ORGANIZATIONAL PATH DEPENDENCE: OPENING THE BLACK BOX
←
→
Page content transcription
If your browser does not render page correctly, please read the page content below
姝 Academy of Management Review
2009, Vol. 34, No. 4, 689–709.
ORGANIZATIONAL PATH DEPENDENCE:
OPENING THE BLACK BOX
JÖRG SYDOW
GEORG SCHREYÖGG
Freie Universität Berlin
JOCHEN KOCH
European University Viadrina
To enable a better understanding of the underlying logic of path dependence, we set
forth a theoretical framework explaining how organizations become path dependent.
At its core are the dynamics of self-reinforcing mechanisms, which are likely to lead
an organization into a lock-in. By drawing on studies of technological paths, we
conceptualize the emergent process of path dependence along three distinct stages.
We also use the model to explore breakouts from organizational path dependence and
discuss implications for managing and researching organizational paths.
The discourse on organizational innovation mean exactly? In organization research the term
and change has become more complex. On the is used mostly as a broad label indicating all
one hand, there is an ever-increasing demand kinds of imprinting effects of the past on orga-
for more flexible or even fluid “new” organiza- nizational behavior (e.g., recently, Beckman &
tional forms. On the other hand, studies stress- Burton, 2008). A closer examination quickly re-
ing organizational inertia and the historical im- veals that the predominant usage is more met-
printing of decision making (“history matters”) aphorical than theoretical in nature. A clear
have come to the fore in management and or- specification is usually missing. This means, at
ganization theory. There seems to be a broadly the same time, that no indicators are available
shared feeling that we need to understand bet- that allow for examining whether or not the or-
ter how organizations can lose their flexibility ganizational phenomena in question are actu-
and become inert or even locked in. Among the ally path dependent. If we want “path depen-
most referred to conceptions, path dependence dence” to provide more than a synonym for
has recently gained prominence. Many contribu- persistence, then we need a theoretical frame-
tions refer to path dependence to illuminate or- work clarifying the notion and helping us better
ganizational rigidities, stickiness, or inflexibili- understand the conditions and dynamics under
ty.1 But what is path dependence supposed to which organizations become path dependent. By
addressing this gap in management and orga-
nization research, we aim to offer a framework
We thank the four anonymous reviewers for their helpful designed to explain organizational path depen-
comments, and we particularly thank former associate edi- dence.
tor Pamela Tolbert for her thoughtful advice. Earlier versions
of the manuscript profited significantly from discussions in
The endeavor to explain organizational rigid-
the subgroup on path dependence and creation at the 21st ities and structural inertia is not new in man-
EGOS Colloquium in Berlin, 2005, and at the annual meeting agement and organization research. Over the
of the Academy of Management in Atlanta, 2006. We are years scholars have accumulated ample evi-
grateful to the German Research Foundation (DFG) for fund-
dence on change-inhibiting forces. Various stud-
ing the doctoral program Research on Organizational Paths
at Freie Universität Berlin and to its members for providing ies have highlighted cases of persistence and
a stimulating environment for the research. irreversibility of organizational strategies, de-
1
A quick search for references to path dependence in signs, and competences by drawing, for in-
papers published between 1995 and 2008 in three leading stance, on awkward routines, groupthink, or
scholarly journals (Administrative Science Quarterly, Orga- fixed cognitive maps (e.g., Beckman & Burton,
nization Science, and Organization Studies) showed that
more than eighty papers referred to this concept. That is
2008; Burgelman, 2002; Collinson & Wilson, 2006;
about 4.3 percent of the articles published in those journals Gilbert, 2005; Hannan & Freeman, 1984; Helfat,
over this time span—an average of 0.3 papers per issue. 1994; Huff & Huff, 2000; Stimpert, Wasserman, &
689
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.690 Academy of Management Review October Jayaran, 1998; Tripsas & Gavetti, 2000). Path de- predominance for more than 100 years. This pendence, however, is supposed to mean more standard has spread around the world and, puz- than the mere existence of timeworn routines, zlingly enough, has never been seriously chal- cognitive rigidities, or structural inertia. It is, lenged by all the newly developed, technically first of all, a process. Its distinguishing features more efficient alternatives. David explains this need elaboration. inefficient long-term predominance as being the The starting point of any advanced path de- result of a path-dependent process, which was pendence thought stresses the importance of set up owing to some initial events and ad- past events for future action or, in a more fo- vanced mainly through network externalities cused way, of foregoing decisions for current leading to a technological lock-in early on. and future decision making. Hence, decisions The QWERTY case and similar case studies are conceived of as historically conditioned— from technology diffusion, economic history, “bygones are rarely bygones” (Teece, Pisano, & and evolutionary economics (e.g., Antonelli, Shuen, 1997: 522). In short, the basic thesis holds 1999; Callon, 1992; Castaldi & Dosi, 2006; Dosi, that history matters (e.g., Nooteboom, 1997; 1982; Hughes, 1987) offer intriguing evidence of Sewell, 1996). similar persistence in national and global con- This essential insight has certainly advanced texts. Arthur (1989, 1994) was the first to model a the understanding of emerging organizational formal theory of path dependence and to expose phenomena and has helped to overcome the increasing returns as the major process driver. ahistorical and unbounded view of rational Later on, this thinking was extended to the eco- choice thought. We learn that history can be nomics of institutions (North, 1990). However, up quite important for explaining strategic choices to now, studies of path dependence (in this spec- and organizational failures. While we appreci- ified sense) neither addressed the persistence of ate this insight, merely focusing on the fact of organizations nor explored the logic and dy- past dependence (Antonelli, 1999) implies taking namics of internal organizational processes a fairly broad view—too broad a theoretical per- leading to a lock-in. We fill this gap by elabo- spective: if we base path dependence explana- rating a theory of organizational path depen- tions on the history matters argument only, the dence and lock-in. Building on the evidence and notion is likely to become indistinct. All human insights from research on technological paths, activity and organizational processes are im- we develop a theoretical framework to gain a printed by their history in a way, so we would better understanding of how organizational end up by