Escolar Documentos
Profissional Documentos
Cultura Documentos
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at .
http://www.jstor.org/page/info/about/policies/terms.jsp
.
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of
content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms
of scholarship. For more information about JSTOR, please contact support@jstor.org.
Academy of Management is collaborating with JSTOR to digitize, preserve and extend access to The Academy
of Management Journal.
http://www.jstor.org
MELISSA E. GRAEBNER
University of Texas at Austin
The Academy ofManagement Journal has taken accounts, such as Weick's (1993) study of theMann
the lead among major journals in spotlighting alter Gulch fire, but they are more likely to be contem
native methods that take advantage of rich empiri porary descriptions of recent events, such as Gil
cal data. In a series of "From the Editor" commen bert's(2005) study of adaptation to discontinuous
taries, scholars cogently have explicated related environmental change by newspaper organizations.
topics such as research (Gephart, 2004), The central notion is to use cases as the basis from
qualitative
grounded theory building (Suddaby, 2006), the which to develop theory inductively. The theory is
value of richness (Weick, 2007) and the persuasive emergent in the sense that it is situated in and
power of the single case (Siggelkow, 2007). In this developed by recognizing patterns of relationships
commentary, we focus on the related research strat among constructs within and across cases and their
egy of theory building from cases, particularly mul underlying logical arguments.
cases.
tiple Central to building theory from case studies is
Scholars have used case studies to develop the
replication logic (Eisenhardt, 1989b). That is, each
ory about topics as diverse as group process (Ed case serves as a distinct experiment that stands on
mondson, B?hmer, & Pisano, 2001), internal organ its own as an analytic unit. Like a series of related
ization (Galunic & Eisenhardt, 2001; Gilbert, 2005), cases are discrete
laboratory experiments, multiple
and strategy (Mintzberg & Waters, 1982). Classic that serve as contrasts,
experiments replications,
scholars (Chandler, 1962; Whyte, 1941) as well as and extensions to the emerging theory (Yin, 1994).
the authors of highly regarded AMJ papers (Dutton But while isolate the phe
laboratory experiments
& Dukerich, 1991; Sutton & Raphaeli, 1988) have nomena from their context, case studies emphasize
used the method. Indeed, papers that build theory the rich, real-world context in which the phenom
from cases are often regarded as the "most interest ena occur. The theory-building process occurs via
ing" research (Bartunek, Rynes, & Ireland, 2006) recursive cycling among the case data, emerging
and are among the most highly cited pieces in AMJ and later, extant literature. Although some
theory,
(e.g., Eisenhardt, 1989a; Gersick, 1988), with im times seen as "subjective," well-done theory build
to their numbers. Not sur
pact disproportionate ing from cases is surprisingly "objective," because
prisingly then, thewinning authors (Ferlie, Fitzger its close to the data keeps researchers
adherence
ald, Wood, & Hawkins, 2005; Gilbert, 2005) of the "honest." The
data provide the discipline that
most recent AMJ Best Article Award relied on this mathematics does in formal analytic modeling.
