Content uploaded by Thomas Rixen
Author content
All content in this area was uploaded by Thomas Rixen on May 13, 2016
Content may be subject to copyright.
Article
Putting path dependence in its
place: toward a Taxonomy of
institutional change
Journal of Theoretical Politics
2015, Vol. 27(2) 301–323
©The Author(s) 2014
Reprints and permissions:
sagepub.co.uk/journalsPermissions.nav
DOI:10.1177/0951629814531667
jtp.sagepub.com
Thomas Rixen
University of Bamberg, Germany
Lora Anne Viola
Free University Berlin, Germany
Abstract
Stalled progress on explaining institutional change is, in part, the result of two conceptual chal-
lenges that hinder effective theory building: concept stretching and concept proliferation. These
problems affect a hallmark concept of institutional change, path dependence, whose usefulness has
been curtailed by the variety of meanings attributed to it. This article seeks to remedy concept
stretching and proliferation by developing a taxonomy of institutional change explanations. Start-
ing with the core attributes of path dependence, increasing returns and endogeneity, we use the
procedure of ‘negative identification’ to derive a logically complete set of possible change expla-
nations. The result is a taxonomy in which the scope of path dependence is delimited vis-à-vis
other change explanations. We illustrate the usefulness of the taxonomy by assessing stretching
in the literature.
Keywords
Concept formation; endogenous change; increasing returns; institutional change; path dependence
1. Introduction
The study of institutions over the last several decades has greatly advanced our under-
standing of political processes (Hall and Taylor, 1996; Pierson, 2004; Streeck and Thelen,
2005). Nevertheless, there is a growing awareness that, despite the institutional turn, we
still lack good explanations of how institutions themselves change (Mahoney and Thelen,
2010). As Mahoney and Thelen write, ‘the vast literature that has accumulated provides
us with precious little guidance in making sense of processes of institutional change’
Corresponding author:
Thomas Rixen, University of Bamberg, Professorship of Comparative Public Policy Feldkirchenstr, 21 96052
Bamberg, Germany.
Email: thomas.rixen@uni-bamberg.de
at Universitaetsbibliothek Bamberg on April 8, 2015jtp.sagepub.comDownloaded from
302 Journal of Theoretical Politics 27(2)
(Mahoney and Thelen, 2010: 2). In their view, this shortcoming is the result of the lit-
erature’s focus on institutional stability and exogenous sources of change. Given this
diagnosis, it makes sense to respond, as they do, by theorizing endogenous change. In
our estimation, however, stalled progress on explaining institutional change is not the
result of a lopsided focus on one set of sources and therefore cannot readily be amelio-
rated by simply theorizing ‘the other side’. Indeed, a wide range of causes and causal
mechanisms can already be identified in the literature.
We argue, instead, that understanding institutional change is hindered by conceptual
challenges faced by any maturing field: concept stretching and concept proliferation.
On the one hand, as existing concepts get applied to a range of change phenomena,
the boundaries of those concepts are liable to get stretched, r esulting in a diversity
of meanings ascribed to a single concept. On the other hand, as new phenomena and
processes are identified, there is a tendency to capture them by developing new con-
cepts, resulting in descriptive rather than explanatory concepts. These pitfalls of concept
development hinder effective and cumulative theory building because logically distinct
mechanisms and processes of change get subsumed or obscured, thus artificially narrow-
ing the range of potential explanations. The challenge for theory-building, then, is to find
a balance between stretching the boundaries of existing concepts and engaging in concept
proliferation.
These challenges are visible in one of the hallmark concepts of institutional change,
path dependence. Path dependence has served as the point of departure for the recent and
ongoing debate on how to theorize institutional change. In addition to contributions that
focus on better theorizing path dependence as such (Boas, 2007; Page, 2006), a number
of important contrib utions to the institutional change literature borrow elements from
path dependence and combine these with other mechanisms and processes ( e.g. Bednar
et al., 2012; Greif and Laitin, 2004; Mahoney, 2000). Perhaps most prominently, path
dependence is used as a foil against which newer concepts such as layering, conversion,
displacement or drift, are developed (Mahoney and Thelen, 2010; Streeck and Thelen,
2005; Thelen, 2003). As a number of scholars have noted, however, path dependence is
often used in a variety of ways to mean a variety of things (Bennett and Elman, 2006;
Greener, 2005; Mahoney and Schensul, 2006). In its broadest form, path dependence
is used to refer to the vague notion that history matters or that the past influences the
future (Sewell, 1996). Some scholars use path dependence simply to express the idea
of gradual or incremental change in one direction over time (Nor t h, 1990). But path
dependence is also more narrowly understood to refer to the idea of institutional ‘lock-
in’ by which change becomes impossible or unlikely (see, for example, the discussion of
Pierson, 2004). Some authors have added that this period of stasis and path reproduction
must be preceded by a critical juncture (Mahoney, 2000). Still others, however, argue that
path dependence is in fact compatible with a number of mechanisms for path change,
and that path dependence may even sometimes be characterized by a self-undermining
process (e.g. Beyer, 2010; Page, 2006). This variety of meanings and usages indicates,
in the first instance, the great resonance that the concept has had among scholars. At the
same time, as Pierson has suggested, ‘[T]he fuzziness that has marked the use of this
concept in social science suggests that the greater range offered by looser definitions
has come at a high price in analytical clarity’ (Pierson, 2004: 21). We too argue that the
at Universitaetsbibliothek Bamberg on April 8, 2015jtp.sagepub.comDownloaded from
Rixen and Viola 303
variety of meanings taken on by path dependence is symptomatic of concept stretching
and indeed problematic.
We propose that a necessary first step for building better theories of institutional
change is pre-theoretical and conceptual; it requires assessing the logical limits and
connections among existing concepts. The key to avoiding concept stretching and prolif-
eration is to delimit the analytical scope of a concept and to use additional concepts only
when that scope is exceeded. A solution, therefore, requires clarifying the core meaning
of the concept so that we have a reference point for stretching, and then embedding this
meaning in the context of the larger range of institutional change explanations in which it
already implicitly exists. We carry out this conceptual exercise by developing a taxonomy
of institutional change explanations in which the scope of path dependence is delimited
vis-à-vis other types of institutional change explanations.
1
We show that the two defining
attributes of path dependence, according to its most parsimonious and widely used defi-
nition, are that it is endogenous and exhibits increasing returns. Path dependence can then
be distinguished from a number of other mechanisms and processes that are characterized
by other attributes, namely exogeneity and non-increasing returns.
We use path dependence as a starting point for the taxonomy because it is one of
the best developed and most explicitly theorized explanations of institutional change that
is available, and it is often used as the basis for further development of theor y. In fact,
in the process of developing and applying the concept of path dependence a great deal
has been learnt about institutional dynamics. Our point is that many of these compelling
accounts are in fact making use of distinguishable change mechanisms, important in their
own right, that get obscured by subsuming them under the label of path dependence. The
taxonomy exercise, then, aims to enhance the explanatory power of path dependence and
to enable us to build more self-reflective, and therefore stronger, theories of institutional
change.
We begin, in the following section, by introducing the idea of a taxonomy as a useful
tool for addressing the problems associated with concept stretching. In section 3 we then
use existing theory t o identify the defining attributes of path dependence. These core
characteristics form the basis for developing a taxonomy, in section 4, which systematizes
various concepts and mechanisms of institutional change. In section 5, then, we use the
taxonomy to diagnose the contours of concept stretching in the literature. In conclusion,
section 6, we discuss how the distinctions developed in this article can shed fresh light
on a number of controversial issues currently being discussed in the institutional change
literature, including the relationship of path dependence to critical junctures, and the
relationship between continuity and change.
2. Diagnosing concept stretching: the problem and a solution
Creating and applying concepts is a central part of theorizing that is, however, fraught
with challenges. Concepts, such as path dependence, gain analytical rigor when they are
clearly delimited. But at t he same time, knowledge is often built by applying concepts
to a range of cases, which in turn often requires adapting or re-interpreting the original
concept. In his classic work on concept formation, Sartori calls the application of con-
cepts to new cases ‘travelling’ (Sar tori, 1970: 1033). Applying and adapting a concept
at Universitaetsbibliothek Bamberg on April 8, 2015jtp.sagepub.comDownloaded from
304 Journal of Theoretical Politics 27(2)
across cases is useful in that it allows us to make comparisons and to impose analyti-
cal order on the world without requiring new concepts for each phenomenon of interest.