concluding that all organizational path dependence comes into existence. In pur- decisions and actions are path dependent. Such suing this aim, we also integrate insights from a ubiquitous, all-embracing understanding of institutional economics (North, 1990), as well as path dependence would bring us close to a tru- from political science (in particular, Mahoney, ism. Path dependence relates to more specific 2000; Pierson, 2000, 2004; Thelen, 1999), theories constellations; it includes features such as sus- of institutionalization (Lawrence, Winn, & Jen- tained persistency and lock-in, which are defi- nings, 2001; Powell & DiMaggio, 1991; Tolbert & nitely not a common characteristic of decision Zucker, 1996), and related organization studies processes. A theory of organizational path de- (Boeker, 1989; Johnson, 2007; Stinchcombe, 1965). pendence therefore needs a more elaborated In a subsequent section we will, however, also framework, which takes us beyond the mere in- show where the differences to these organiza- sight that past events influence subsequent tional approaches can be found. actions. In essence, we suggest a framework that dif- To gain a deeper understanding of the orga- ferentiates three developmental phases of path nizational patterns considered to be path depen- dependence, starting with (1) singular historical dent, along with their underlying causal mech- events, (2) which may, under certain conditions, anisms, it is instructive to explore the cases and transform themselves into self-reinforcing dy- conceptual suggestions provided by studies on namics, and (3) possibly end up in an organiza- technological paths. Paul David (1985, 1986) pro- tional lock-in. The three phases are each as- vides the most prominent example of technolog- sumed to be governed by different regimes. The ical path dependence—the well-known stan- suggested model aims at providing an explan- dard of the QWERTY keyboard and its amazing atory framework but also an operational scheme
2009 Sydow, Schreyögg, and Koch 691
for investigating claimed path dependence in dictable. Later on, as the path is increasingly
and of organizations. Furthermore, we explore formed, by implication, the actions become more
whether and how organizational path depen- and more predictable. Having arrived at the
dencies and lock-ins can be overcome (un- lock-in stage, the behavior even becomes fully
locked). We conclude by considering implica- predictable. The reverse is true for nonergodic-
tions for research and management. ity and inflexibility: it is only at the later stages
that a path process rigidifies. In the beginning
the process is assumed to be flexible. And, sim-
ADVANCING A DYNAMIC FRAMEWORK OF
ilarly, inefficiency becomes a feature of the later
ORGANIZATIONAL PATH DEPENDENCE
stages only; initially, before a path is shaping,
Valuable insights into the dynamic nature of the situation is open (unpredictable) and—as
entrapping or locking processes have already Arthur (1994: 116) himself stresses—the choices
been provided by studies from evolutionary eco- may well be efficient. It is only at a later stage
nomics and economic history, although these, as that a more efficient option may emerge, which
already indicated, focus almost exclusively on actors can no longer choose because they are
technological innovation at the field or market locked in, thus causing inefficiency.
level (Arthur, 1989, 1994; David, 1985, 1986; Dosi, These brief considerations advise us to differ-
1982, 1997). At the core, these studies identify entiate explicitly among different stages in the
self-reinforcing processes as drivers that are formation of a path and to specify their struc-
likely to accumulate in a specific path of action. tural properties. To elaborate on a theory of or-
These inherent self-reinforcing dynamics that ganizational path dependence, we therefore
eventually lead to an irreversible state of total suggest subdividing the whole process of evolv-
inflexibility or lock-in (David, 1985) are seen as ing path dependence into three stages governed
becoming increasingly systemic forces, beyond by different causal regimes and constituting dif-
the control of the individual actor. In other ferent settings for organizational action and de-
words, the individual actor becomes entrapped cision making.
in the system’s dynamics. Phase I—the Preformation Phase—is charac-
It is difficult to conceptualize the general logic terized by a broad scope of action. The effect of
of this type of entrapping process. Arthur (1994; a choice of options cannot be predicted (see also
see also Pierson, 2000: 253) has advanced its Mahoney, 2000: 511). Once a decision is made,
most explicit characterization. In his view the this choice may, however, amount to a small
process of becoming path dependent can be
event that unintentionally sets off a self-
characterized by four general properties:
reinforcing process. This moment of entering
1. Nonpredictability—there is an indetermi- into the dynamics of a self-reinforcing process
nacy of outcome. can be thought of as a “critical juncture” (Collier
2. Nonergodicity—several outcomes are possi- & Collier, 1991), and it indicates the end of the
ble (multiple equilibria), and history selects Preformation Phase. Drawing on complexity the-
among the possible alternatives.
3. Inflexibility—the actors are entrapped, so a ory, this transition comes close to “bifurcation”
shift to another option is impossible. (Kauffman, 1993).
4. Inefficiency—actions resulting from the In Phase II—the Formation Phase—a new re-
path lock the market into an inferior solu- gime takes the lead: the dynamics of self-
tion.
reinforcing processes (Arthur, 1994). A dominant
These four properties provide a first orienta- action pattern is likely to emerge, which renders
tion for differentiating between path-dependent the whole process more and more irreversible.
and non-path-dependent processes. However, By implication, the range of options narrows,
the properties seem to be somewhat overgener- and it becomes progressively difficult to reverse
alized, and they do not actually apply to the the initial choice or the initial pattern of ac-
whole process of becoming locked into a path. tion—that is, a path is evolving. Decision pro-
Rather, they appear to cover specific episodes in cesses in Phase II are, however, still contingent;
this process. Take, for instance, nonpredictabil- they are nonergodic (not accidental)—that is,
ity; this trait applies only to the beginning of the they do not yet fully converge to a fixed-point
process, when the outcome is actually unpre- distribution (David, 1985).692 Academy of Management Review October
The transition from Phase II to Phase III—the question that arises is how this initial state can
Lock-in Phase—is characterized by a further be conceptualized in more distinctive terms. The
constriction, which eventually leads to a lock- technological path studies—if at all— have con-
in—that is, the dominant decision pattern be- ceived of the initial situation as being unre-
comes fixed and gains a deterministic charac- stricted. The search for alternatives starts from
ter; eventually, the actions are fully bound to a scratch, and decisions are unconstrained.