method. A major reason for the popularity and relevance
Building theory from case studies is a research of theory building from case studies is that it is
strategy that involves using one or more cases to one of the best (if not the best) of the bridges from
create theoretical constructs, propositions and/or rich qualitative evidence tomainstream deductive
midrange theory from case-based, empirical evi research. Its emphasis on developing constructs,
dence (Eisenhardt, 1989b). Case studies are rich, measures, and testable theoretical propositions
empirical descriptions of particular instances of a makes inductive case research consistent with the
that are typically based on a variety of
phenomenon emphasis on testable theory within mainstream de
data sources (Yin, 1994). Cases can be historical
ductive research. In fact, inductive and deductive
logics are mirrors of one another, with inductive
theory building from cases producing new theory
We appreciate helpful comments from Diane Bailey,
Steve Barley, Chris Bingham, Jason Davis, Nathan Furr, from data and deductive theory testing completing
and Ben Hallen as well as the sponsorship of theNational the cycle by using data to test theory. Moreover,
Science Foundation IOC Award #0621777 and the Stan since it is a theory-building approach that is deeply
fordTechnology Ventures Program. embedded in rich empirical data, building theory
25
research areas particularly well (Edmondson & Mc ing is simply idiosyncratic to a single case or con
Manus, 2007). By contrast, the research strategy is sistently replicated by several cases (Eisenhardt,
ill-equipped to address the questions "how often," 1991). Multiple cases also create more robust the
and "how many," and questions about the relative ory because the propositions are more deeply
empirical importance of constructs. grounded in varied empirical evidence. Constructs
and relationships are more precisely delineated be
cause it is easier to determine accurate definitions
Theoretical Sampling of Cases
and appropriate levels of construct abstraction
Another frequent
challenge to theory building from multiple cases. For example, Brown and
from cases concerns
case selection. Some readers Eisenhardt (1997) found that, although some firms
make the faulty assumption that the cases should used alliances to experiment with the future, others
be representative of some population, as are data in used futurists and exploratory products. With mul
testing research. In other
large-scale hypothesis tiple cases, the authors set an appropriate level of
words, they ask, How can the theory generalize if abstraction (i.e., probes) that was more accurate
the cases aren't representative? than the individual instantiations (e.g., alliances,
A key response to this challenge is to clarify that cases also enable
exploratory products). Multiple
the purpose of the research is to develop theory, not broader exploration of research questions and the
to test it, and so theoretical (not random or strati
oretical elaboration. For example, Brown and
fied) sampling is appropriate. Theoretical sampling Eisenhardt (1998) added successful and unsuccess
simply means that cases are selected because they ful turnaround cases that enabled them to add fur
are suitable for illuminating and ex
particularly ther longitudinal elements to their theory. Because
tending relationships and logic among constructs.
case numbers are
as are not ran typically small, a few additional
Again, just laboratory experiments cases can
from a of significantly affect the quality of the
domly sampled population experiments,
but rather, chosen for the likelihood that they will emergent theory. For example, adding three cases
to a single-case study is modest in terms of num
offer theoretical insight, so too are cases sampled
bers, but offers four times the analytic power. Thus,
for theoretical reasons, such as revelation of an
theory building from multiple cases typically
unusual phenomenon, replication of findings from
yields more robust, generalizable, and testable the
other cases, contrary replication, elimination of al
ternative explanations, and elaboration of the emer ory than single-case research.
But although multiple cases are likely to result in
gent theory.
better theory, theoretical sampling ismore compli
Theoretical sampling of single cases is straight
cated. The choice is based less on the uniqueness of
forward. They are chosen because they are unusu
a given case, and more on the contribution to the
ally revelatory, extreme exemplars, or opportuni
access ory development within the set of cases. That is,
ties for unusual research (Yin, 1994). For
multiple cases are chosen for theoretical reasons
example, Weick (1993) used an extreme case of lost
such as replication, extension of theory, contrary
sensemaking in the wilderness fire-fighting disaster
at Mann Gulch; Galunic and Eisenhardt (1996, replication, and elimination of alternative explana
in an ex tions (Yin, 1994). For example, Graebner and Eisen
2001) examined organizational adaptation
hardt (2004) studied acquisition from the seller per
emplar firm that was the highest performing tech
in the world for several spective by examining three replicated cases in
nology-based corporation
and Dutton and Dukerich which the executives sold their companies, a con
decades; (1991) studied
in which executives could have
the New York Port Authority, where they had un trary replication
usual access re sold their companies but did not, and then further
through friends. Thus, single-case
search typically exploits to explore a cases in different industries that explored industry
opportunities
under rare or extreme level explanations. A particularly important theo
significant phenomenon
circumstances. retical sampling approach is "polar types," in
But while single-case studies can richly describe which a researcher
samples extreme (e.g., very high
the existence of a phenomenon and very low performing) cases in order to more
(Siggelkow, 2007),
multiple-case studies typically provide a stronger easily observe contrasting patterns in the data. Al
base for theory building (Yin, 1994). Again, to use though such an approach can surprise reviewers
the analogy of laboratory experiments, the theory is because the resulting theory is so consistently sup
better grounded, more accurate, and more general ported by the empirical evidence, this sampling
izaba (all else being equal) when it is based on leads to very clear pattern recognition of the central
multiple case experiments. Multiple cases enable constructs, and of the focal
relationships, logic
comparisons that clarify whether an emergent find phenomenon.