The danger of fitting a concept to new cases, however, is that the meaning of the con-
cept can become distorted in a process that Sartori calls ‘conceptual stretching’ (Sartori,
1970: 1034). Concept stretching runs at least two risks: that of undermining the valid-
ity of coding and measurement across cases because scholars use the same concept to
refer t o different empirical phenomena; and that of hindering effective and cumulative
theory building because of the diversity of meanings ascribed to a single concept. The
challenge, then, is to achieve ‘the virtue of conceptual travelling without committing the
vice of conceptual stretching’ (Collier and Mahon, 1993: 845).
In trying to strike this balance it is important not to overcompensate for stretching
by creating a multitude of new concepts. In the extreme, this could lead to creating a
new concept for every new case. We certainly need conceptual richness to help explain
social phenomena, but there are at least two dangers associated with concept prolifera-
tion: concepts become increasingly descriptive and lose the analytical value inherent in
abstraction; and the relationship between multiple concepts may become attenuated and
less systematic.
One strategy for avoiding concept stretching and concept proliferation is to systemize
concepts into a taxonomy which delineates the scope of concepts relative to one another
(Elman, 2005; Sartori, 1970).
2
The idea of a taxonomy is not to create a theory or expla-
nation, but rather to map out the available explanatory concepts and their logical status
with respect to one another. We build our taxonomy in two steps. The first step is to
determine a concept’s boundaries or, in other words, the set of meanings or attributes that
define the category and identify which cases belong to the category. (Collier and Mahon,
1993; Sartori, 1970).
3
The attributes of a concept are extracted from pre-existing theories
and explanations. We take path dependence as a starting point for our taxonomy because
it is the best developed and most explicitly theorized explanation of institutional change
that we have, and it is therefore possible to identify its core causal attributes (see section
3).
Second, once we identify the core attributes of path dependence, we can begin to
identify and expand the implicit typology that it belongs to in order to provide a fuller
account of the entire explanatory space (see section 4). To do this we unpack the core
attributes by doing what Sartori calls ‘negative identification’ and Elman calls ‘reverse
engineering’ (Elman, 2005: 308; Sartori, 1970: 1042). Path dependence has explanatory
power precisely because it refers to something rather than to everything; its attributes are
determinate and bounded. This means that it is possible to determine what the concept
is not, and the negation of the concept, in turn, reveals an alternative concept (Sartori,
1970: 1042). For example, if we understand increasing returns to be a core attribute
of path dependence, then implicit in this attrib ute is the (at least theoretical) existence
of alternative concepts characterized by constant or decreasing returns. Thus, we work
backwards to fill out the entire logically given domain while keeping clear analytical
distinctions between path dependence and other explanatory concepts.
A taxonomy is a useful tool for avoiding both concept stretching and concept pro-
liferation because it identifies each concept’s central attributes and thus helps to clarify
the boundaries between multiple concepts. It allows for the discovery of cells in the tax-
onomy that logically exist but are currently empty (that is, have not been theorized or
at Universitaetsbibliothek Bamberg on April 8, 2015jtp.sagepub.comDownloaded from
Rixen and Viola 305
conceptualized), and it allows us to discover missed or obscured combinations (Elman,
2005). It also helps to determine whether a particular explanation really belongs to the
type ‘path dependence’. We want to emphasize that there is no special virtue in having
a ‘pure’ path dependence explanation of institutional change; sometimes a combina-
tion of concepts will provide the most convincing explanation. What is important is to
reflect on whether and when the concepts we use are being combined with other, distinct
explanatory concepts.
With the steps needed to build a taxonomy in mind, we turn now to a discussion of
path dependence and its core attributes which will serve as the starting point for devel-
oping a taxonomy of institutional change explanations in section 4. As Gerring points
out, ‘the most coherent definitions are those identifying a ‘core’ or ‘essential’ meaning’
(Gerring, 1999: 374).
3. The core meaning of path dependence and its adoption in
institutional theory
The idea that ‘history matters’ has a long tradition in the social sciences. Scholars like
Max Weber, Karl Marx and Joseph Schumpeter have constructed arguments in which
past decisions limit the available range of current choices, or in which some social pattern
causes its own reproduction over time (Stinchcombe, 1968). While the general idea can
thus be traced back at least to these classic works, it was not until the 1970s and 1980s that
a systematic and deductive theory of why and how past decisions can matter for future
ones was developed by W Brian Arthur and Paul David. It is from their formulation of
the concept that we extract two core causal attributes of path dependence.
According to Arthur (1994) and David (1985; 2007), path dependence is caused by
a process of self-reinforcement. The term self-reinforcement conveys the two central
causal attributes of path dependence. First, there is the reinforcing element, which is
caused by increasing returns. This refers to economists’ understanding of a production
process that exhibits increasing returns to scale, defined as a reduction in t he cost per unit
resulting from increased production. The relevant definition of increasing returns in the
context of institutional change is ‘that the more a choice is made or an action is taken,
the greater its benefit’ (Page, 2006: 88). We can regard this as a situation in which the
returns actors derive from following an institutional rule or practice increase relative to
the initial investment. This is in contrast both to constant returns to scale, in which initial
choices always yield proportional benefits, and to negative returns to scale, in which
initial choices yield decreasing benefits.
The second causal attribute of path dependence refers to the ‘self’ in self-
reinforcement. A path dependent process reinforces itself through endogenous variables.
Endogeneity is given when there is a cause-effect circuit or feedback loop. The idea is
that the system being studied (whether a technological standard or an institution) has
effects which then become the causes of subsequent effects on the same system, which
in turn become causes once again ( Büthe 2002). This is in contrast to processes in which
factors exogenous to the system are the causes of change.
4
These two core attributes – increasing returns and endogeneity – give rise to cer-
tain features commonly associated with path dependent processes. In particular, Arthur
(1994) has shown t hat equilibrium selection under conditions of increasing returns and
at Universitaetsbibliothek Bamberg on April 8, 2015jtp.sagepub.comDownloaded from
306 Journal of Theoretical Politics 27(2)
endogeneity exhibits three effects. First, it is unpredictable and contingent. In contrast to
negative or constant returns processes, which have a unique equilibrium, a system sub-
ject to endogenous increasing returns has multiple equilibria which are initially equally
likely to be realized. Second, they are non-ergodic. This means that the order in which
choices are made matters for the equilibrium that eventually emerges. This is so because
under endogenous increasing returns the choices made in the beginning of the selection
process are amplified so that it pays to go along with early choices (even though they may
not be efficient).
5
Third, for that reason, once initial choices have been made, the system
eventually locks in to one equilibrium.
6
Path dependence provides a rigorous theoretical basis for the claim that ‘history mat-
ters’ in the analysis of economic, political and social institutions. Past choices may have
been purely accidental or have at the time been viewed as unimportant events, but these
early choices may prove to have important consequences later on. This also means that
the sequence of events is significant and influences outcomes. Moreover, early events get
amplified due to increasing returns and get locked in. Thus, there is determinacy at the
end of the process (see for example Mahoney, 2000; Pierson, 2004).
Institutionalist scholars of all disciplines tend to converge on this model of path
dependence. In doing so, scholars often invoke path dependence to emphasize one or
more of its consequences; e.g., some scholars emphasize the constraints path depen-
dence puts on future choices, others the contingency of path dependent processes and
others the stability of path dependent processes. All of these consequences, however,
derive from the two core attributes of increasing returns and endogeneity. Increasing
returns and endogeneity, which are the characteristics on which we will build our tax-
onomy, are causal attributes of path dependence, meaning that they describe the types
of variables and mechanisms that cause path dependence. In contrast, initial contin-
gency, the importance of sequence, and lock-in are consequences or effects of endogenous
increasing retur ns. While identifying these effects is an important component of path
dependence theory, a path dependence explanation will only be complete if it can be
shown that there are endogenous increasing returns that cause it.