path. One particular choice or action pattern has Such framing of the first stage in the rational
become the predominant mode, and flexibility choice tradition, however, paradoxically ignores
has been lost. Even new entrants into this field the fact that the development of a path is em-
of action cannot refrain from adopting it. When bedded and connected with other developments;
more efficient alternatives are available, indi- it cannot be considered a completely separate
viduals’ and organizations’ decision processes process without any imprints from the past. In
and established practices continue to reproduce brief, history matters in the Preformation Phase
this and only this particular outcome. The occur- too. In organizations initial choices and actions
rence of a lock-in renders a system potentially are embedded in routines and practices; they
inefficient, because it loses its capability to reflect the heritage—the rules and the culture—
adopt better alternatives. making up those institutions (e.g., Child, 1997;
Figure 1 illustrates the process across the March, 1994; Tolbert & Zucker, 1996). Institutions
three stages. This differentiated framework is are “carriers of history” (David, 1994), and his-
intended as a general model of path depen- tory cannot be intermittent; it does not matter
dence; its functioning, however, is likely to differ only occasionally—it always matters! A concep-
from context to context according to the prevail- tualization of the activities in the Preformation
ing conditions, particularly market versus hier- Phase thus cannot start from scratch; it has to
archy. The contextual specifics when applied to account for institutional imprints.
an organizational context—the target field of On the other hand, history in this broad sense
this contribution—will be outlined in subse- is not destiny; we have to draw a clear distinc-
quent sections. tion between historical-institutional influences
and imperatives. The notion of path dependence
does not refer to a state of determinacy from the
Preformation Phase
beginning; it sheds light on a tapering process
Phase I can be characterized as an open situ- that possibly ends in a lock-in. Increasing path
ation with no significantly restricted scope of dependence implies an initial scope of choice.
action. From a theoretical point of view, the Otherwise, the theory would lose its very point:
FIGURE 1
The Constitution of an Organizational Path2009 Sydow, Schreyögg, and Koch 693
to make tapering processes in organizational cases invite a rethinking of the small event
reality better understood. Thus, Phase I should framing: path dependence may be triggered by
build on a historically framed or imprinted con- “bigger” events or even strategies as well.
tingency and, therefore, neither on the assump- A less randomized modeling of these initial
tion of determinacy nor on that of completely activities thus seems advisable—at least for or-
unrestricted choice. The shadow in Phase I in ganizational contexts. It is, however, important
Figure 1 is intended to indicate this institutional to realize that in the nonlinear logic of path
heritage. dependence, irrespective of whether the initial
A related issue is the triggering of further re- actions are big or small, they can never be con-
actions. The initial choice in a process that be- sidered causal determinants. A determined pro-
comes path dependent later on is not simply a cess would follow a prescribed course of events
single event; it is an impetus, a trigger stimulat- right from the beginning, as is the case with
ing further actions, which may accumulate in an linear cause-and-effect laws. Opposed to that,
organizational path. Arthur (1994: 14) character- the very point of early path developments is that
izes these initial choices as “small events.” they are contingent in character. Moreover, their
Drawing on complexity theory (Kauffman, 1993), outcomes are unforeseeable consequences of
we conceptualize the triggering as bifurcation: purposeful action (Merton, 1936). The outcome
small events may cause unintended, far-reach- cannot be known unless the process has been
ing consequences—as is the case, for instance, formed.
with the well-known butterfly effect.2 Because
several outcomes initially are possible, the his-
Formation Phase
torical sequence of choices becomes decisive in
determining the final outcome; the first choice Phase II is characterized by the gradual emer-
(or action), however, is random (David, 1985). gence of an organizational path. The scope of
It is doubtless appealing to conceive of trig- action is assumed to narrow increasingly be-
gers of path dependence as small and random cause of the “pull” of the evolving path. An ini-
events, but for the purpose of organizational tially unknown regime3 happens to take the
analysis, we need to expand the scope. Since lead, which favors a particular type of decision
organizations are social systems and not mar- or action pattern and reproduces it over a cer-
kets or natural entities, triggering events in or- tain period of time. This phase commences with
ganizations are likely to prove to be not so in- a critical juncture at the passage from Phase I to
nocent, random, or “small” (cf. also Bassanini & II. A decision made or an action taken in Phase
Dosi, 2001). For instance, in the case of the VHS I amounts to a trigger for the further develop-
monopoly that has been intensively studied ment of the organization or an organizational
from a path dependence perspective (Cu- subsystem. However, not all cases of competing
sumano, Mylonadis, & Rosenbloom, 1992), the solutions culminate in path dependence. It is
triggering event was neither a random nor a therefore of critical importance to indicate such
small one. Rather, it was Matsushita’s initial cases in which path dependence is likely to de-
move to secure content delivery through an velop.
agreement with major Hollywood studios that Early studies on technological path depen-
happened to become the crucial step in defeat- dence (David, 1985, 1986) highlighted the central
ing the technologically superior Sony Beta stan- role of self-reinforcing processes for path build-
dard. Similarly, in Cowan’s (1990) study of nu- ing. Arthur (1989, 1994) elaborated on these driv-
clear power plants, initial choices reflected ing forces and specified “increasing returns” as
intentions, not randomness. These (and other) the decisive feature; this builds on the assump-
tion that the decision to reproduce a particular
2
Here the flap of a butterfly’s wings represents a small
3
random change in the initial condition of the system (atmo- It should be stressed that in Arthur’s well-known Polya
sphere), which sets in motion a chain of events eventually urn model, the self-reinforcing mechanisms are already set
causing a large-scale change (tornado). Had the small event up right from the beginning (Arthur, 1989); the experimenter
not occurred, the development of the whole system might determines the rules. From our point of view, however, this is
have been vastly different (for a more detailed account, see owing to the necessities of a formal model rather than a
Hilborn, 2004). theoretical statement.694 Academy of Management Review October
option is suggested by a utility calculus. While still prevalent situation of contingency). One of
in no way we exclude this case, it seems too the subsequent sections elaborates on major
restrictive a starting point for the organizational self-reinforcing processes in organizations.