Dealing with Interview Data A more subtle challenge arises from the confu
sion between qualitative data and qualitative re
Case studies can accommodate
a rich variety of
search. Theory-building cases usually rely exten
data sources, including interviews, archival data,
and observations. For sively on qualitative data from interviews and other
survey data, ethnographies,
sources, such as observations, historical books, ar
example, Hargadon and Sutton (1997) combined
chives, and so forth. This research is often termed
observations of brainstorming sessions, interviews
"qualitative" simply because it relies significantly
with corporate actors, and ethnographies of two
on qualitative data. But qualitative research can
projects in their case study of routine innovation at
also refer to the use of qualitative data in research
Ideo. But as research incorporates more cases and
moves strategies other than organizing data into cases and
from everyday phenomena
away such as
work practices to intermittent and strategic phe using replication logic to build theory. For exam
ple, Elsbach and Kramer (2003) accumulated qual
nomena such as acquisitions and strategic decision
itative data on individual "pitches" in their study
making, interviews often become the primary data
of face-to-face interviews in Hollywood, but they
source. Interviews are a highly efficient way to
pooled their data rather than organize it into cases.
gather rich, empirical data, especially when the some
Adding to the confusion, scholars have a very
phenomenon of interest is highly episodic and in
specific definition of what constitutes "qualitative
frequent. But interviews also often provoke a
research" that goes well beyond the type of data.
"knee-jerk" reaction that the data are biased in
For example, Gephart (2004) described qualitative
which impression management and retrospective
research as "multimethod research that uses an in
sensemaking are deemed the prime culprits. The terpretive, naturalistic approach to its subject mat
prototypical reader asks, Is the theory Just retro ter (Denzin & Lincoln, 1994)" and "addresses ques
spective sensemaking by image-conscious tions about how social experience is created and
informants? given meaning" (Gephart, 2004: 454-455). Accord
The challenge of interview data is best mitigated to this research is
ing view, qualitative highly de
by data collection approaches that limit bias. A key the social construction of re
scriptive, emphasizes
is numerous and highly knowl
approach using ality, and focuses on revealing how extant theory
edgeable informants who view the focal phenom
operates in particular examples. This view is dif
ena from diverse perspectives. These informants ferent in terms of research activities, goals, and
can include organizational actors from different hier from the more objective and positiv
epistemology
archical levels, functional areas, groups, and geog ist stance of theory building from cases as well as
as well as actors from other relevant organ from other research strategies also termed "qualita
raphies,
izations and outside observers such as market tive." The key implication is that some readers will
analysts. It is unlikely that these varied informants confuse different kinds of research that seem simi
informants will engage in convergent retrospective lar because they use qualitative data, and these
sensemaking and/or impression management. For readers may be disappointed if the research does
example, in our study of acquisitions from the not then match their understanding of "qualitative
seller perspective, Graebner and Eisenhardt (2004) research."
we relied on interviews with executives from two A straightforward approach for coping with the
hierarchical levels at the selling firms, executives varied meanings of "qualitative research" is to
from two hierarchical at the buying firms,
levels avoid the term. Rather, clarify the research strategy
board members from both
the buying and selling and contrast itwith
other "qualitative"
being used,
firms, and investment bankers who provided back approaches withdiffering epistemological assump
ground information about M&A. tions. Specifically, when inducting theory from
Another key approach to mitigating bias is to cases, be explicit about the theory-building goal
combine retrospective and real-time cases (Leonard and to liberally use footnotes that sharpen the dis
cases rely on inter tinctions
Barton, 1990). Retrospective among the multiple meanings of qualita
views (and archival data) that build up the number tive research. The key here is to convey the theory
and depth of cases efficiently and so enable a re building strategy clearly while avoiding confusion,
searcher to cover more informants and include philosophical pitfalls, and unrealistic reader
more cases. Such interviews are particularly accu expectations.
