been developed. AMR’s Special Topic edged, but they were rarely the focus. Only Barden
Forum on Corporations as Social Change Agents and Mitchell focused on upward influences, by
illustrates how multilevel thinking is informing examining the effects of individual ties on organi-
new conceptual developments in the field of man- zational ties. They found that individual ties influ-
agement. In the introduction to that special topic enced future interorganizational exchanges but
forum (STF), the editors asserted that an under- only when integrated with organizations’ prior ties.
standing of corporate social change activities “in- As organizations continue to evolve in an envi-
volves examination of corporate social agency at ronment characterized by global networks of indi-
multiple levels of analysis: the micro level (focus- viduals and special interest groups, opportunities
ing on psychological and social psychological for studying bottom-up phenomena are likely to
bases), the meso level (involving relational and net- grow. Scholars’ understanding of organizations
work issues), and the macro level (involving polit- could be enriched immensely by multilevel studies
ical, economic, institutional and societal dynam- that investigate the forces of upward influence in
ics)” (Bies, Bartunek, Fort, & Zald, 2007: 789). The addition to the top-down forces that shape complex
multilevel and multidisciplinary character of social phenomena. In a discussion of this point, Hackman
change is reflected in both the STF as a whole and (2003) identified the following benefits of looking
in specific articles. Examining the gestalt, one finds both one level down and one level up for scholars
a collection of articles that address phenomena at interested in understanding groups or teams: an
different levels of analysis. For example, Brickson enriched understanding of a focal phenomenon,
(2007) presented a discussion of organizational the discovery of nonobvious forces that influence
identity orientations and their implications for var- the focal phenomenon, discovering interactions
ious groups of stakeholders, and King (2007) dis- that involve higher-level forces that shape the focal
cussed how transaction costs can influence a firm’s phenomenon, and developing better theories. We
relationships with stakeholders. Campbell (2007) believe that scholars examining phenomena at any
and Marquis, Glynn, and Davis (2007) described level of analysis could reap these same benefits by
how community-level institutional pressures can thinking carefully about how the phenomena both
shape corporate-level action. Aguilera, Rupp, Wil- influence and are influenced by forces operating at
liams, and Ganapathi (2007) considered four levels higher and lower levels of analysis.
of analysis in a discussion of the motives of indi-
vidual, organizational, national, and intergovern-
Collaborate across Disciplines
mental agents that pressure firms to engage in so-
cial change through socially responsible corporate Our third suggestion for readers interested in
behavior. Of course, AMR is not the only source of pursuing multilevel empirical work is to become
good ideas for multilevel research. Often, edited involved in multidisciplinary collaborations with
books that focus on single topics of interest inte- people interested in similar topics. As the field of
grate articles that address the same phenomena at management continues to grow, it becomes increas-
different levels of analysis. ingly important to consider and integrate the devel-
opments that are occurring outside of specialty
areas and in adjacent disciplines. However, reading
Look Up
all of the published research in one’s own specialty
Work that considers how phenomena at lower (within our discipline) with relevance to one’s own
levels of analysis influence higher-level phenom- work is a significant challenge. As a consequence,
ena is much less common than work that does the future excellent multilevel research is more likely
reverse. In the AMR forum on social change, Den to be conducted by multidiscipline teams of schol-
Hond and de Bakker (2007) illustrated this type of ars who are motivated to investigate complex or-
multilevel theorizing. They integrated the litera- ganizational phenomena.