context (see also critical comments by Crouch &
Farrell, 2004; Eden, 2004; Ortmann, 1995). Focus-
Lock-in Phase
ing on utility-driven behavior only implies a dis-
regard for important insights of organizational The transition from Phase II to Phase III is
studies. Self-reinforcing patterns in organiza- characterized by a further restriction of the
tions have been shown to result from other fac- scope. The focal action pattern is replicated
tors as well, such as emotional reactions (uncer- even more, which eventually leads the whole
tainty avoidance, intergroup revenge, etc.), setting into a lock-in. Because of the circum-
cognitive biases (selective perception, blind stances, this lock-in may be of a predominantly
spots, implicit theories, etc.), and even political cognitive, normative, or resource-based nature
processes (gaining and maintaining power, re- (Giddens, 1984). Although organizational studies
ciprocal negotiation). These aspects have to be mostly emphasize the role of managerial cogni-
included in a theory of organizational paths to tions or beliefs or resources, organizational lock-
represent the scope of organizational behavior ins are also likely to be combinations of all three
possibly activating self-reinforcing effects. More dimensions.
precisely, we suggest including different forms In its extreme form the dominant pattern
of positive feedback cycles based on specific gains a deterministic character, and alternative
organizational forces. courses of action are no longer feasible for var-
A related problem of the technological path ious reasons: high switching costs, sunk costs,
dependence studies results from their focus on monopoly, and so forth. By implication, further
individual decision making. This exclusive fo- decisions (owing to lack of alternatives, they are
cus on individuals does not account for the in- actually no longer decisions) are bound to rep-
stitutional setting in which organizational posi- licate the path. Even newcomers are forced to
tive feedback processes happen to occur. It is adopt it. Agents continue to reproduce this and
the broader organizational context (e.g., hidden only this particular outcome. This extreme form
assumptions of the organization, organizational of lock-in has been found with technological
culture, status and role system, and institution- solutions (e.g., the QWERTY keyboard).
alized practices) that informs decision makers Considering organizational paths, however,
and provides the basis, indirectly and inadver- the context seems to be significantly different,
tently, for the development of self-reinforcing requiring a somewhat modified conception of
loops. lock-in. Organizational settings cannot readily
On a general level the concept of increasing be equated with markets and monopoly. Be-
returns highlights positive feedback processes— cause of their social character, organizational
that is, the increase of a particular variable processes are more complex and ambiguous in
leads to a further increase of this very variable. nature. They are not likely to amount to a state
More specifically, the notion of increasing re- of full determinacy, which excludes any alterna-
turns indicates self-reinforcing processes with tive choices. Rather, self-reinforcing dynamics
increasing benefits; repetitive pursuits to earn are expected to bring about a preferred action
this increasing rent are likely to culminate in a pattern, which then gets deeply embedded in
patterned dynamic. Eventually, a dominant so- organizational practice and replicated. Hierar-
lution emerges in terms of recursive action pat- chy provides formal authority and legitimate in-
terns (Giddens, 1984). The flip side of these re- fluence on members’ behavior; orders can poten-
turns is that the whole process becomes more tially stop inefficient replication. On the other
and more irreversible, particularly in cases of hand, fixed and inflexible behavior is a widely
high investments and/or high fixed costs (Ghe- recognized feature in the organizational change
mawat, 1991). Decision processes in Phase II, literature. It is well known that, from time to
however, are still contingent or nonergodic time, despite hierarchical control, it is extremely
(David, 1985)—that is, while essentially con- hard to change organizational action patterns
strained, choices are still possible (the shadow (e.g., Beer & Nohria, 2000; Kaufman, 1995); they
of Phase II in Figure 1 is designed to indicate the are quasi locked in. We should nevertheless re-2009 Sydow, Schreyögg, and Koch 695 frain from reifying organizational paths and at- dress nonergodic inflexible processes (as op- tributing an objective quality to social rigidify- posed to linear ergodic processes). A separate ing processes. In organizational settings, question is whether the state finally reached is therefore, we suggest conceptualizing the final efficient or inefficient, and the answer to this stage of a path-dependent process in a less re- question is not considered part of the theory of strictive way—as a predominant social influ- path dependence. Although we appreciate this ence, leaving some scope for variation (Pierson, argument, from our point of view it misrepre- 2000; Thelen, 1999; for illustrative examples see sents the very intention of path analyses. The Bruggeman, 2002; Burgelman, in press; Eden, primary interest is not in the formal logic of 2004; Hollingsworth, 2006). nonlinear nonergodic processes as such; rather, In more detail, it seems promising to conceive it is nourished from congealing processes and of the lock-in stage in terms of an underlying puzzling persistencies that are likely to hamper core pattern (invisible “deep structure”), with present and future scopes of action. In other some variation in practicing it (visible activity words, it is at least potential inefficiency that is level). Actors in the final phase do not simply worrying and makes path dependence a matter experience the path; rather, as “knowledgeable of high importance. We therefore advocate in- agents” (Giddens, 1984), they have scope in in- cluding inefficiency in an organizational theory terpreting the organizational patterns. This in- of path dependence. dividual interpretation of the core (path) is likely It is true that a narrowing organizational pro- to bring about some variation in actual organi- cess and lock-in do not automatically mean im- zational action patterns. While the underlying mediate inefficiency or losses. Path dependence path structure is fixed, its replicative practice is and efficiency, however, do not refer to a certain subject to some variation. In a way, this argu- point in time; instead, a longer time horizon is ment echoes the conception of routines ad- covered, necessarily including the alerting risk vanced by Feldman and Pentland (2003), stress- of becoming dysfunctional. From a strategic, fu- ing, on the one hand, the ostensive side as a ture-oriented point of view, rigidity therefore al- fixed, overarching pattern and, on the other ways means potential inefficiency. If an organi- hand, the performative side as the actual prac- zation or a significant practice (e.g., combining ticing of a routine involving some variation. In specific R&D capabilities with marketing skills) conclusion, for organizational settings it seems has become locked in, there is inherently the more adequate to conceive of the lock-in state danger of becoming inefficient, either in the face not in terms of total rigidity but, rather, as a of new, more efficient alternatives or changed matter of degree, accounting for variance in the internal or external circumstances calling for actual practicing of the organizational path. A new solutions. Latent inefficiency becomes corridor may best serve to illustrate this reason- manifest when an organization confronted with ing; the shadow in Phase III in Figure 1 is de- these change requirements cannot adopt new signed to indicate this adaptation. Although measures because it is confined to the existing highlighting these differentiations, the lock-in path of action, which binds it to the historical phase is nevertheless constitutive for path de- solutions (Leonard-Barton, 1995; Schreyögg & pendence. If actors were not locked in, one Kliesch-Eberl, 2007). This dysfunctional flip or would not call the process path dependent. rationality shift from initial reinforcing earnings Whatever the best conceptualization of orga- to strong barriers to change and losses should nizational lock-ins, the more controversial fea- therefore be considered a constitutive element ture of this stage is efficiency. In David’s initial of organizational path analyses. framework (1985) inefficiency was considered a In any case, calling a lock-in “inefficient” al- necessary element, because he set out to ex- ways implies a base of reference—a comparison plain a puzzle: how could an inferior solution with another standard. The base of reference like the QWERTY keyboard endure in a market can differ; it is not a fact but, rather, depends on economy? So he started with inefficiency right the perspective taken (focusing on a group, a from the beginning. Subsequent work called this department, the whole organization, the field). element into question and suggested a modified By implication, discussing the inefficiency of an perspective (Arthur, 1994; Pierson, 2000)—the ar- organizational path always requires the expo- gument being that path analyses merely ad- sure of the base of reference applied.