7
Indeed, increasing
returns and endogeneity are the two main causal attributes invoked in the literature on
path dependence.
8
Arthur and David identify four mechanisms that can produce endogenous increasing
returns: high fixed or setup costs (leading to decreasing costs for each additional unit
produced); learning effects (also leading to decreasing unit costs); coordination effects
(when actors derive utility from going along with the decisions of other actors); and
adaptive expectations (when all actors expect positive coordination effects in the future)
(Arthur, 1994: 112; David, 1985).
Almost all scholars working with path dependence explicitly discuss increasing
returns as a defining attribute of path dependence (see the short discussion of the major
contributions to the social science literature in Page, 2006; also Pierson, 2004; Greif
and Laitin, 2004). Accordingly, these scholars explicitly try to show that many polit-
ical, social, and cultural institutions are subject to increasing returns. They argue that
high fixed or setup costs, learning, coordination and adaptation, are often present in
such institutions and that path dependence should therefore be widely observable (North,
1990; Mahoney, 2000: 508; Pierson, 2000; Pierson, 2004). Pierson, for example, argues
that politics generally involves increasing returns because democratic politics requires
at Universitaetsbibliothek Bamberg on April 8, 2015jtp.sagepub.comDownloaded from
Rixen and Viola 307
actors to make high initial investments in institutions (set-up costs), that these institutions
tend to complement one another (coordination effects), and that they tend to reinforce
shared understandings (adaptation). This may explain, for instance, why party systems
are relatively stable over time (Pierson, 2004).
Endogeneity is also widely used as an attribute of path dependence in the literature.
This may at first appear to be an unusual claim because many authors do not explic-
itly use the terms ‘endogenous’ or ‘endogeneity’ when discussing the causes of path
dependence.
9
But the major contributions to the literature very prominently refer to
the fact that the processes they are interested in rely on self -reinforcement or positive
feedback (see, for example,. David, 1985; Arthur, 1994; Thelen, 1999; Pierson, 2004;
Page, 2006). These authors make clear that self-reinforcing processes are characterized
by endogeneity in the sense of a closed cause and effect circuit (Greif and Laitin, 2004).
Endogeneity is what makes the idea that ‘history matters’ different from a simple time-
series notion of causality: cause and effect amplify each other. One example of such
endogeneity may be an institution creating a constituency that comes to condition its
actions on that institution, and that experiences increasing returns from the continua-
tion of the institution. Thus, the constituency created by the institution has an interest
in maintaining and expanding the institution. It has been argued, for example, that the
expansion of welfare states in the post-WWII era created a new political constituency of
welfare state beneficiaries which successfully rallied against conservative parties’ efforts
at welfare state retrenchment (Pierson, 1996).
While a close reading of the major contributions to the path dependence litera-
ture thus confirms that increasing returns and endogeneity are broadly accepted as core
attributes of path dependence, some contributions to the literature expand or refine the
concept to include additional attributes. Greener, for example, suggests that path depen-
dence cannot be characterized by increasing returns alone because this would make path
change impossible even though we do know that institutions change (Greener, 2005).
His conclusion, therefore, is that ‘we need some modification of the idea’ to include
other returns to scale (Greener, 2005: 69). Similarly, Mahoney’s (2000: 526–535) sug-
gestion that ‘reactive sequences’ be considered instances of path dependence can be seen
as an attempt to include decreasing returns as an attribute of path dependence.
10
Such
modifications of the path dependence concept allow it to fit a wider range of empirical
observations, but also lead to complaints of concept stretching (Mahoney, 2000; Pierson,
2004).
We do not propose our definition of path dependence with the aim of denying other
possible usages. We do argue, however, that a lax use of the concept of path dependence
comes with costs. First, it makes case comparisons across the literature difficult. It is
valuable to have recourse to a clearly delimited definition when confusions arise so that
these can be adjudicated. Second, expanding the meaning of path dependence runs the
risk of obscuring distinct causal processes. Clearly defining path dependence is important
especially because it might be necessary to employ multiple processes and mechanisms
to fully explain an institution’s development. In the following section, we argue that the
core attributes that we have derived can be delimited from alternatives that deserve their
own conceptual space – and their own theorizing.
at Universitaetsbibliothek Bamberg on April 8, 2015jtp.sagepub.comDownloaded from
308 Journal of Theoretical Politics 27(2)
4. A taxonomy of institutional change explanations
The utility of defining the core attributes of a concept is that it can then be distinguished
from other concepts and positioned into a broader typology. Our aim in developing a
taxonomy is to capture and systematize the range of possible categories available for
explaining institutional change, and within this to make clear where path dependence
belongs and how it relates to other explanations of change. By ‘institutional change’ we
mean the full variation in institutional dynamics – from change to no change. Following
common practice in the literature, we define ‘institutions’ as ‘the rules of the game in a
society’ (North, 1990: 3); the ‘implicit or explicit principles, norms, rules, and decision-
making procedures around which actors’ expectations converge’ (Krasner, 1982: 186).
This definition is purposefully broad and can accommodate a variety of more specific
understandings of institutions, so that we expect our taxonomy to be robustly applicable
across a large range of alternative definitions.
Our taxonomy is meant to capture the building blocks for explanations of why and
how institutions change. We are not interested in capturing concepts whose main function
is to describe particular patterns of change, such as layering, conversion, displacement or
drift.
11
We also are not interested in including concepts that describe the pace of institu-
tional change; that is, whether change is fast or gradual. Path dependence may empirically
be a slow-moving process, but this is best s een as a description rather than an explanation
of the process. Excluding such characteristics helps to avoid the conceptual confusion of
using path dependence both to describe the dependent variable (type of change) and to
explain that change.
Path dependence gains its explanatory power by specifying a particular type of causal
variable (endogenous) and a particular type of causal mechanism (increasing returns) to
explain change. Endogeneity is not itself a causal variable but an attribute that describes
a particular category or type of cause. Increasing returns, on the other hand, refers not to
a type of variable but to a type of causal mechanism. It is a particular kind of relation-
ship between independent and dependent variables. In focusing on specific categories
of causal variables and mechanisms, path dependence logically implies alternative cat-
egories not specified in the theory. We star t from the core characteristics of the given
theory and, using ‘reverse engineering’, search for the alternative categories that provide
logical completeness. By making explicit the alternatives implicit in path dependence, we
can map out a larger explanatory space. We depict the set of alternatives in a taxonomical
tree (see Figure 1).
4.1. Categories of causes
Understanding path dependence as a process driven by endogenous causes implies a
distinction from other processes in which the causal variables are not endogenous but
exogenous. We argue that these two categories, endogenous and exogenous, describe
the full range of sources from which the independent variables available for explaining
institutional change can be drawn.
The categories ‘endogenous’ and ‘exogenous’ distinguish variables on the basis of
their causal relationship to the institution in question. A variable is endogenous when its
value is determined or influenced by an institution, and it in turn affects that institution’s
at Universitaetsbibliothek Bamberg on April 8, 2015jtp.sagepub.comDownloaded from
Rixen and Viola 309
development. Legitimacy, for example, may be a property cultivated by an institution
but it can in tur n affect the development of that institution by increasing the institu-
tion’s attractiveness to investors. An exogenous variable, in contrast, is one whose value
is itself causally independent of the institution in question. Variables are exogenous to
an institution when they are not controlled or determined by that institution, but they
may nevertheless affect its development. Unlike endogenous change, which is based on
feedback, causality in exogenous change is unidirectional. Environmental parameters (as
long as they are independent of the institution in question) are often a source of exoge-
nous change. Technological variation, too, often occurs independently of a particular
institution and yet might be a primary cause of change in that institution.
Some institutional changes may best be explained by a combination of exogenous
and endogenous factors. For example, Greif and Laitin (2004) identify change processes
in which an initially exogenous variable eventually becomes coopted by an institution
and therefore acts endogenously over the long term. Whereas they treat this combination
as a type of variable in its own right (called a quasi-parameter), in our view exogenous
and endogenous variables do not ‘mate’ to form a new hybrid source of change (even
in Greif and Laitin’s account quasi-parameters are not simultaneously endogenous and
exogenous). Rather, any given variable at any given point in time can be either exogenous
or endogenous. Whether a given variable is exogenous or endogenous must be empirically
determined based on what relationship it has to the institution itself. Exogenous and
endogenous variables may both be at work simultaneously and may interact to produce a
change, or a single variable may change from being exogenous to endogenous over time.