rate when the focal events are recent. In contrast,
real-time cases employ longitudinal data collection
Presenting Empirical Evidence
of interviews and, often, observations, both of
which help to mitigate retrospective sensemaking A critical aspect of empirical research is present
and impression management. ing the evidence from which the theory of interest
was inducted. In large-scale deductive studies, idence are central to signaling the depth and detail
there is a widespread norm of presenting theory of empirical grounding. In other words, the use of
and then empirical evidence in compact numerical summary tables and aids that summarize the case
tables that summarize statistical analyses of large evidence complements the selective story descrip
amounts of data. But case data cannot be so tightly tions of the text and further emphasizes the rigor
summarized, because much of it consists of rich and depth of the empirical grounding of the theory.
qualitative detail. A separate table that summarizes the evidence for
In a single-case study, the challenge of presenting each theoretical construct is a particularly effective
rich qualitative data is readily addressed by simply way to present the case evidence. These "construct
presenting a relatively complete rendering of the tables" summarize the case evidence and indicate
story within the text. The story typically consists of how construct is "measured,"
the focal thus in
narrative that is interspersed with quotations from creasing the "testability" of the theory and creating
key informants and other supporting evidence. The a particularly strong bridge from the qualitative
story is then intertwined with the theory to dem evidence to theory-testing research. Graebner
onstrate the close connection between empirical (2004),Gilbert (2005), and Zott and Huy (2007) are
evidence and emergent theory. This intertwining excellent examples of blending construct tables
keeps both theory and evidence at the forefront of with selected text descriptions.
the paper. Gersick (1994), Hargadon and Douglas Summarizing case evidence within tables and
(2001), and Mintzberg and Waters (1982) are exem organizing the text around the theory can be, how
plars of this approach.1 ever, disappointing to readers who are expecting
But presenting a relatively complete and un the "richness" of detailed narratives from the em
broken narrative of each case is infeasible for pirical data. This is particularly likely among read
multiple-case research, particularly as the number ers whose research predilections favor description
of cases increases. If the researcher relates the nar over theory. So, although itmay seem trivial, it is
rative of each case, then the theory is lost and the usually helpful to remind reviewers that the objec
text balloons. So the challenge in multiple-case tive is theory development. More significantly, it is
research is to stay within spatial constraints while critical to invest in developing well-crafted tables,
also conveying both the emergent theory that is the appendixes, and visual aids to demonstrate the the
research objective and the rich empirical evidence ory's underlying empirical support and the antici
that supports the theory. Coping with the trade-off pated richness of the case data, and to tie those
between rich story and well-grounded theory is tables clearly to the text.
easier to do in a multicase book or a single-case
paper. But in journal articles, multicase researchers
face a particularly difficult trade-off between the
Writing the Emergent Theory
ory and empirical richness. It can be especially
challenging to satisfy readers who expect the exten The objective of building theory from cases is
sive narratives of single-case research. They ask, theory. But unlike in large-scale hypothesis
Where's the rich story? testing research, there is no
"sure-to-please"
stan
The best way to address this challenge of "better dard template for writing emergent theory in
stories vs. better theories" is to develop a theory in theory-building research. Since different readers
sections or by distinct propositions in such a way have their own preferences, they often ask, Why did
that each is supported by empirical evidence. Thus, you format the theory this way?