696 Academy of Management Review October
To sum up, the proposed theory conceptual- Beckman & Burton, 2008), the process of becom-
izes an organizational path as a tapering social ing path dependent is governed by a different
process. Starting (Phase I) with contingency, a logic.
critical event (decision, accident, etc.) favors a First, the replicated pattern in the imprinting
solution leading unpredictably to a critical junc- approach is ready-made at the beginning; it is a
ture. If it triggers a regime of positive, self- specific scheme that persists and continues to
reinforcing feedback, this solution progressively influence future processes. In contrast, the ge-
gains dominance (Phase II). This pattern is stalt of an organizational path is not clear at all
likely to become persistently reproduced and to in the early stage; it is an unforeseeable product
crowd out alternative solutions to an extent that of later processes, which are initially unknown.
it gets locked in (Phase III) and is accompanied Path dependence is an offspring of the nature of
by immediate or future inefficiency. In short, the process. Second, because of this, a theory of
organizational path dependence can be defined organizational paths—as opposed to the im-
as a rigidified, potentially inefficient action pat- printing approach— has to explain the unfold-
tern built up by the unintended consequences of ing process of path formation, not only the re-
former decisions and positive feedback pro- production of structural properties because of
cesses. either efficiency or a lack of competition (Stinch-
combe, 1965), or the presence of institutionaliza-
tion processes (Johnson, 2007). Nevertheless, im-
COMPARING RELATED CONCEPTIONS prints doubtless play an important role in many
organizational processes. In path-dependent or-
The suggested framework needs discus-
ganizational processes they can, for instance, ex-
sion and refinement to further clarify the
plain the restrictions in the Preformation Phase.
causal logic of path-building processes. In
particular, the concrete forms of self-reinforc-
ing organizational dynamics need elabora-
tion. In a first step, however, it seems advis- Escalating Commitment
able to sharpen the model’s distinguishing
Another concept that shares striking similari-
features by contrasting it with related concep-
ties with organizational path dependence is es-
tions that also highlight the importance of ini-
calating commitment (Ross & Staw, 1993; Staw,
tial conditions and events for organizational
1976). As happens in cases of path dependence,
development, such as imprinting or escalating
particularly in the inefficient Lock-in Phase, es-
commitment.
calating commitment prevents organizational
decision makers from changing their course of
action, despite continued negative feedback on
Imprinting
the outcome. Instead of stopping, the agents rep-
The concept of imprinting (Beckman & Burton, licate the inefficient solution—in particular, the
2008; Boeker, 1989; Johnson, 2007; Stinchcombe, tendency to throw good money after bad (see
1965) figures prominently among approaches Guler, 2007)—for various reasons.
that seem to address a process very similar to There is, however, a major difference between
organizational path dependence. Basically, this escalating commitment and path dependence
concept postulates that either initial cognitive explanations. The latter consider a process with
schemes, competences, and so forth— of a a more or less accidental beginning and a
founding entrepreneur or team, for instance— or longer phase of success; it is only in the final
specific contextual conditions (organizational stage that the persistent course of action shifts
structure, postwar depression, internet boom, into inefficiency. In contrast, escalating commit-
etc.) at the time of founding imprint organiza- ment captures situations where the course of
tional processes at later stages and, eventually, action fails from the very beginning. Since there
amount to a replicated pattern. Although there are no increasing returns or similar enhancing
are doubtless striking similarities that lead effects, it highlights another problem area—
many authors to either simply equate imprint- namely, pathological decision behavior based
ing and path dependence or to conceive the on the dynamics of self-justification and fears of
former as a specific variant of the latter (e.g., losing face.2009 Sydow, Schreyögg, and Koch 697
Commitment/Sunk Cost organizational feature that develops in the
course of structuring the organization. Routin-
There is another related stream of thought high-
izing and institutionalizing organizational ac-
lighting persistence through resource commit-
tivities are seen as imperative in order to
ments and subsequent exit barriers (Ghemawat,
guarantee stakeholders reliability, account-
1991). Early investment is assumed to restrict the
ability, and, finally, survival in competitive
future scope of action. This argument comes close
environments. Inertia is considered a precon-
to sunk cost. Sunk cost thought, however, has its
dition for effective organizational acting but,
own ambiguity. As we can learn from microeco-
paradoxically enough, eventually threatens
nomics, sunk costs are only a psychological out-
the organization’s survival, because it is likely
come of imagination; from a rational choice point
to bring about a mismatch with changing en-
of view (e.g., Pindyck & Rubinfeld, 2005), they are
vironmental conditions.