tures on social movements and institutional change In the call for papers for this special research
2007 Hitt, Beamish, Jackson, and Mathieu 1395

forum, we invited articles that link management The Academy of Management’s recent discus-
research to other disciplines. The field of manage- sions with the U.S. Institute of Medicine (IOM)
ment is firmly rooted in several disciplines, includ- illustrates how engaging big real-world problems
ing economics, sociology, and psychology. For reveals the need for knowledge created through
many of us, doing research that encompasses at interdisciplinary and multilevel research (Acad-
least two of these disciplines is a substantial chal- emy of Management, 2007). Among its many activ-
lenge. Perhaps for these reasons, we received very ities, the IOM is centrally involved in ongoing dis-
few submissions that adopted a larger multidisci- cussions about how to improve the quality and
plinary perspective and found none to include in efficiency of health care in the United States. Un-
this SRF. As our review of the evolution of multi- fortunately, in developing its policy recommenda-
level work revealed, the management discipline tions for improving health care, the IOM has not
has evolved and benefited from developments in effectively used relevant and potentially helpful
other fields. We have been good at adapting ad- management research. To remedy this situation, a
vances from other fields (e.g., analytical tech- committee of AOM scholars prepared a report that
niques), but we have been slower to integrate works would “begin to demonstrate how the application
from other disciplines. Echoing calls from long ago of management research might inform and expand
for such integration (Roberts et al., 1978), we sug- the IOM’s work, particularly with regard to the
gest that our field could be enriched by building development of recommendations that may be suc-
such bridges. In fact, Agarwal and Hoetker (2007) cessfully implemented” (Academy of Management,
argued that the management field could be en- 2007). Members of the AOM reviewed several re-
hanced by integrating the theory and empirical cent IOM reports that outlined a vision for the U.S.
work of other, related social science disciplines health care system. The IOM reports reflected the
(e.g., economics, sociology). accumulated knowledge of economists and medi-
cal scholars but cited almost none of the manage-
ment literature.
Tackle Major Real-World Problems
Even without technical knowledge about the U.S.
Becoming involved in multidisciplinary collabo- health care system, we know that improving the
rations is an aspect of our fourth and final sugges- delivery of health care will require making changes
tion for the future—namely, tackle important, real- at multiple levels. At the individual level are pa-
world problems. In his 2006 Academy of tients and their family members; physicians,
Management presidential address, Tom Cummings nurses, and myriad other health professionals; em-
asserted that “the future vitality and success of our ployees of health insurance companies; human re-
profession depends on making sure our research- source management professionals responsible for
based knowledge is relevant and useful. This will administering health care benefits; and so on.
require the Academy of Management . . . to be far These individuals are embedded in various types of
more engaged with the real world than has tradi- organizations. Because employers are a major
tionally been the case” (Cummings, 2007: 355). To- source of health care insurance, employment set-
ward this end, Cummings encouraged the develop- tings are relevant. Of course, hospitals, private
ment of alliances between the Academy of medical practice groups, and insurance companies
Management and organizations such as the United are central players, as are a variety of professional
Nation’s Global Compact, the Aspen Institute, the associations, such as the American Medical Asso-
Business Roundtable, and the American Associa- ciation. For some aspects of medical care, funding
tion of Collegiate Schools of Business. He envi- from the federal government is administered
sioned management scholars working with such through the states, so state-level dynamics must be
organizations tackling problems such as interna- addressed also.
tional trade, bribery and corruption, emergency Improving health care requires intervening in a
preparedness, and global outsourcing. These are complex, multilevel system. To be effective, inter-
large, complex issues. If management scholars hope ventions must be designed to correctly anticipate
to provide useful, actionable knowledge to such responses by the many elements comprising the
organizations, we must think seriously about the system. Can management scholars accumulate the
questions we ask. Are we satisfied with helping to type of compelling empirical evidence necessary to
resolve only small pieces of problems, hoping to make accurate predictions about the consequences
publish the results in one or two AMJ articles? Or of specific changes in the U.S. health care system?
are we willing to commit to conducting much Can we confidently make recommendations about
larger and more complex research projects to ad- how to design financial incentives to improve
dress problems similar to those identified above? health outcomes and/or organizational efficiency?