Different theoretical approaches favor looking at particular categories of variables
for the sources of institutional change. Rationalists, for example, tend to favor exoge-
nous explanations of change. According to rationalists, an institution is an equilibrium
from which no actor has an incentive to deviate given other actors’ behavior. Logically,
therefore, in this account change can only ever be caused by exogenous factors (Greif
and Laitin, 2004). Moreover, change in this account means moving from being in equi-
librium to being out of equilibrium or from being in one equilibrium to being in another
equilibrium. Historical institutionalists, on the other hand, have argued that this under-
standing of change is too narrow. The institution itself may give rise to dynamic forces
because it affects actors and actor behavior. This in turn implies that internal factors may
change an institution, and this change may be something less than a move out of or to a
new institution. Rather, an institution may become more stable or expand its domain of
activity. Historical institutionalists argue that, in order to capture these types of changes,
we need to consider endogenous variables (Mahoney and Thelen, 2010: 4–7). What nei-
ther of these approaches does well, however, is to recognize that if we can identify two
sources of variables (endogenous and exogenous), then there is the possibility that change
is the result of an interaction of these two. Carefully distinguishing these possibilities can
help us to better tease out and then re-combine the different explanations of institutional
change without blurring important distinctions. Thus we argue that our understanding
of institutional change can be advanced by being explicit about the circumstances under
which endogenous and exogenous variables interact or under which they transform into
one or the other over time.
at Universitaetsbibliothek Bamberg on April 8, 2015jtp.sagepub.comDownloaded from
310 Journal of Theoretical Politics 27(2)
4.2. Categories of mechanisms
Understanding path dependence as characterized by increasing returns, or reinforcing
processes, helps us to distinguish it from processes driven by different causal mecha-
nisms, such as constant returns to scale (enforcement) and decreasing returns to scale
(undermining). We consider these distinctions in turn.
The reinforcement of an institution occurs by means of increasing returns.
12
In terms
of cooperative equilibria, this means that cooperative payoffs increase from one round of
play to the next so that an institutional equilibrium is not simply maintained but amplified.
Thus, reinforcement is a process that increases the stability of an institution by deepening
or expanding the range of the equilibrium/institution over time. This leads some to argue
that path dependence is really about stability and not about change (see, for example,
Greener, 2005; Quack and Djelic, 2007). This is, in our view, a misunderstanding. Path
dependence does not mean ‘no change’, which would simply be a constant, equilibrium
situation. Rather, path dependence captures a particular dynamic of reproduction over
time in which the equilibrium is deepened. In path dependence, what changes is that as
it reproduces itself the institution becomes increasingly stable, locked-in, and resistant to
reversal. So although path dependence does not provide an explanation of path switching
– that is, the movement to a different equilibrium – it does provide an explanation of
change along a path – that is, the increasing entrenchment of an equilibrium.
The enforcement of an institution means that the institution is maintained in stasis. No
actor has an incentive to deviate from the agreed upon behaviors or strategies. If we think
of institutions as cooperative equilibria, then this means that cooperative payoffs remain
the same in each round of play. Enforcement, as opposed to reinforcement, is character-
ized by constant returns to scale; when we change the independent variable by a certain
value, the dependent variable changes by a constant value. Whereas reinforcement leads
to institutional amplification, enforcement describes a mechanism that allows particular
rules or behavior to come to or stay in equilibrium. It refers, then, to institutional creation
– understood both as the move from no institution to an institution and the move from an
existing institution to a new institution – and maintenance.
The undermining of an institution means that the institution is neither being main-
tained nor amplified but is, in fact, crumbling. In terms of cooperative equilibria, this
means that cooperative payoffs decrease from one round of play to the next until even-
tually cooperation (and the institution) ceases. Thus, undermining is characterized by
decreasing returns to scale such that the value of the institution is decreasing relative to
the initial investment. Mechanisms of undermining eventually lead to the breakdown of
an institution.
4.3. Combining causes and mechanisms in a taxonomy of
institutional change
We can now combine these alternative mechanisms of change with our first distinc-
tion between exogenous and endogenous sources of change (see Figure 1).
13
On the
exogenous side, enforcement, reinforcement, and undermining are all the result of vari-
ation in factors not themselves causally affected by the institution. On the endogenous
side, self-enforcement, self-reinforcement, and self-undermining are all dynamics which
at Universitaetsbibliothek Bamberg on April 8, 2015jtp.sagepub.comDownloaded from
Rixen and Viola 311
Figure 1. Taxonomy of Institutional Change Explanations.
affect and are affected by the institution. Each pair of processes – enforce/self-enforce,
reinforce/self-reinforce and undermine/self-undermine – shares a common logic that can
be expressed in terms of returns-to-scale, b ut each pair is separable according to whether
the variables are exogenously or endogenously determined. Below we use the prisoners’
dilemma (PD) game to illustrate the categories in our taxonomy.
In a one-off PD game, there is a unique Nash equilibrium. This equilibrium (mutual
defection) is a self-enforcing outcome since no player has an incentive to deviate from
it. Self-enforcement is the attempt to stabilize an equilibrium from within and can be
considered constant feedback since all information results in a return to a goal state. The
Nash equilibrium in a PD, however, is Pareto inefficient because, according to the pay-off
structure, there is another outcome that would make some players better off without mak-
ing others worse off. To achieve this outcome in a one-off game, the optimal equilibrium
would have to be externally enforced by, to take some typical examples, a Leviathan, the
Mafia or the police.
14
In infinite iteration, the optimal outcome becomes self-enforcing
when players use a tit-for-tat strategy; that is, when infinitely iterated all players have an
incentive to cooperate and no player has an incentive to defect. Note, however, that the
iterated PD is not self-reinforcing. This is because in a classic iterated PD, the pay-off in
every round is the same as in the previous round. Cooperation is maintained not because
of increasing returns, but because constant return pay-offs accumulate over time.
A PD which is only finitely iterated, in contrast, would be characterized by self-
undermining. That is, even if cooperation were possible, the possibility of defection in the
last round of play converts all previous strategies into defection thus unravelling coopera-
tion. This is self-undermining because the impetus for the unraveling comes from within
at Universitaetsbibliothek Bamberg on April 8, 2015jtp.sagepub.comDownloaded from
312 Journal of Theoretical Politics 27(2)
the logic of the game (or institution) itself. In order to prevent self-undermining and
maintain mutual cooperation under conditions of finite play, an external enforcer would
be necessary. Some institutions might be purposefully self-undermining, like the March
of Dimes. The March of Dimes was established to eradicate polio, and when the disease
was in fact eradicated the March of Dimes logically faced collapse.
15
An eradication
mandate can thus be a source of institutional self-undermining. Another possible source
of self-undermining is institutional overstretch. When an institution’s original mandate
is expanded and it takes on additional tasks that it cannot fulfill, the entire institution
might suffer. Snyder, for example, argues that this is one cause of the decline of empires
(Snyder, 1991).
In contrast to self-undermining, undermining occurs when an external factor desta-
bilizes an existing equilibrium. This factor, either material or normative, gives players
incentives to deviate from their agreed upon behavior. The ‘live-and-let-live’ approach
that spontaneously broke out among WWI trench soldiers, inducing a cooperative cease-
fire, was broken by officers who forced their soldiers into battle by demanding to see
either prisoners or casualties (Axelrod, 1984). Slavery and apartheid, it has been argued,
are institutions that have been undermined by society’s changing normative beliefs
(Klotz, 1995). Other tactics which attempt to change the payoffs of cooperating actors,
such as shaming or boycotting, can have similar institution-undermining effects.