irrelevant. Organizations can ignore them be-
Again, the phenomenon is somewhat similar
cause they are, in fact, not relevant for their future
to organizational path dependence, but the fo-
decisions. If there are new and better projects, the
cused process and its explanations are clearly
capital market will provide fresh money to over-
at variance. Inertia occurs via the intended es-
come the old structure.
tablishment of reliable organizational struc-
Apart from this idealized counterargument (it
tures; there are no structural dynamics. It is a
takes an efficient capital market for granted),
universal requirement that all organizations
the simple fact of past investments (in terms of
have to fulfill. And all organizations, especially
sunk costs) cannot be equated with path depen-
when growing and aging, are also expected to
dence, since, in consequence, all investments
become hyperstable, with difficulties in meeting
would bring about path dependence (for a more
new environmental challenges. Opposed to
sophisticated argument in this direction, see
that, the suggested framework of path depen-
Arrow, 2004). This refers back to our initial state-
dence does not apply to all organizations (it
ment that we should refrain from conceptualiza-
highlights special cases only) and requires an
tions that end up considering all past depen-
avalanchelike process to bring about a lock-in.
dence as path dependence.
The focus is on explaining the process and its
In a similar vein, Pierson (2000) raises the is-
various stages.4
sue of whether increasing returns or other self-
reinforcing processes should be made a neces-
sary element of path theory. Among others, he Reactive Sequences
highlights pure complementarities as being
While subscribing to the suggested type of
likely to bring about persistency. From our point
path-building process, Mahoney (2000) develops
of view, this argument refers to a different type
a second type—namely, efficient or inefficient
of rigidity. Pure complementarities without self-
trajectories built up by reactive sequences. This
reinforcing processes characterize a stable situ-
process is characterized by a chain of modular
ation of fitting resources, but not a process that
events governed by singular cause-and-effect
eventually leads to a lock-in. If there is no esca-
relationships. A focal event, B, is assumed to be
lating self-reinforcing process, switching to new
the effect of a prior event, A, and at the same
and better opportunities may be difficult but not
increasingly impossible.
4
Carroll and Harrison (1994) point to the importance of
positive feedback in the ecological model, but only with
Structural Inertia respect to density dependence; a more general consider-
ation of the importance of self-reinforcing processes does
There is another well-known argument on not seem to be intended. Other evolutionary and, more re-
organizational persistency that has been ad- cently, coevolutionary theorists make explicit use of the no-
vanced by population ecology (Gresov, Have- tion of path dependence and tend to prefer it to other con-
man, & Olivia, 1993; Hannan & Freeman, 1984; cepts (e.g., Helfat, 1994; Nelson & Winter, 1982; Volberda &
Lewin, 2003). Given its relatedness to economic evolutionary
Hannan, Plos, & Carroll, 2004; Ruef, 1997). In
theory, one seedbed of path dependence research (e.g., Dosi,
this perspective structural inertia—the hyper- 1982; Witt, 1997), this certainly comes as no surprise. Never-
stability of organizational arrangements in theless, the concept of path dependence has not yet been
spite of environmental change—is a universal fully utilized in this stream of research either.698 Academy of Management Review October
time the cause of a future event, C, etc., accu- ertia and stability by revealing how a specific
mulating in a reaction chain: A ⬎ B ⬎ C ⬎ D ⬎ organizational structure or form becomes
E ⬎ F. Thus, initial event A is expected to affect sedimented and taken for granted over time,
B, but it unintentionally triggers a multistage preferably across sets of organizations.
development. The final state or, better, an inter- While neoinstitutional theory elucidates im-
mediate result can—very much like pursuing a printing and stabilizing processes and, in partic-
lawsuit— be traced back to the releasing event. ular, sensitizes us to the relevance of symbolic-
This intermediate state is also likely to shape normative contexts (e.g., Hargadon & Douglas,
future action; it is, however, not in any way 2001), in its present form it does not address the
locked in or inefficient. systemic logic of an escalating reinforcement of
In contrast to the path dependence model ad- an action pattern or a path (see, however, Eden,
vanced above, the intermediate state of a se- 2004, and Holm, 1995). The theoretical focus,
quence of causal reactions is not reached by therefore, differs significantly and explains
increasingly reproducing a specific pattern, and other constellations.
there is no connecting logic that explains the
succession of the singular sequences. Although
the idea of reactive sequences doubtless pro- AT THE HEART OF ORGANIZATIONAL PATH
vides insights into the evolvement of historical DEPENDENCE: SELF-REINFORCING
processes, it does not fit into a theory of path MECHANISMS
dependence. Without path drivers and the
So far, we have conceptualized path-building
causal logic of a lock-in, a theory of organiza-
processes as processes of a diminishing scope
tional paths loses its very point.
of action that unintentionally develop their own
Furthermore, the sequence argument raises
pull and are driven by positive feedback. It is a
some conceptual questions. First of all, se-
time-based theoretical concept differentiating
quences seem simply to occur. In contrast to
between different states of flexibility/choice and
processes explained by the regime of self-
stability/determinism, respectively. The dy-
reinforcing mechanisms, the concept of causal
namic eventually flips over into rigidity. At their
reactive chains does not provide a logic that
heart, such processes can be explained by one
explains why the sequences take place in this
or a combination of several self-reinforcing so-
way and not in another way. Why do reactive
cial mechanisms. In this section we elaborate on
sequences accumulate? Superimposing, ex post,
these mechanisms in an organizational con-
a trajectory on reactive sequences does not pro-
text.5
vide an explanation. Another problem is gener-
In the field of technology development and
alization. A theory of path dependence aims at
diffusion, different types of self-reinforcing dy-
explaining a particular class of processes (Pet-
namics have been identified (Arthur, 1994;
tigrew, Woodman, & Cameron, 2001). Drawing
Cowan, 1990; David, 1985; Katz & Shapiro, 1985;
on reactive sequences does not, however, tran-
North, 1990). Since they have been developed at
scend a singular case: singular reasons are sup-
the market level, these mechanisms cannot
posed to explain singular events only.
readily be transferred to organizational analy-
sis. In our view four mechanisms in particular
Institutionalizing are likely to contribute to the development of
organizational path dependence: coordination
Contextual shaping forces play a major role in
effects, complementarity effects, learning ef-
neoinstitutional theory, and its concept of insti-
fects, and adaptive expectation effects. Below
tutionalization also seems to come close to path
we aim to combine different streams of thought
dependence (e.g., Powell & DiMaggio, 1991;
to build a framework of self-reinforcing dynam-
Scott, 2001; Tolbert & Zucker, 1996). Most impor-
ics at the level of single organizations and or-
tant, this theory highlights the relevance of the
ganizational subunits. We discuss these four
symbolic-normative environment of organiza-
tions and how this influences the formal and
informal structuring of organizations over time.