1396 Academy of Management Journal December

Can we confidently predict how individual physi- international entry: Advantages of multilevel meth-
cians will react to new measurement and reporting ods. Management International Review, 5: 557–
requirements? A close reading of the Academy’s 618.
report for the IOM suggests that our conceptual Bailey, W., & Spicer, A. 2007. When does national iden-
models should be helpful to the IOM as it tackles tity matter?: Convergence and divergence in interna-
these problems, but our empirical base of knowl- tional business ethics. Academy of Management
edge at this stage is anemic. Members of the Acad- Journal, 50: 1462–1480.
emy could be ideal participants in future research Barden, J. Q., & Mitchell, W. 2007. Disentangling the
designed to evaluate the consequences of imple- influences of leaders’ relational embeddedness on
menting new management practices in the health interorganizational exchange. Academy of Manage-
care field. Out of such engagement with this very ment Journal, 50: 1440 –1461.
pressing real-world problem, management scholars Barney, J. 1991. Firm resources and sustained competi-
would almost certainly develop a deeper knowl- tive advantage. Journal of Management, 17: 99 –
edge base. That knowledge would reflect more of 120.
the multilevel complexities inherent in large real- Barney, J. B. 1992. Integrating organizational behavior
world problems, and it also would be informed by and strategy formulation research: A resource-based
the knowledge of collaborators from other disci- analysis. In P. Shrivastava, A. Huff, & J. Dutton
plines. We have both much to offer to such efforts, (Eds.), Advances in strategic management, vol. 8:
and much to learn from the exchanges. 39 – 61. Greenwich, CT: JAI Press.
The example of the AOM’s developing relation- Bauer, D. J., Preacher, K. J., & Gil, K. M. 2006. Conceptu-
ship with the IOM illustrates the importance of and alizing and testing random indirect effects and
the need for more multilevel research. Clearly, im- moderated mediation in multilevel models: New
proving the U.S. health care system is only one of procedures and recommendations. Psychological
many large and important problems of potential Methods, 11: 142–163.
interest to management scholars. Regardless of the Bies, R. J., Bartunek, J. M., Fort, T. L., & Zald, M. N. 2007.
particular problem one chooses, it seems likely that Introduction to special topic forum: Corporations as
collaborating with scholars from other disciplines social change agents: Individual, interpersonal, in-
to address major real-world problems will lead to stitutional, and environmental dynamics. Academy
more useful multilevel research. Furthermore, in- of Management Review, 32: 788 –793.
creasing multilevel research is critical to enrich Bliese, P. D. 2000. Within-group agreement, non-inde-
and enhance our understanding of organizational pendence, and reliability: Implications for data ag-
phenomena. Thus, we call for and expect multi- gregation and analyses. In K. J. Klein & S. W. J.
level research to be one of the major steps that will Kozlowski (Eds.), Multilevel theory, research, and
move the management field into prominence in the methods in organizations: Foundations, exten-
future. sions, and new directions: 349 –381. San Francisco:
Jossey-Bass.
Bommer, W. H., Dierdorff, E. C., & Rubin, R. S. 2007.
REFERENCES Does prevalence mitigate relevance? The moderating
Academy of Management. 2007. Applying organization- effect of group-level OCB on employee performance.
al research and theory to Institute of Medicine Academy of Management Journal, 50: 1481–1494.
(IOM) reports: A review of three IOM reports. Bou, J. C., & Satorra, A. 2007. The persistence of abnor-
Briarcliff Manor, NY: Academy of Management. mal returns at industry and firm levels: Evidence
Agarwal, R., & Hoetker, G. 2007. A Faustian bargain? The from Spain. Strategic Management Journal, 28:
growth of management and its relationship with re- 707–722.
lated disciplines. Academy of Management Jour- Brickson, S. L. 2007. Organizational identity orientation:
nal, 50: 1304 –1322. The genesis of the role of the firm and distinct forms
Aguilera, R. V., Rupp, D. E., Williams, C. A., & Ganapathi, of social value. Academy of Management Review,
J. 2007. Putting the S back in corporate social respon- 32: 864 – 888.
sibility: A multilevel theory of social change in or- Campbell, J. L. 2007. Why would corporations behave in
ganizations. Academy of Management Review, 32: socially responsible ways? An institutional theory of
836 – 863. corporate social responsibility. Academy of Man-
Almeida, P., & Phene, A. 2004. Subsidiaries and knowl- agement Review, 32: 946 –967.
edge creation: The influence of the MNC and host Cannella, A. A., & Holcomb, T. R. 2005. A multi-level
country on innovation. Strategic Management analysis of the upper-echelons model: Multi-level
Journal, 25: 847– 864. issues in strategy and methods. Research in Multi-
Arregle, J.-L., Hebert, L., & Beamish, P. W. 2006. Mode of Level Issues, 4: 197–237.
