Chapter 1
Process-Tracing in the Social Sciences
You know a conjurer gets no credit when once he has explained his
trick; and if I show you too much of my method of working, you will
come to the conclusion that I am a very ordinary individual after all.
—Sherlock Holmes (A. C. Doyle 2010: 33)
The essence of process-tracing research is that scholars want to go beyond
merely identifying correlations between independent variables (Xs) and outcomes
(Ys). For example, a strong statistical correlation has been found between
democracy and peace (Oneal, Russett, and Berbaum 2004). Yet how
do we know that mutual democracy was the cause of peace between two
nations? How does democracy produce more peaceful relations? Answering
these questions requires that we unpack the causal relationship between
mutual democracy and peace to study the causal mechanism linking the two
concepts.
Process-tracing in social science is commonly defined by its ambition to
trace causal mechanisms (Bennett 2008a, 2008b; Checkel 2008; George and
Bennett 2005). A causal mechanism can be defined as “a complex system,
which produces an outcome by the interaction of a number of parts” (Glennan
1996: 52). Process-tracing involves “attempts to identify the intervening
causal process—the causal chain and causal mechanism—between an independent
variable (or variables) and the outcome of the dependent variable”
(George and Bennett 2005: 206–7).
Investigating causal mechanisms enables us to go a step further when
studying causal relationships, allowing us to “peer into the box of causality
to locate the intermediate factors lying between some structural cause and its
purported effect” (Gerring 2007a: 45). Yet process-tracing methods are argu-
2 Process-Tracing Methods
ably the only method that allows us to study causal mechanisms. Studying
causal mechanisms with process-tracing methods enables the researcher to
make strong within-case inferences about the causal process whereby outcomes
are produced, enabling us to update the degree of confidence we hold
in the validity of a theorized causal mechanism. Process-tracing therefore
represents “an invaluable method that should be included in every researcher’s
repertoire” (George and Bennett 2005: 224).
Process-tracing methods have recently experienced a surge in popularity
within qualitative social science, with numerous doctoral students and
established scholars attempting to use process-tracing methods in their research
(e.g., Bennett and Elman 2006a, 2006b; Elman 2004; Hall 2008; Jacobs
2004; Khong 1992; Lehtonen 2008; Owen 1994). Yet despite the widespread
use of process-tracing in empirical research and an increasing body
of methodological literature on process-tracing and causal mechanisms, we
still do not possess a clear and coherent framework for how and when valid
inferences can be made using process-tracing. We also lack a set of concrete
guidelines for using the methods in practice. This deficiency has prevented
process-tracing from fulfilling its potential of enabling us to open up the
black box of causality using in-depth case study methods to make strong
within-case inferences about causal mechanisms.
In this book, we seek to reveal how the trick is performed. In so doing,
we show readers that process-tracing is an “ordinary” social science method,
like many others, with comparative strengths and weaknesses. It is not a
panacea, but when applied in appropriate research situations, it can enable
us to make strong within-case causal inferences about causal mechanisms
based on in-depth single-case studies that are arguably not possible with
other social science methods.
1.1. Defining Process-Tracing
Process-tracing methods are tools to study causal mechanisms in a singlecase
research design. While scholars generally agree that process-tracing
methods can be defined by their ambition to trace causal mechanisms, the
existing literature retains considerable confusion about both the ontological
and epistemological foundations of process-tracing methods and guidelines
for what good process-tracing entails in practice. Basic questions such
as what types of causal mechanisms are being traced and to what degree
Process-Tracing in the Social Sciences 3
process-tracing case studies can be nested in broader, mixed-method research
designs have been left relatively unanswered. The resulting lack of
coherent foundations and concrete guidelines has prevented the method
from fulfilling its potential.
This confusion results partly from the literature’s definition of processtracing
as a single research method. A lot of the murkiness about what
process-tracing is and how it should be used in practice can be cleared up
by differentiating process-tracing into three variants within social science:
theory-testing, theory-building, and explaining-outcome. The three differ
along several dimensions, including whether they are theory- or case-centric,
the types of inferences being made, how they understand causal mechanisms,
and whether and how they can be nested in mixed-method designs.
Theory-testing process-tracing deduces a theory from the existing literature
and then tests whether evidence shows that each part of a hypothesized
causal mechanism is present in a given case, enabling within-case inferences
about whether the mechanism functioned as expected in the case and
whether the mechanism as a whole was present. No claims can be made,
however, about whether the mechanism was the only cause of the outcome.