Apart from the pace of the development (Law- 5
For the more general debate on social mechanisms in
rence et al., 2001), it addresses institutional in- organization theory, see, for instance, Pajunen (2008).2009 Sydow, Schreyögg, and Koch 699
mechanisms and show how they apply to orga- in changing their R&D priorities into a new
nizational settings. product development competence. Similarly,
Gilbert (2005) described newspaper companies
who stuck to self-reinforcing rules for producing
Coordination Effects
a newspaper and thereby became unable to ex-
Initially put forward in institutional econom- ploit new online opportunities. More recently,
ics (North, 1990), these effects relate to the heart Koch (2008) provided evidence of similar pat-
of organizational functioning. They build on the terns in German quality newspapers. Adopting
benefits of rule-guided behavior: the more ac- joint rules of quality journalism brought about
tors adopt and apply a specific institution (i.e., significant coordination advantages and the
an organizational rule or routine), the more effi- lasting constitution of a once successful busi-
cient the interaction among these actors is, ness model. Nowadays, the flip side of this path
since the behavior of the actors is rule guided is broadly discussed.
and can therefore be anticipated and reactions
can be considered in advance. Coordination
Complementary Effects
costs can be significantly reduced. In conse-
quence, it becomes more attractive to adopt A well-known explanation for complementari-
these rules the more other individuals also fol- ties are economies of scope, which exist when
low them. the cost of producing and selling two or more
The best-known illustrative example of this goods or services together is lower than the cost
effect at the institutional level is the decision of producing and selling them separately (Pan-
regarding right-hand traffic versus left-hand zar & Willig, 1981). On a more general level,
traffic; the institution became fixed early on be- complementarities mean synergy resulting from
cause of the obvious benefits of following it— the interaction of two or more separate but in-
uncertainties involved in human interaction terrelated resources, rules, or practices (Pierson,
could successfully be reduced (North, 1990: 23). 2000; Stieglitz & Heine, 2007). In the case of
Another well-known example that applies di- complementarities, the advantages of repeat-
rectly to the level of single organizations is edly combining interrelated activities do not
working-time regimes, which guarantee effi- simply add up; they produce an additional sur-
cient cooperation. There is a striking similarity plus: K(x ⫹ y) ⬎ K(x) ⫹ K(y). Take, for instance,
to the economies of scale effect (North, 1990): marketing skills and R&D capabilities, which
increasing the number of participants results in may add up to a “core competence” (Prahalad &
decreasing (coordination) cost per unit. Hamel, 1990) of a company or a division. David
Miller and Friesen (1984) developed the con- (1994: 214) calls such combinations “institutional
cept of internal consistency, which comes very clusters.” In complementary settings self-
close to that of coordination effects. This propo- reinforcing processes occur when routines
sition stresses the advantages of an internal fit and/or practices are interconnected in such a
among the various elements of an organization way that it becomes ever more attractive to ex-
(see also Miller, 1992). Coordination effects thus ploit the synergies or—when referring to the re-
result from the benefits of following the same verse side—to save misfit costs caused by solu-
single rule or set of related rules to which others tions deviating from the established cluster/
are willing to conform. As a result, through the organizational capability. As a result, distinct
advantages of continued replication, a specific sets of activity patterns become progressively
pattern of practices is likely to become fixed. dominant (Leonard-Barton, 1995) and, addition-
The fixing power of such arrangements has ally, deeply embedded in an organization
been proved in cases where organizational (“deep structure”)—that is, they become organi-
members have recognized new challenges and zationally path dependent.
set out to change their practices but failed to do There are many other examples that can fur-
so because they could not get rid of their well- ther illustrate this effect. Take, for instance,
attuned activity sets and routines. “Fordism,” which is characterized by comple-
A striking example of such path dependence mentary management systems in human re-
was provided by Tripsas and Gavetti (2000), who sources (hiring and firing of low skilled labor),
portrayed the difficulties Polaroid experienced operations (mass production), and organization700 Academy of Management Review October
(hierarchy of control), and which for quite some increasing simplicity. Ultimately, “it turns into a
time constituted, through repeated practicing, a monolithic, narrowly focused version of its
specific capability amounting to a competitive former self, converting a formula for success into
organizational advantage. Ultimately, the insti- a path toward failure” (Miller, 1993: 116). Learn-
tutional cluster became path dependent (Piore & ing effects are often reinforced and extended by
Sabel, 1984). earnings from coordination costs and comple-
mentarities.
Learning Effects
Adaptive Expectation Effects
The learning effect theory holds that the more
often an operation is performed, the more effi- These self-reinforcing effects relate to the in-
ciency will be gained with subsequent itera- teractive building of preferences. With this con-
tions. The operation becomes more skillfully cept, as opposed to neoclassical economics, in-
performed (faster, more reliable, and with less dividual preferences are not considered to be
errors), which, in turn, means decreasing aver- fixed; instead, they are assumed to vary in re-
age costs per unit of output (Argote, 1999). And sponse to the expectations of others. Often
the more attractive the chosen solution becomes quoted examples highlight the need for social
because of accumulated skills and decreasing belonging and the desire to end up on the win-
cost, the less attractive it is to switch to new ning side. The more people are expected to pre-
learning sites (where the actors would have to fer a particular product or service (and not an-
start from scratch). Only sticking with the once other), the more attractive that product or service
chosen solution promises continued returns— becomes (Leibenstein, 1950). Since users are of-
although, as is well known, the resulting cost ten uncertain about the right choice, they feel
curve flattens after a while. rewarded by the fact that others are likely to
Self-reinforcing learning effects can be found prefer the same. Because of this self-reinforcing
at various organizational levels. A well-known dynamic, a dominant solution is likely to
example from organizational learning points to emerge, more often than not by way of a self-
the fact that a focus on the advantages of ex- fulfilling prophecy (in most cases on the basis of
ploitative learning may increasingly drive out more or less random first choices and from hear-
explorative learning (March, 1991, 2006). For var- say).