2007 Hitt, Beamish, Jackson, and Mathieu 1397

Chan, D. 1998. Functional relations among constructs in The organization as a reflection of its top managers.
the same content domain at different levels of anal- Academy of Management Review, 9: 193–206.
ysis: A typology of composition models. Journal of Hawawini, G., Subramanian, V., & Verdin, P. 2003. Is
Applied Psychology, 83: 234 –246. Performance driven by industry- or firm-specific fac-
Chen, G., Mathieu, J. E., & Bliese, P. D. 2004. A frame- tors? A new look at the evidence. Strategic Manage-
work for conducting multilevel construct validation. ment Journal, 24: 1–16.
In F. J. Dansereau & F. Yammarino (Eds.), Research Hitt, M. A., Dacin, M. T., Levitas, E., Arregle, J.-L., &
in multi-level issues: The many faces of multi-level Borza, A. 2000. Partner selection in emerging and
issues, vol. 3: 273–303. Oxford, U.K.: Elsevier Sci- developed market contexts: Resource-based and or-
ence. ganizational learning perspectives. Academy of
Chreim, S., Williams, B. E., & Hinings, C. R. 2007. Inter- Management Journal, 43: 449 – 467.
level influences on the reconstruction of profes- Hoetker, G., & Agarwal, R. 2007. Death hurts, but it isn’t
sional role identity. Academy of Management Jour- fatal: The postexit diffusion of knowledge created by
nal, 50: 1515–1539. innovative companies. Academy of Management
Cohen, M. P. 2005. Sample size considerations for mul- Journal, 50: 446 – 467.
tilevel surveys. International Statistical Review, Hofmann, D. A. 1997. An overview of the logic and
73(3): 279 –287. rationale of hierarchical linear models. Journal of
Cummings, T. G. 2007. 2006 Presidential address: Quest Management, 23: 723–744.
for an engaged academy. Academy of Management Hofmann, D. A., & Gavin, M. B. 1998. Centering decisions
Review, 32: 356 –360. in hierarchical linear models: Implications for re-
Dansereau, F., & Yammarino, F. J. 2000. Within and search in organizations. Journal of Management,
between analysis: The variant paradigm as an under- 24: 623– 641.
lying approach to theory building and testing. In K. J. Hoskisson, R. E., & Hitt, M. A. 1990. Antecedents and
Klein & S. W. J. Kozlowski (Eds.), Multilevel theory, performance outcomes of diversification: Review
research, and methods in organizations: Founda- and critique of theoretical perspectives. Journal of
tions, extensions, and new directions: 425– 466. San Management, 16: 461–509.
Francisco: Jossey-Bass. Hoskisson, R. E, Hitt, M. A., Ireland, R. D., & Harrison,
den Hond, F., & de Bakker, F. G. A. 2007. Ideologically J. R. 2008. Competing for advantage. Mason, OH:
motivated activism: How activist groups influence Thomson South-Western Publishing.
corporate social change activities. Academy of Man- House, R., Rousseau, D. M., & Thomas-Hunt, M. 1995.
agement Review, 32: 901–924. The meso paradigm: A framework for the integration
Dutton, J. E., & Dukerich, J. M. 1991. Keeping an eye on of micro and macro organizational behavior. In L. L.
the mirror: Image and identity in organizational ad- Cummings & B. M. Staw (Eds.), Research in organ-
aptation. Academy of Management Journal, 34: izational behavior, vol. 17: 71–114. Greenwich, CT:
517–554. JAI Press.