Theory-building process-tracing seeks to build a generalizable theoretical
explanation from empirical evidence, inferring that a more general causal
mechanism exists from the facts of a particular case. Although this type of
process-tracing is analytically useful, to our knowledge, the literature offers
no guidelines about how to proceed with this approach.
Finally, explaining-outcome process-tracing attempts to craft a minimally
sufficient explanation of a puzzling outcome in a specific historical
case. Here the aim is not to build or test more general theories but to craft
a (minimally) sufficient explanation of the outcome of the case where the
ambitions are more case-centric than theory-oriented. This distinction reflects
the case-centric ambitions of many qualitative scholars and echoes
arguments found in the burgeoning literature on topics such as eclectic
theorization (where the case is front and center) (Sil and Katzenstein 2010)
and pragmatism as a research strategy (Friedrichs and Kratochwill 2009).
Accounting for the outcome of a case usually requires an eclectic combination
of different mechanisms, some of them case-specific/nonsystematic (see
chapters 2 and 4).
We do not suggest this differentiation for its own sake. Instead, by identifying
three variants, we can bring alignment between what we practice and
what we preach, as these differences have important methodological impli-
4 Process-Tracing Methods
cations for research design that are masked when we treat process-tracing as
a single method.
1.2. How Process-Tracing Differs from Other Case Study Methods
Taken together, process-tracing methods can be distinguished from most
other small-n case study methods by the types of inferences being made.
Process-tracing seeks to make within-case inferences about the presence/
absence of causal mechanisms in single case studies, whereas most small-n
methods attempt cross-case inferences about causal relationships. These different
inferential ambitions require different logics of inference, resulting in
fundamentally different methodologies (see chapter 5).
Few case study methods enable within-case inference, and the most
prominent alternative to process-tracing is what George and Bennett term
the congruence method (2005: chapter 9). In the congruence method, based
on the value of the independent variable (X), researchers test whether the
prediction about the outcome that should follow from the theory is congruent
with what is found in the case, investigated either temporally or other
across aspects of the outcome(s) (181–204; Büthe 2002).
The congruence method is often used as a way of structuring a narrative
of a historical process, testing predicted values of X and Y at different times
during an empirical process (t0
, t1
, . . . tn
) (Büthe 2002). “In addition to presenting
information about correlations at every step of the causal process,”
this type of narrative case study “can contextualize these steps in ways that
make the entire process visible rather than leaving it fragmented into analytical
stages” (486). For example, Tannenwald’s (1999) study of the nuclear
taboo involves congruence case studies where she investigates whether the
observable implications of X (norms against using atomic weapons) measured
as “taboo talk” or Z (material factors) measured as “materialist arguments”
are present in decision-making processes within the U.S. government.
She uses a historical narrative of four cases of nuclear use and nonuse
and finds a strong correlation between the presence of taboo talk (X) and
nonuse of nuclear weapons (Y) in three cases where nuclear weapons could
conceivably have been used.
What marks the difference between the congruence method and
process-tracing methods is the explicit focus on investigating causal mechanisms.
Congruence investigates correlations between X and Y, whereas
process-tracing investigates the workings of the mechanism(s) that con-
Process-Tracing in the Social Sciences 5
tribute to producing an outcome. Process-tracing methods go beyond
correlations by attempting to trace the theoretical causal mechanism(s)
linking X and Y.
Process-tracing case studies usually cannot be presented in narrative
form, in contrast to what Rubach (2010) and others have argued. While
evidence in the form of events or temporal sequences can be relevant in
testing the presence of one part of a causal mechanism, depending on the
type of observable implications that are predicted (see chapter 6), other
types of evidence such as pattern evidence (e.g., the number of documents
produced by different agencies) can be relevant for testing other parts of
the mechanism. Process-tracing case studies should therefore usually be
presented as a stepwise test of each part of a causal mechanism, especially
in the theory-testing variant. For example, Owen’s (1994) study of the
democratic peace mechanism is presented as a step-by-step test of each
part of his theorized mechanism instead of a narrative of events in the case
(see chapter 5).
1.3.Themes of the Book
Process-tracing methods are used when we want to gain a greater understanding
of the nature of causal relationships than can be provided by other social
science case study methods, such as comparative cross-case methods. However,
a key deficiency in the existing methodological literature on processtracing
is the absence of sufficient exposition of the logical foundations of
the method or research design, especially with regard to how process-tracing
differs from other qualitative case study methods.