The final combination to be discussed is reinforcement/self-reinforcement. Both of
these are characterized by increasing returns. An institution might be reinforced, for
example, through the payment of subsidies or bonuses which reward cooperation by
increasing payoffs in future rounds conditional on past success. In self-reinforcing pro-
cesses, in contrast, the increasing returns have an endogenous source. Self-reinforcement
is, as established in the previous section, what is commonly referred to as path depen-
dence. Endogenous factors which can lead to increasing returns include sunk costs,
learning effects, and coordination effects.
Although we have used the prisoners’ dilemma to illustrate the range of change expla-
nations, it is important to note that this taxonomy does not prefer and is not prejudiced
against any theoretical predisposition. The taxonomy can be used by any approach, for
example, those that Mahoney discusses under the labels of ‘utilitarian’, ‘functional’,
‘power’ and ‘legitimation’ approaches (Mahoney, 2000: 517–526). The taxonomy can,
for example, accommodate traditional rational actor accounts that argue change is the
result of exogenously determined actor preferences, but also constructivist accounts that
argue preferences are endogenous to an institution. This is because the taxonomy is not
itself an explanation but a logical exercise that orders explanatory concepts and makes
explicit which explanatory building blocks are available.
At least three significant conclusions can be read from the taxonomy. First, path
dependence is clearly only one of a number of endogenous explanations of institutional
change. Second, it is also clear that not all explanations that rely on increasing returns are
also path dependent. Third, although many path dependence explanations rely on exoge-
nous factors or non-increasing returns to scale for explanatory power, s trictly speaking
these should be seen as additions that stretch the concept of path dependence. These
additions may be necessary to convincingly explain change, but rather than making room
for them within path dependence we should instead theorize the multiple ways in which
different mechanisms and processes interact to explain change.
at Universitaetsbibliothek Bamberg on April 8, 2015jtp.sagepub.comDownloaded from
Rixen and Viola 313
5. Applying the taxonomy to the path dependence literature
With the taxonomy in hand, we now have a tool for assessing the path dependence litera-
ture. We identify two types of conceptual stretching that risk obscuring a wider range of
institutional change explanations. First, there are those who agree that the core character-
istics of path dependence are endogeneity and increasing returns, but who then effectively
take these characteristics to be ubiquitously present. Second, there are those who argue
that the core attributes of path dependence cannot be identified in the phenomenon to be
explained and therefore respond by extending path dependence to include also constant
and decreasing returns, and/or exogenous causal factors. In both instances, the desire to
apply path dependence to an ever wider range of institutional change cases tends to result
in stretching.
In this section, we purposefully focus on major contributions to the path dependence
literature that clearly aim to rigorously employ the concept of path dependence, as this
is where stretching has the strongest implications.
16
Our goal is not to show that those
contributions that rely on an extended rather than parsimonious understanding of path
dependence are ‘wrong’, but rather to identify instances in which the use of path depen-
dence subsumes and obscures distinct causal processes that are important in their own
right. Our argument is that theory building would benefit from recognizing when distinct
explanations are at work so that we can theorize those processes and hypothesize how
these distinct processes interact with path dependence.
5.1. The two attributes of path dependence: a broad interpretation
Two landmark contributions to the study of path dependence have come from Douglass
North and Paul Pierson. North’s theory of institutional development in economic history
introduced the concept to the study of institutions, while Pierson’s work was instrumen-
tal i n introducing path dependence to political science. Both of these contributions to
the theoretical debate accept increasing returns and endogeneity as defining attrib utes
of path dependence. In broadening the empirical application of path dependence from
issues of technological development to institutions – that is, concept traveling – North
and Pierson make arguments for why institutions are particularly likely to be subject to
path dependence. In their search for more variables that are endogenous to institutions
and more reasons why institutions exhibit increasing returns, they come close to arguing
that institutions are per se path dependent. But the extensions proposed by North and
Pierson can only be subsumed under path dependence if they can be shown to fulfill the
two conditions of endogeneity and increasing returns. In the following we show that this
is not always the case.
North’s substantive interest is in providing an explanation for the long-term success
of some economies and the lack of development of others. North i s interested in national
economies and he takes the individual institutions and organizations that make up an
economy as endogenous to this unit of analysis. He then develops the idea that different
institutions within a society will complement each other, and these complementarities
have coordination effects that produce increasing returns: ‘the interdependent web of an
institutional matrix produces massive increasing returns’ (North, 1990: 95). In addition,
actors are only limitedly rational and seek to minimize transaction costs. Institutional
at Universitaetsbibliothek Bamberg on April 8, 2015jtp.sagepub.comDownloaded from
314 Journal of Theoretical Politics 27(2)
structures enable actors to deal with the problems of bounded rationality and transaction
costs. Since actors get more adept at using institutions over time, national economies
exhibit increasing returns and are thus likely to remain on their respective paths (North,
1990; North, 1981).
So far, these arguments imply that institutions always reproduce themselves and leave
no scope for change. But since North acknowledges that institutions may change, he
makes room for this possibility by arguing that actors are not only boundedly rational
but also innovators and lear ners. As a consequence, there will permanently be efforts at
institutional creation and change driven by actors who stand to gain from a change in the
institutional structure (North, 1990). Nevertheless, institutional creation and change will
be incremental and bounded rather than rapid and radical because actors interested in
minimizing transaction costs will be sensitive to start-up and fixed costs and their efforts
will be slowed down by the increasing returns dynamics of existing institutions (North,
1990).
This account has two implications which strike us as problematic. First, North sees
path dependence everywhere. Because bounded rationality, the desire to minimize trans-
action costs, and the interdependence and complementarity of institutions are, in North’s
account, general features of the social world, all i nstitutions must be subject to path
dependence all the time (North, 1990: 101). While these factors are certainly important,
it is doubtful that the minimization of transaction costs will always be actors’ overriding
interest or that this motivation will be strong enough to create increasing returns. In addi-
tion, as we will discuss below, i t is not necessarily the case that institutions will always
complement each other. Rather than showing that these factors are present in the specific
case analyzed, North assumes them to be ubiquitous. Consequently, path dependence
becomes a general feature of all institutions and is not seen as a particular explanation for
specific institutional developments. Significantly, this move means that we can no longer
discriminate between path dependence and other processes of institutional change.
Second, North gives up the notion that path dependence is about the reproduction and
deepening of the same institutional outcome. Instead he subsumes gr adual, bounded or
incremental change under the concept of path dependence. However, on the basis of our
taxonomy it becomes clear that his account of incremental and bounded change cannot
be fully subsumed under path dependence. This is because in his model those agents
responsible for change are not sources of increasing returns. One may accept that actors’
bounded rationality, their desire to minimize transaction costs, and even actors’ interests
in changing an institutional structure are endogenous, since North’s unit of analysis is
an economy in its entirety. One may also accept that the institutional structure exhibits
increasing returns because it enables actors to minimize transaction costs and deal with
problems of bounded rationality. But it cannot be argued that actors interested in changing
the institutional structure are sources of increasing returns. I n contrast, since they want to
undermine or change the current institution, they may be sources of decreasing returns.
This is not to deny that North presents a plausible account of institutional change,
but it would have to be retold in a way that recaptures the variety of processes at work:
the desire of actors to change the institutional structure to further their interests is an
endogenous undermining force. This undermining force, however, is counterbalanced by
an endogenous, self-reinforcing (i.e. path dependent) process of coping with transac-
tion costs and bounded rationality. Thus, without changing the substantive claims of his
at Universitaetsbibliothek Bamberg on April 8, 2015jtp.sagepub.comDownloaded from
Rixen and Viola 315
account, our taxonomy helps to disentangle the different processes that interact to pro-
duce incremental and bounded change, whereas North s ubsumed all of them under path
dependence.
In contrast to North, Pierson proposes a more specific set of factors that make path
dependence likely. He argues that the concept of path dependence should be even more
relevant in political science than it is in economics (Pierson, 2004), due to additional
sources of increasing returns in the political sphere. Four factors of political life are
relevant.