Ellis, A. P. J., Bell, B. S., Ployhart, R. E., Hollenbeck, J. R., Jackson, S. E., Joshi, A., & Erhardt, N. L. 2003. Recent
& Ilgen, D. R. 2005. An evaluation of generic team- research on team and organizational diversity:
work skills training with action teams: Effects on SWOT analysis and implications. Journal of Man-
cognitive and skill-based outcomes. Personnel Psy- agement, 29: 801– 830.
chology, 58: 641– 672. Jackson, S. E., May, K. E., & Whitney, K. 1995. Under-
Gamble, J. 2003. Transferring human resource practices standing the dynamics of diversity in decision mak-
from the United Kingdom to China: The limits and ing teams. In R. A. Guzzo & E. Salas (Eds.), Team
effectiveness and decision making in organiza-
potential convergence. International Journal of Hu-
tions: 204 –261. San Francisco: Jossey-Bass.
man Resource Management, 14: 369 –387.
James, L. R., & Williams, L. J. 2000. The cross-level op-
Griffin, M. A. 1997. Interaction between individuals and
erator in regression, ANCOVA, and contextual anal-
situations: Using HLM procedures to estimate recip-
ysis. In K. J. Klein & S. W. J. Kozlowski (Eds.), Mul-
rocal relationships. Journal of Management, 23:
tilevel theory, research, and methods in
759 –773.
organizations: Foundations, extensions, and new
Hackman, J. R. 1990. Groups that work (and those that directions: 382– 424. San Francisco: Jossey-Bass.
don’t). San Francisco: Jossey-Bass.
Joshi, A., Liao, H., & Jackson, S. E. 2006. Cross-level
Hackman, J. R. 2003. Learning more about crossing lev- effects of workplace diversity on sales performance
els: Evidence from airplanes, hospitals, and orches- and pay. Academy of Management Journal, 49:
tras. Journal of Organizational Behavior, 24: 905– 459 – 481.
922. Katz, D., & Kahn, R. L. 1978. The social psychology of
Hambrick, D. C., & Mason, P. A. 1984. Upper echelon: organizations (2nd ed.). New York: Wiley.
1398 Academy of Management Journal December

King. A. 2007. Cooperation between corporations and design for cluster randomized trials. Psychological
environmental groups: A transaction cost perspec- Bulletin, 2: 173–185.
tive. Academy of Management Review, 32: 889 – Raudenbush, S. W., Bryk, A. S., Cheong, Y. F., Congdon,
900. R. T., & du Toit, M. 2004. HLM6: Hierarchical lin-
Klein, K. J. 1987. Employee stock ownership and em- ear and nonlinear modeling. Lincolnwood, IL: Sci-
ployee attitudes: A test of three models. Journal of entific Software International.
Applied Psychology, 72: 319 –332. San Francisco: Roberts, K. H., Hulin, C. L. & Rousseau, D. M. 1978.
Jossey-Bass. Developing an interdisciplinary science of organ-
Kozlowski, S. W. J., & Klein, K. J. 2000. A multi-level izations. San Francisco: Jossey-Bass.
approach to theory and research in organizations: Rousseau, D. M. 1985. Issues of level in organizational
Contextual, temporal, and emergent processes. In research: Multi-level and cross-level perspectives. In
K. J. Klein, S. W. J. Kozlowski (Eds.), Multilevel L. L. Cummings & B. M. Staw (Eds.), Research in
theory, research, and methods in organizations: organizational behavior, vol. 7: 1–37. Greenwich,
Foundations, extensions, and new directions: 3–90. CT: JAI Press.
San Francisco: Jossey-Bass.
Rousseau, D. M. 2000. Multilevel competencies and
Makino, S., Isobe, T., & Chan, C. M. 2004. Does country missing linkages. In K. J. Klein & S. W. J. Kozlowski
matter? Strategic Management Journal, 25: 1027– (Eds.), Multilevel theory, research and methods in
1043. organizations: 572–582. San Francisco: Jossey-Bass.
Marquis, C., Glynn, M. A., & Davis, G. F. 2007. Commu- Ruefli, T. W., & Wiggins, R. 2003. Industry, corporate,
nity isomorphism and corporate social action. Acad- and segment effects and business performance: A
emy of Management Review, 32: 925–945. non-parametric approach. Strategic Management
Marrone, J. A., Tesluk, P. E., & Carson, J. B. 2007. A Journal, 24: 861– 879.
multi-level investigation of antecedents and conse- Schneider, B., Hanges, P. J., Smith, D. B., & Salvaggio,
quences of team member boundary spanning behav- A. N. 2003. Which comes first: Employee attitudes or
ior. Academy of Management Journal, 50: 1423– organizational financial and market performance?