the overarching organizing frame of the paper is the A useful way to cope with this challenge is to
theory, and each part of the theory is demonstrated write the theory inmultiple ways. First, sketch the
by evidence from at least some of the cases. But emergent theory in the introduction. Then, in the
since it is generally not realistic to support every body of paper, write each proposition (implicitly or
theoretical proposition with every case within a explicitly stated), and link it to the supporting em
text itself, the use of extensive tables and other pirical evidence for each construct and for the pro
visual devices that summarize the related case ev posed relationship between the constructs. When
the research is well done, the propositions will be
consistent with most (or even all) of the cases be
1 cause the researcher has
An alternative approach is to present the story and effectively "pattern
then the theory. But this approach moves the theory off matched" between theory and data. It is also crucial
center stage and makes the empirical grounding of the to write the underlying theoretical arguments that
theory less apparent. Nonetheless, it is a reasonable and provide the logical link between the constructs
common
approach. within a proposition. These arguments can be
plaining thelogic) and/or from more detached ported with transparent description, particularly
a visual
logic. Finally, provide theory summary regarding how the theory was inducted from the
such as a "boxes and arrows" diagram or summary data of cross-case
(e.g., description comparison
table. Eisenhardt (1989a), Gilbert (2005), and The is to convey the rigor,
techniques). key here
Maurer and Ibers (2006) offer exemplars of themul and of the research
creativity, open-mindedness
tiple ways of writing theory within a single paper. processes while sidestepping confusion and philo
Using these multiple ways to present the theory is sophical pitfalls.
often a safe starting point for initial manuscript a surprising challenge can arise from
Finally,
submissions. are
readers who disappointed by parsimonious
A more subtle challenge arises from confusion
theory. Particularly when readers are more fa
about the meaning of "grounded theory build miliar with the idiosyncratic detail of some single
ing." For some scholars, grounded theory build case research, they may expect the complicated
ing simply means creating theory by observing
theory that can arise from such cases. Somewhat
patterns within systematically collected empiri
cal data. This view often includes some notion of surprisingly, single cases can enable the creation
of more complicated theories than multiple cases,
recursively iterating between (and thus con
because single-case researchers can fit their theory
stantly comparing) theory and data during anal
cases exactly to the many details of a particular case.
ysis, and theoretically sampling (as de
scribed earlier). As Langley this is a In contrast, multiple-case researchers retain only
(1999) noted,
held view of In the relationships that are replicated across most
widely grounded theory building.
or all of the cases. Since there are typically fewer
this view, the quality of the theory and the
of these relationships than there are details in a
strength of its empirical grounding are more cen
tral to research quality than the specifics of the richly observed single case, the resulting theory is
theory-building process. often more parsimonious (and also more robust
But for other scholars, grounded theory building and generalizable). A key approach to dealing with
has a more precise meaning that stems from the this challenge is to ensure that the theory fully
original focus of Glaser and Strauss (1967) on the exploits the available evidence in terms of pos
interpretation of meaning by social actors. For ex sible nuances and alternative interpretations. It
ample, Suddaby described grounded theory build also helps to remind readers that parsimony, ro
ing as "most suited to efforts to understand the and generalizability characterize
bustness, superior
process by which actors construct meaning out of
theory.
intersubjective experience" (Suddaby, 2006: 634).
Others go further to emphasize elaborate processes
(and terminology) for how researchers should
Conclusion
gather field data and discover theory using a hier
archical structure of categories (Corbin & Strauss,
Theory building from case studies is an increas
1990). Constant comparison and theoretical sam and relevant research strategy that
ingly popular
pling take on precise meanings: "constant compar
forms the basis of a disproportionately large num
ison" means simultaneous collection and analysis
ber of influential studies. But like the adherents of
of data, and "theoretical sampling" means that de
any research method, its adherents face some pre
cisions about which data to collect next are deter
dictable challenges, some ironi
of which
have,
mined by the theory in progress (Suddaby, 2006).
In this view, adherence to specific grounded theory cally, emerged precisely because research relying
on rich qualitative data is becoming more common.
building processes is important in judging research
The good news is that these often very legitimate
quality. But strict adherence can also result in the
challenges can be mitigated through precise lan
ory with limited generalizability (Langley, 1999)
and idiosyncratic path dependence on the particu guage and thoughtful research design: careful jus
lar empirical starting point. tification of theory building, theoretical sampling
As when of cases, interviews that limit informant bias, rich
coping with the multiple meanings
of "qualitative research," it is often helpful to deal presentation of evidence in tables and appendixes,
with the multiple meanings of "grounded and clear statement of theoretical arguments. The
theory
the term unless one is ac result is fresh theory that bridges well from rich
building" by avoiding
the Glaser and Strauss (1967) ap qualitative evidence to mainstream deductive re
tually using
proach. It is also helpful to preempt misunder search. This is the hallmark of building from case
standing by engaging in systematic data collection studies.