First, the central role of collective action and collective action problems makes insti-
tutional reforms less likely than in the more competitive and thus flexible environment of
a market. Once actors have overcome a collective action problem to form an institution,
the operation of the institution will be subject to decreasing costs per unit of the (public)
good produced. Second, the high density of political institutions is a source of increas-
ing returns, if the institutions complement each other. Third, political authority and power
asymmetries can be sources of positive feedback. Actors may use their power positions to
change the rules of the game in a way that further enhances their power positions. Fourth,
another source of positive feedback can lie in the complexity and opacity of politics. In
contrast to the economic sphere, where success can be measured in terms of monetary
units, such a coherent and easily observable indicator of success does not exist in the
political sphere. This makes lear ning through trial-and-error more difficult, which makes
it more likely that actors’ shared understandings will be reinforced over time rather than
changed (Pierson, 2004).
But even though each of these four factors can potentially result in path dependence,
there is no a priori reason to believe that they are always endogenous and always result
in increasing returns. Pierson’s first argument about collective action problems is that
solving them involves high fixed or sunk costs and therefore institutions are subject to
increasing returns. This is consistent with our taxonomy as long as the specific collective
action is also driven by endogenous factors. Note, however, that the level of fixed or
sunk costs will vary with the nature of the collective action problem, and so increasing
returns cannot be assumed across the board. An institution that is designed to deal with
assurance problems, for example, should be cheaper to set up than one that deals with
prisoners’ dilemma problems. Also, as Schwartz points out, even if institutions have high
setup costs this does not mean that there will be decreasing unit costs (i.e. increasing
returns) for all of its activities. For example, if an old organization starts to mobilize for
a new cause, the unit costs may well be constant (Schwartz, 2004).
Pierson’s second argument about the high density of institutions is similar to North’s
argument about institutional complementarities. This is consistent with path dependence
for those cases in which high density does in fact lead to increasing returns. It is impor-
tant to note, however, that a high density of institutions does not necessarily lead to
complementarities. We may just as well observe competition among institutions, which
may involve decreasing or constant returns for the competing institutions. Thus, in the
specific case analyzed, it will have to be established that institutions actually complement
each other and benefit from coordination effects. Which one of the two, complementari-
ties or competition, is more likely in a dense institutional field cannot be determined ex
ante.
at Universitaetsbibliothek Bamberg on April 8, 2015jtp.sagepub.comDownloaded from
316 Journal of Theoretical Politics 27(2)
It is more difficult to see why, as Pierson’s third point contends, power asymmetries
are likely to be endogenous and create increasing returns. First, power asymmetries may
or may not be endogenous to an institution. If the power of certain actors is derived from
the rules of the institution (e.g. voting shares), then this would be the case. If, however, it
is derived from sources external to the institution (e.g. private wealth, personal network,
etc.) then power asymmetries are exogenous (cf. Beyer, 2010). Second, even if they are
endogenous, power asymmetries are not always sources of self-reinforcement,butmay
also be sources of self-undermining. This is apparent in Pierson’s own illustration: actors
may use their power positions to change the institution in a way that further enhances
their power but undermines the stability of the institution (see also Mahoney, 2000: 523;
Pierson, 2004).
Finally, Pierson’s fourth point about the opacity of politics leading to positive feed-
back is also ambiguous. One may indeed argue that shared understandings or social
norms are endogenously reinforced, because by applying norms actors also constitute
and sustain them. And it may also be argued that applying those shared understandings
leads to increasing returns. However, Pierson combines this idea with an argument about
the difficulties of learning from bad political outcomes and subsumes both under path
dependence. On the basis of our taxonomy, however, we can see that these are actually
two distinct processes: the (endogenous) path dependent development of norms and an
endogenous undermining process of learning. Whether these opposing processes will
lead to institutional stability or to change is an empirical outcome that depends on the
strength of the respective processes.
As both North and Pierson argue, there are good reasons to believe that institutions
may be subject to path dependent processes. It does not logically follow, however, that
all institutional change processes are therefore path dependent. In order to determine
which processes are actually path dependent, we need to identify whether the causal
variables at work are endogenous and whether the mechanism relating independent and
dependent variables is characterized by increasing returns. As the discussion has shown,
this often cannot be identified ex ante but has to be shown empirically in the specific
case analyzed. When the two conditions are not met, then institutional change is indeed
caused by another process or the result of a combination of processes.
5.2. Adding attributes to path dependence
Another kind of stretching occurs when additional attributes are subsumed under the
concept of path dependence. This tends to happen when researchers identify variables
or mechanisms beyond the t wo core attributes of path dependence to be influential in
institutional change. Rather than implicitly subsuming these additional factors under the
two core attributes of path dependence, as discussed in the previous section, these studies
respond by explicitly extending the concept. We first discuss examples in which path
dependence is extended to include the full range of causal variables (i.e. endogenous
and exogenous factors), and then turn to cases in which path dependence is extended to
include a range of causal mechanisms (i.e. constant and decreasing returns).
An important instance of exogenous causal factors being added to the concept of path
dependence occurs in the critical junctures debate. A critical juncture refers to “relatively
at Universitaetsbibliothek Bamberg on April 8, 2015jtp.sagepub.comDownloaded from
Rixen and Viola 317
short periods of time during which there is a substantially heightened probability that
agents’ choices will affect the outcome of interest,” because in these exceptional situa-
tions structural constraints are loosened (Capoccia and Kelemen, 2007: 348). Examples
of critical junctures are wars, revolutions, or natural disasters. While most historical insti-
tutionalists acknowledge that such events are typically exogenous to the i nstitution in
question (Hall and Taylor, 1996: 942), many of them integrate critical junctures into the
concept of path dependence.
17
For example, exogenous economic crises play an impor-
tant role in arguments about the institutional development of party-political systems
(Thelen, 1999: 392). Equally, exogenous shocks are seen as key elements in a ‘functional
explanation’ of path dependence (Mahoney, 2000: 519–522). For these authors, critical
junctures are the founding moments of path dependent institutional developments and
capture the initial contingency of the process (see, for example, Mahoney, 2000; Thelen,
1999).
According to our taxonomy, however, a critical juncture cannot be both exogenous
to the institution analyzed and also be subsumed under path dependence. While expla-
nations based on combinations of exogenous critical junctures and (endogenous) path
dependence processes are empirically plausible, the taxonomy reminds us that critical
junctures as such are not part of the concept of path dependence. Instead of adding criti-
cal junctures to path dependence explanations in order to allow path dependence to work
in a way that it would not on its own,
18
theory building on institutional change should
observe the conceptual distinction between endogenous path dependence and exogenous
critical junctures. It could then more rigorously theorize combinations of these causal
factors.
The second kind of explicit stretching is the attempt to add decreasing or constant
returns to path dependence. Proponents of this strategy often follow the same impulse as
Douglass North; they aim at reconciling the notion of path dependence with the empirical
fact that institutional change occurs (Greener, 2005; see also Thelen, 2003). However, in
contrast to North, who accepted increasing returns as an attribute but then smuggled in
decreasing returns, others explicitly argue that increasing, decreasing, and even constant
returns ought to be considered attributes of path dependence. One example is Mahoney’s
concept of a ‘path dependent reactive sequence’ (Mahoney, 2000: 527). In contrast to
self-reinforcement, reactive sequences are marked by ‘backlash processes that transform
and perhaps reverse early events’ (Mahoney, 2000: 526, emphasis in the original). In
other words, Mahoney includes decreasing returns as a possible characteristic of the con-
cept of path dependence.
19
While there can be no doubt that such undermining processes
exist and should be analyzed to understand institutional change, there is, in our view, a
real risk in subsuming such processes under the notion of path dependence. Stretching
path dependence to explain both lock-in and reversal, for example, expands the number
of cases the concept covers b ut limits its analytical power.