ious reasons (e.g., the prevailing organizational In the context of organizations, the informal
culture and reward system) the motivation to diffusion of best practices often follows this
improve everyday practices is likely to gain logic (Szulanski, 1996). Organizational members
more acceptance or legitimacy (and, thus, more are willing to adopt these practices because
rewards) from the organization, whereas the mo- they expect others to do the same and wish to
tivation to look for fresh alternatives and to crit- end up on the side of the winners. This tendency
ically examine well-established organizational is reinforced by other drivers, such as legitimacy
practices is likely to shrink progressively. This seeking or signaling; individuals or subsystems
myopia or preference for repetitive exploitative not subscribing to the mainstream practices are
learning builds on the self-reinforcing dynamics afraid of losing legitimacy and—if associated
of learning effects, eventually ending up in an with failure— of becoming stigmatized as “out-
organizational path along the familiar prac- siders” (Kulik, Bainbridge, & Cregan, 2008).
tices. Early on, McGregor’s (1960) Theory X nicely
A related effect has been highlighted by the illustrated the dynamics of such self-reinforcing
“architecture of simplicity” (Miller, 1993), in adaptive expectations and subsequent self-
which an organization develops a successful set fulfilling prophecies in organizations. The start-
of strengths and tends to focus all learning abil- ing point of his Theory X spiral is managers’
ities on refining this success; it exploits these implicit assumptions about the nature of their
strengths through gaining learning effects employees—as being interested only in mone-
while neglecting other opportunities. The ex- tary rewards, hating to take on responsibility,
ploitation is easier (more efficient) the simpler and shirking wherever they can. This implicit
the institutional cluster; therefore, the self- theory of human behavior not only defines the
reinforcing dynamics bring about unintended set of expected managerial behaviors but also2009 Sydow, Schreyögg, and Koch 701
essentially frames management’s decisions on are relevant contextual factors, but they should
reward systems and organizational design (in not be equated with self-reinforcing mecha-
particular, a strong emphasis on control and nisms. Enhancing contexts— however important
authority), which, in turn, evoke corresponding they may be—neither lead directly to path de-
reaction patterns (especially passivity, indo- pendence nor represent a necessary or even suf-
lence, and apathy). Observing those reactions is ficient condition for the occurrence of path de-
likely to confirm and reconfirm exactly those pendence (see also Arthur, 1989). A theory of
assumptions about behavior managers have organizational path dependence has to differen-
made, based on their implicit Theory X. These tiate properly between self-reinforcing mecha-
confirmed expectations then reinforce the em- nisms on the one hand and enabling institu-
phasis on restrictive organizational structures tional contexts on the other. In consequence,
and controls, thereby unconsciously advancing Pierson’s insights should encourage further re-
a vicious circle (see also Leonard-Barton, 1995; search to explore the contextual conditions en-
Masuch, 1985; Repenning & Sterman, 2002). In hancing (or hindering) the unfolding of self-
this case a dominant organizational design reinforcing mechanisms and subsequent
emerges because of a self-reinforcing spiral that constitution of organizational paths.
is based on “expectations of expectations” (Luh-
mann, 1995).
IMPLICATIONS FOR MANAGEMENT
AND RESEARCH
Adding an Enhancing Context?
Path dependence and its far-reaching conse-
Some authors add contextual conditions as a quences doubtlessly constitute an issue of high
further reinforcing effect of and in institutions. relevance in strategic management and organi-
Pierson (2000), for instance, highlights institu- zational decision making. From a managerial
tional density as a salient determinant likely to point of view, the fatal consequences of being
converge into self-reinforcing effects in organi- locked in raise the pressing question of whether
zations. In his view organizations (in particular, organizational paths can be dissolved or in any
formal political institutions) are more prone to way escaped. It is true that no path is forever,
bring about path-building forces than markets, but this is no relief from the perspective of a
because they act in “a far, far murkier environ- particular organization, since path dependence
ment” (Pierson, 2000: 260) with weaker forces to may exist for quite some time.
correct inefficient courses of action over time. Path dissolution may occur through unfore-
The complexity of organizational goals and the seen exogenous forces, such as shocks, catastro-
uncertainty of the causal links between actions phes, or crises; these are likely to shake the
and outcomes render the organizational field system, thereby causing the organization to
inherently ambiguous, and organic corrections break away from the path (Arthur, 1994: 118).
of inefficient action are less likely to occur here However, path dissolution may also occur be-
than in markets. Therefore, practices, once es- cause of an insidious change in organizational
tablished, gain momentum more easily and cre- demography or the “incomplete” socialization of
ate a fertile ground for developing increasing new organizational members (Tolbert, 1988). In
returns or other types of positive self-reinforcing this vein, Castaldi and Dosi (2006) refer to the
feedback. More generally, Pierson considers possibility of coincidental delocking in terms of
ambiguity and complexity important conditions, a by-product of other organizational decisions. A
which amount to self-reinforcing effects and nice illustration of such coincidental path disso-
subsequent path dependence. lution at an organizational level is provided by
Although addressing doubtless significant the Intel case and its moves in the memory busi-
contextual conditions for path development, the ness (Burgelman, 1994, 2002; Burgelman &
conditions of ambiguity and complexity should Grove, 1996). With this perspective, however,
not be misconceived as self-reinforcing mecha- path dissolution amounts to an accidental pro-
nisms in their own right. This also holds true for cess, be it revolutionary or evolutionary,
other factors addressed in the literature as self- which—nobody knows—may or may not occur.
reinforcing dynamics, such as “uncertain expec- Adopting this view clearly has a fatalistic or at
tations” or “power structure” (Beyer, 2005). These least a passive flavor to it. We are condemned toYou can also read