Bingham, C. B., & Eisenhardt, K. M. 2006. Unveiling the Gephart, R. P. 2004. Qualitative research and the Acad
creation and content of strategic processes: How and emy ofManagement Journal. Academy ofManage
ment 47: 454-462.
what firms learn from heterogeneous experience. Journal,
Proceedings of the Academy ofManagement. Gersick, C. J.G. 1988. Time and transition inwork teams.
Brown, S. L., & Eisenhardt, K. M. 1997. The art of contin Toward a new model of group development. Acad
uous 31: 9-41.
change: Linking complexity theory and time emy of Management Journal,
paced evolution in relentlessly shiftingorganizations. Gersick, C. J.G. 1994. Pacing strategic change. Academy
Administrative Science Quarterly, 42: 1-35. 9-45.
of Management Journal,
& Eisenhardt, on
Brown, S. L., K. M. 1998. Competing Gilbert, C. G. 2005. Unbundling the structure of inertia:
the edge: Strategy as structured chaos. Boston: Resource versus routine rigidity. Academy of Man
Harvard Business School Press. 48: 741-763.
agement Journal,
D.
Chandler, A. 1962. Strategy and structure. Cam
Glaser, B., & Strauss, A.
The discovery of 1967.
bridge, MA: MIT Press. grounded theory: Strategies in qualitative re
Corbin, J.,& Strauss, A. 1990. Grounded search. London: Wiedenfeld and Nicholson.
theory research:
canons
Procedures, and evaluative criteria. Qualita Graebner, M. E. 2004. Momentum and serendipity: How
tive Sociology, 13: 3-21.
acquired leaders create value in the integration of
Denzin, N. K., & Lincoln, Y. S. 1994. Introduction: Enter technology firms. Strategic Management Journal,
ing the field of qualitative research. In N. K. Denzin
25: 751-777.
& Y. W. Lincoln (Eds.), Handbook of qualitative M. E., & Eisenhardt, K. M. 2004. The seller's
Graebner,
research: 1-17. Thousand Oaks, CA: side of the story:Acquisition as courtship and gov
Sage.
themirror: The role of image and identity in organ ministrative Science Quarterly, 49: 366-403.
izational adaptation. Academy of Management Greenwood, R., & Suddaby, R. 2006. Institutional entre
34: 517-554.
Journal,
preneurship inmature fields: The Big Five account
Edmondson, A. C, B?hmer, R. M., & Pisano, G. P. 2001. ing firms. Academy ofManagement Journal, 49:
Disrupted routines: Team learning and new technol
27-48.
Eisenhardt, K. M. 1989a. Making fast strategic decisions ing and innovation in a product development firm.
in high-velocity environments. Academy of Man
Administrative Science Quarterly, 42: 716-749.
32: 543-576.
agement Journal,
Langley, A. 1999. Strategies for theorizing from process
Eisenhardt, K. M. 1989b. Building theories from case data. Academy of Management Review, 4: 691-710.
Suddaby, R. 2006. What grounded theory is not. Acad Kathleen M. Eisenhardt (kme@stanford.edu) is the Stan
emy ofManagement Journal, 49: 633-642. fordW. Ascherman M.D. Professor of Strategy and Or
Sutton, R. I., & Raphaeli, A. 1988. Untangling the rela ganizations in the Department of Management Science
and Engineering and the codirector of the Stanford Tech
tionship between displayed emotions and organiza
tional sales: The case of convenience stores. Acad nology Ventures Program, Stanford University. She re
ceived her Ph.D. from Stanford's Graduate School of
emy ofManagement Journal, 31: 461-487. Business. Her research interests include alliance and net
Weick, K. E. 2007. The generative properties of richness. ing both theory building from cases and simulation
methods.
Academy ofManagement Journal, 50: 14-19.
Melissa E. Graebner
Whyte, W. F. 1941. Corner boys: A study in clique be (melissa.graebner@mccombs.
havior. American Journal of Sociology, 46: 647 utexas.edu) is an assistant professor of management at