20
Finally, it is worth noting that some scholars have reacted to the inability of path
dependence to explain more instances of institutional change not by stretching the con-
cept but by abandoning it. Consider, for example, Thelen’s arguments about incremental
change along ‘historical trajectories’. She is careful not to label these trajectories path
dependent, but her term ultimately attempts to explain similar phenomena. Thelen wants
to reconcile the notion of institutional stability with the fact that institutions also undergo
at Universitaetsbibliothek Bamberg on April 8, 2015jtp.sagepub.comDownloaded from
318 Journal of Theoretical Politics 27(2)
gradual change (Thelen, 2003). Thus, she proposes to focus on mechanisms of institu-
tional ‘reproduction’ and ‘undermining’ rather than path dependence (Thelen, 1999). The
distinction between reproduction and undermining is in line with ours between decreas-
ing and increasing returns (but neglects constant returns). However, Thelen does not
distinguish endogenous from exogenous causal factors. Consequently, she only offers two
types of explanations, whereas we distinguish six different types. Moreover, it strikes us
as undesirable that this (implied) typology intentionally sacrifices the thoroughly theo-
rized concept of path dependence. Our taxonomy reserves a clearly delimited space for
path dependence while also recognizing other institutional change processes as distinct
concepts.
The discussion that has evolved around the concept of path dependence within the
social sciences has been highly productive in that it has motivated scholars to develop
theoretical accounts of institutional change. The problem with some of these accounts,
however, is that they either implicitly or explicitly stretch the concept to accommodate
empirical observations rather than to argue that explaining empirical complexity may
require a combination of different mechanisms and processes that need not be [bun-
dled] into one concept.
21
This conclusion in no way diminishes these contributions to
our understanding of institutional change. In fact, it is highly plausible that explain-
ing institutional change will often involve a combination of distinct processes. Rather,
what we argue is that this explanatory richness can be made more fruitful when expla-
nations are combined in a theoretically-conscious and not ad hoc fashion. With our
taxonomy in hand, we can begin to generate hypotheses about how different explanations
of institutional change might work together and with what effects.
6. Conclusions
Recent literature has observed that despite the institutional turn we still do not have strong
theories of institutional change. In this article, we have suggested that theory develop-
ment has been hindered by conceptual challenges that arise as a field matures; namely,
existing concepts get stretched and a range of new concepts get introduced. To remedy
these problems, we constructed a taxonomy that captures a logically complete palette
of institutional change explanations and offers a basis for delimiting them from one
another. We built the taxonomy by taking the well-developed theory of path dependence
as our starting point for identifying a distinct set of change variables and mechanisms. We
argued that path dependence has two core attributes; it is characterized by endogeneity
and increasing r eturns. Then, using the technique of ‘negative identification’, we inferred
a logically complete set of causal variables and mechanisms (see Figure 1).
The value of this conceptual exercise is that it provides a tool for systematically
theorizing institutional change. First, it provides a remedy for the problem of concept
stretching. In particular, it strengthens the analytic value of path dependence by clearly
delimiting its scope. It is important to address concept stretching because, unheeded, it
increases the likelihood that cases of path dependence become over-diagnosed and other
concepts of change get smuggled in haphazardly or remain obscured altogether. Sec-
ond, it distinguishes among existing concepts and mechanisms so that they can be more
reflectively combined. This in turn should help scholars identify multiple or configurative
at Universitaetsbibliothek Bamberg on April 8, 2015jtp.sagepub.comDownloaded from
Rixen and Viola 319
causation in the explanation of particular phenomena, developing hypotheses about the
relationships among various causal factors and mechanisms. Finally, it helps us to iden-
tify logically missing concepts and mechanisms of institutional change. Interestingly, the
path dependence branch of the taxonomy tree is the most and best theorized. Other cells
have been used ‘a la carte’ but systematic theorization of alternatives is as of yet lacking.
Developing these cells and developing hypotheses for when and how various branches
interact are tasks for future research.
The taxonomy also sheds light on three particular issues that trouble the institu-
tional change literature: endogeneity versus exogeneity; temporality; and non-change.
First, although path dependence is widely understood to be a self-reinforcing process, the
literature has paid much more attention to the increasing returns that characterize ‘rein-
forcement’ than to the endogeneity that characterizes ‘self-reinforcement’. The power of
the path dependence logic, however, rests at least in part on the insight that institutional
change can be endogenous. When current theories neglect to distinguish or unwittingly
mingle endogenous and exogenous factors, we lose analytical rigor. It is certainly the case
that endogenous and exogenous factors work together in reality. But to gain a coherent
understanding of how these factors interact, we first need to reflect on which ones are at
work when and where.
Second, we have explicitly built the taxonomy around possible explanations of insti-
tutional change, including types of causes and mechanisms, rather than descriptions of
change, which focus on the pace of change (e.g. whether it is gradual or punctuated) or
the scope of change (e.g. whether it is broad or deep). Nevertheless, it may well be that
particular causes of change (e.g. exogenous or endogenous) and particular mechanisms
of change (e.g. constant, increasing or decreasing returns) influence the pace or scope of
change. Endogenous change, for instance, may be correlated with gradual change. These
relationships need to be systematically worked out in future work.
Finally, thinking about path dependence as one explanation of institutional change
reveals an important difference between the notions of no change, change, and stability.
As we have discussed, some interpret path dependence to be a concept of non-change. But
in our view this conflates stability and stasis. Stability itself is a property of institutions
that is subject to change, even though a highly stable institution might appear to be static.
Our discussion suggests, then, that path dependence does indeed capture change – an
increase in the stability of an institutional equilibrium (i.e., its robustness to shocks grows
over time).
In systematizing various types of institutional change explanations, our hope is that
future work can more effectively address outstanding puzzles of institutional change,
including those outlined above, by self-consciously theorizing the interaction of distinct
variables and mechanisms.
Acknowledgements
For their helpful comments on previous drafts of this article we thank Sebastian Botzem, Matthias
Ecker-Erhardt, Tim Gemkow, Philipp Genschel, Johannes Gerschewski, Tine Hanrieder, Joaquín
Herranz, Jacint Jordana, Jon Pevehouse, Justin Powell, Georg Schreyögg, Jörg Sydow, Kathleen
Thelen and Michael Zürn. For research assistance we thank Mary Kelley-Bibra and Xaver Keller.
at Universitaetsbibliothek Bamberg on April 8, 2015jtp.sagepub.comDownloaded from
320 Journal of Theoretical Politics 27(2)
We also thank the Social Science Center Berlin (WZB), where this research was completed, for its
support.
Funding
This research received no specific grant from any funding agency in the public, commercial or
not-for-profit sectors.
Notes
1. A few other authors, most importantly Scott Page and collaborators (Page, 2006; Bednar et
al., 2012), have also attempted to differentiate path dependence from other kinds of processes.
While their endeavor shares our ambition to clarify the concept in order to preserve its explana-
tory power, it differs from ours in that it (1) aims to expand the explanatory power of path
dependence and (2) focuses only on processes in which histor y matters (in different ways).
In contrast, our taxonomy is an effort to sketch both path dependence and the alternative
explanations it implies.
2.
Another strategy for systematizing concepts is to integrate them into a coherent general theory.
This has been the direction of some recent contributions to the field such as Mahoney and
Thelen (2010).
3. As the number of observations that fit into a category increases (high extension), the broader
the meaning of that category becomes (low intension). Categories with high intension bring
high discriminatory and thus explanatory power, especially within cases; categories with high
extension bring generalizability and explanatory power across cases.
4.
As we will discuss in more detail later, whether a variable is endogenous or exogenous depends
on its position relative to the institution in question; i.e., no variable is inherently exogenous or
endogenous.
5.
While many scholars believe that inefficiency is frequent in path dependent processes (but see
Liebowitz and Margolis, 1990; 1995), there is consensus that inefficiency is not a necessary
effect (David, 2007).
6. This is in contrast to equilibrium selection under constant or decreasing returns. In such ergodic
dynamics, competition can drive out other results. They have a unique equilibrium which,
however, cannot be locked in. Early choices can be corrected, reducing the importance of early
events.
7.
Some of the confusion that has surrounded the application of path dependence is, we maintain,
the result of not properly distinguishing between causes and effects of path dependence.
8.
Page (2006: 88) claims that path dependence has four related causes: increasing returns, self-
reinforcement, positive feedback and lock-in. He does however state himself that positive
feedback is essentially the same as increasing returns, and it is apparent from his definition
of self-reinforcement that it can be usefully understood as increasing returns plus endogeneity.
In contrast to Page, we see lock-in as an effect rather than a cause of path dependence. Over-
all, Page’s discussion underlines our claim that increasing returns and endogeneity are the two
causal attributes of path dependence.