1439. Journal of Applied Psychology, 88: 836 – 851.
Martin, K. D., Cullen, J. B., Johnson, J. L., & Parboteeah, Scott, W. G. 1974. Organization theory: A reassessment.
K. P. 2007. Academy of Management Journal, 50: Academy of Management Journal, 17: 242–254.
1401–1422.
Shimizu, K. 2007. Prospect theory, behavioral theory,
Mathieu, J. E., & Taylor, S. 2007. A framework for testing and threat-rigidity thesis: Combinative effects on or-
meso-mediational relationships in organizational be- ganizational decisions to divest formerly acquired
havior. Journal of Organizational Behavior, 28: units. Academy of Management Journal, 50: 1495–
141–172. 1514.
Moerbeek, M., van Breukelen, G. J. P., & Berger, M. P. Simsek, Z., Veiga, J. F., Lubatkin, M. H., & Dino, R. N.
2000. Design Issues for Experiments in Multilevel 2005. Modeling the multilevel determinants of top
Populations. Journal of Educational and Behav- management team behavioral integration. Academy
ioral Statistics, 25: 271–284. of Management Journal, 48: 69 – 84.
Muthén, L. K., & Muthén, B. O. 2007. Mplus user’s guide Snijders, T. A. B., & Bosker, R. J. 1993. Standard errors
(4th ed.). Los Angeles: Muthén & Muthén. and sample sizes for two-level research. Journal of
Pfeffer, J. 1993. Barriers to the advance of organizational Educational Statistics, 18: 237–259.
science: Paradigm development as a dependable Triandis, H. C. The importance of context in studies of
variable. Academy of Management Review, 18: diversity. In S. E. Jackson & M. N. Ruderman (Eds.),
599 – 620. Diversity in work teams: 225–234. Washington, DC:
Porter, L. W. 1996. Forty years of organization studies: American Psychological Association.
Reflections from a micro perspective. Administra- Tsui, A., & Gutek, B. 2000. Demographic differences in
tive Science Quarterly, 41: 262–269. organizations. New York: Lexington Books.
Porter, M. E. 1980. Competitive strategy. New York: Free Yamin, M., & Otto, J. 2004. Patterns of knowledge flows
Press. and MNE innovative performance. Journal of Inter-
Raudenbush, S. W. 1997. Statistical analysis and optimal national Management. 10: 239 –258.
2007 Hitt, Beamish, Jackson, and Mathieu 1399

Michael A. Hitt ([email protected]) is a Distin- Susan E. Jackson ([email protected]) is a pro-

guished Professor of Management and holds the Joe B. fessor of human resource management in the School of
Foster Chair in Business Leadership at Texas A&M Uni- Management and Labor Relations at Rutgers Universi-
versity. He received his Ph.D. from the University of ty–New Brunswick and a research fellow at GSBA-
Colorado. His research interests focus on managing firm Zurich. She received her Ph.D. from the University of
resources for competitive advantage; international strat- California, Berkeley. Her research interests include
egy, especially the effects of country institutional envi- workforce diversity, knowledge-intensive teamwork,
ronments on the strategic behaviors of firms; corporate and strategic human resource management.
governance; and strategic entrepreneurship.
John E. Mathieu ([email protected]) is a
professor and the Cizik Chair of Management at the Univer-
Paul W. Beamish ([email protected]) holds the sity of Connecticut. He received a Ph.D. in industrial/organ-
Canada Research Chair in International Business at the izational psychology from Old Dominion University. He is a
Ivey Business School, University of Western Ontario. He fellow in the Society of Industrial and Organizational Psy-
received his Ph.D. from the University of Western On- chology and the American Psychological Association. His
tario. His international strategy research interests focus current research interests include models of team and mul-
on equity joint ventures, Japanese foreign direct invest- titeam processes and cross-level models of organizational
ment, internationalization, and emerging markets. behavior.