9.
But see Schwartz (2004: 5), who writes of path dependence’s ‘powerful, endogenous increasing
returns mechanisms’.
10. For a critique of this move see (Schwartz, 2004).
11. For such a classification see (Streeck and Thelen, 2005). As (Mahoney and Thelen, 2010), point
out, these are types or ‘modes’ of change, not explanations of it.
12. It has been argued that both sequential returns and increasing returns can be self-reinforcing
(Hathaway, 2001). In our view, however, it is incorrect to distinguish between sequential and
at Universitaetsbibliothek Bamberg on April 8, 2015jtp.sagepub.comDownloaded from
Rixen and Viola 321
increasing returns. S equencing can be better understood as a way to get increasing returns.
As the QWERTY example shows, a cer t ain sequence of events can produce increasing returns
which would not have been present had the sequence been different (David, 1985).
13. Although we use a taxonomical tree, the typology could also be depicted as a 2x3-matrix, as in
Elman (2005).
14. These traditional examples are of negative enforcement. In fact, external enforcement can be
the result of increasing the costs of defecting (which is how the given examples work), or it can
be the result of increasing the benefits of cooperating (such as through side payments).
15. In fact, rather than collapse, the March of Dimes changed its mandate to one of fighting all
childhood diseases.
16. We do not address studies that use a ‘soft version’ of path dependence, which simply asserts
that history matters (Quack and Djelic, 2007). Such studies have been identified as problematic
by many scholars, see for example Mahoney (2000), Pierson (2004) and Thelen (1999).
17. This understanding of critical junctures has also been criticized by Capoccia and Kelemen
(2007), who argue that the notion of ‘critical juncture’ is an important mechanism in its own
right.
18. That critical junctures serve this function in historical institutionalism is generally acknowl-
edged, see for example Schwartz (2004) and Thelen (2003). What is mostly not acknowledged
is that path dependence has to be conceptually separated from critical junctures.
19. His argument for why a reactive sequence can be subsumed under path dependence is that it is
also marked by initial contingency (Mahoney, 2000).
20. Put more strongly, ‘a concept that attributes the same outcome sometimes to increasing and
sometimes to decreasing retur ns is incoherent’ (Schwartz, 2004: 11). On this point, see also
Drezner (2010)
21. Pierson (2004: 21) makes a similar point with respect to the class of sequencing explanations:
‘Limiting the concept of path dependence to self-reinforcing processes in no way precludes the
investigation of other ways in which sequences can matter in explaining social outcomes’.
References
Arthur W B (1994) Increasing Returns and Path Dependence in the Economy. Ann Arbor, MI:
University of Michigan Press.
Axelrod R M (1984) The Evolution of Cooperation. New York, NY: Basic Books.
Bednar J, Page S, Toole J (2012) Revised-path dependence. Political Analysis 20(2): 146–156.
Bennett A and Elman C (2006) Complex causal relations and case study methods: the example of
path dependence. Political Analysis 14(3): 250–267.
Beyer J (2010) The same or not the same – on the variety of mechanisms of path dependence.
International Journal of Social Sciences 5(1): 1–11.
Boas T C (2007) Conceptualizing continuity and change: The composite–standard model of path
dependence. Journal of Theoretical Politics 19(1): 33–54.
Büthe T (2002) Taking temporality seriously: modeling history and the use of narratives as
evidence. American Political Science Review 96(3): 481–494.
Capoccia G and Kelemen R D (2007) The study of critical junctures: theory, narrative and
counterfactuals in institutional theory. World Politics 59(3): 341–369.
Collier D and Mahon J E (1993) Conceptual ‘stretching’ revisited: adapting categories in
comparative analysis. American Political Science Review 87(4): 845–855.
David P A (2007) Path dependence: a foundational concept for historical social science. Cliomet-
rica 1(2): 91–114.
David P A (1985) Clio and the economics of qwerty. American Economic Review 75(2): 332–337.
at Universitaetsbibliothek Bamberg on April 8, 2015jtp.sagepub.comDownloaded from
322 Journal of Theoretical Politics 27(2)
Drezner DW (2010) Is historical institutionalism bunk? Review of International Political
Economy 17(4): 791–804.
Elman C (2005) Explanatory typologies in qualitative studies of international politics. International
Organization 59(2): 293–326.
Gerring J (1999) What makes a concept good? A criterial framework for understanding concept
formation in the social sciences. Polity 31(3): 357–393.
Greener I (2005) The potential of path dependence in political studies. Politics 25(1): 62–72.
Greif A and Laitin D (2004) A theory of endogenous institutional change. American Political
Science Review 98(4): 633–652.
Hall P A and Taylor R (1996) Political science and the three new institutionalisms. Political Studies
44(2): 936–657.
Hathaway O (2001) Path dependence in the law: the course and pattern of legal change in a common
law system. The Iowa Law Review 86(2): 101–165.
Klotz A (1995) Norms in International Relations. The Struggle against Apartheid. Ithaca, NY:
Cornell University Press.
Krasner S D (1982) Structural causes and regime consequences: regimes as intervening variables.
International Organization 36(2): 185–205.
Liebowitz J S and Margolis E S (1990) The fable of the keys. Journal of Law and Economics 30(1):
1–26.
Liebowitz S J and Margolis S E (1995) Path dependence, lock-in, and history. Journal of Law,
Economics, and Organization 11(1): 205–226.
Mahoney J (2000) Path dependence in historical sociology. Theory and Society 29(4): 507–548.
Mahoney J and Schensul D (2006) Historical context and path dependence. In: Goodin RE and
Tilly C (eds) The Oxford Handbook of Contextual Political Analysis. Oxford: Oxford University
Press, pp. 454–471.
Mahoney J and Thelen K (2010) A theory of gradual institutional change. In: Mahoney J and Thelen
K(eds)Explaining Institutional Change. Ambiguity, Agency, and Power. Cambridge: Cambridge
University Press, 1–37.
North D C (1990) Institutions, Institutional Change, and Economic Performance.NewYork,NY:
Cambridge University Press.
North D C (1981) Structure and Change in Economic History, first edition. New York, NY: Norton.
Page S (2006) Path Dependence. Quarterly Journal of Political Science 1(1): 87–115.
Pierson P (2004) Politics in Time. History, Institutions, and Social Analysis. Princeton, NJ:
Princeton University Press.
Pierson P (2000) The limits of design: explaining institutional origins and change. Governance
13(4): 475–499.
Pierson P (1996) The new politics of the welfare state. World Politics 48(2): 143–179.
Quack S and Djelic M L (2007) Overcoming path dependencies – path generation in open systems.
Theory and Society 36(2): 161–186.
Sartori G (1970) Concept misformation in comparative politics. American Political Science Review
64(4): 1033–1053.
Schwartz H (2004) Down the wrong path: path dependence, increasing returns, and historical
institutionalism. Unpublished manuscript, University of Virginia, Charlottesville, VA.
Sewell W H (1996) Three temporalities: toward an eventful sociology. In: McDonald T (ed.)
The Historic Turn in the Human Sciences. Ann Arbor, MI: University of Michigan Press,
pp. 245–280.
Snyder J L (1991) Myths of Empire: Domestic Politics and International Ambition. Ithaca, NY:
Cornell University Press.
Stinchcombe A L (1968) Constructing Social Theories. New York, NY: Harcourt, Brace and World.
at Universitaetsbibliothek Bamberg on April 8, 2015jtp.sagepub.comDownloaded from
Rixen and Viola 323
Streeck W and Thelen K (2005) Institutional change in advanced political economics. In: Streeck W
and Thelen K (eds) Beyond Continuity. Institutional Change in Advanced Political Economics,
Oxford: Oxford University Press, pp. 1–39.
Thelen K (1999) Historical institutionalism in comparative politics. Annual Review of Political
Science 2: 369–404.
Thelen K (2003) How institutions evolve: insights from comparative historical analysis. In:
Mahoney J and Rueschemeyer D (eds) Comparative Historical Analysis in the Social.
at Universitaetsbibliothek Bamberg on April 8, 2015jtp.sagepub.comDownloaded from