This book rectifies this omission by exploring in detail how the ontological
and epistemological foundations of process-tracing differ from those
of other case study methods, such as congruence methods or structured,
focused comparisons (for more on these two methods, see George and Bennett
2005). Ontology refers to our understanding of the nature of the social
world—specifically, here, the nature of causality. Epistemology refers to arguments
regarding how we should best study causal relationships in the social
world. The argument that we present builds on Hall’s (2003: 374) assertion
that research methodologies and ontology need to be aligned: “Ontology
is ultimately crucial to methodology because the appropriateness of a particular
set of methods for a given problem turns on the assumptions about
the nature of the causal relations they are meant to discover.” As chapter 3
6 Process-Tracing Methods
establishes, adopting the mechanismic and deterministic ontology of causality
of process-tracing implies using quite different methodological tools for
empirical analysis than if a regularity understanding of causality forms the
basis for theorization. Further, the goal of making within-case inferences
about causal mechanisms also implies that a different logic of inference is
adopted than if we are using other small-n methods such as congruence (see
chapter 5).
Chapter 2 explains the three distinct variants of process-tracing, elaborating
on what elements they share as well as their crucial differences, which
have important methodological implications.
Chapter 3 introduces the reader to the ontological debates within the
philosophy of science that deal with the nature of causality to understand
how the mechanismic and deterministic understanding of causality used
in process-tracing methods differs from other social science methods—in
particular, large-n statistical analysis and comparative case study research.
We then explore different ways of investigating causal mechanisms, including
tracing empirical processes, studying them as intervening variables
between X and Y, and using mechanismic, system-oriented understandings.
We contend that to take seriously the study of causal mechanisms,
we should adopt the mechanismic understanding in process-tracing, conceptualizing
causal mechanisms as a series of parts composed of entities
engaging in activities. In so doing, we focus our analytical attention on the
transmission of causal forces through the mechanism. The chapter concludes
with a discussion of the different theoretical levels of mechanisms
along with the question of whether mechanisms can be directly observed
in empirical research.
Chapter 4 deals with questions relating to the theorization of causal
mechanisms. How can causal mechanisms best be conceptualized in a
manner that enables empirical analysis to capture the workings of mechanisms
in a case study? How can causal theories of X→Y be translated into
causal mechanisms composed of a set of parts that describe the theorized
process whereby an explanatory factor (variable or condition) produces an
outcome? Further, how can we work backward from an outcome to build
a sufficient explanation that details the causal mechanisms that produced
that outcome? We discuss how theoretical concepts and causal theories
should be conceptualized in process-tracing before turning to discussion
of the specific challenges in working with each of the three variants of
process-tracing.
In chapter 5, we discuss why mainstream inferential tools used in both
Process-Tracing in the Social Sciences 7
classical statistical analysis and comparative methods cannot be used to
make within-case inferences. Here we continue the argument that methodology
must be brought in line with ontology. In particular, we illustrate that
the inferential tools used in other social science methods are not applicable
in process-tracing, given that we are interested in making within-case inferences
about the presence/absence of causal mechanisms. We then present
the Bayesian logic of inference and how it can be adapted for use as a tool
for making within-case inferences in process-tracing. The chapter concludes
by discussing in more detail the types of inferences that can be made using
different variants of process-tracing methods and, equally important, what
types of inferences cannot be made.
Chapter 6 turns to the question of developing strong empirical tests that
investigate whether a hypothesized causal mechanism is present in a single
case. Based on the Bayesian logic of inference, our goal in process-tracing
is to update our confidence in the presence of a mechanism in light of our
empirical tests. To enable updating to take place, our empirical tests need to
be designed in a manner that maximizes their inferential power. Each test
details the case-specific predictions for what we should expect to see in the
empirical record if each part of the hypothesized causal mechanism is present
in the case.
Empirical material is then gathered to see whether the predicted evidence
is present. However, “raw” empirical observations need to be evaluated for
their content, accuracy, and probability before they can be used as evidence
that enables us to update our confidence. We discuss the evaluation process
in chapter 7, introducing Bayesian-compatible tools for evaluating empirical
material. If there is a strong match between the predicted and found
evidence for each part of the mechanism, we can infer with a certain degree
of confidence that the hypothesized causal mechanism is present in the case
based on the Bayesian logic of inference (Bennett 2008a).
Finally, chapter 8 broadens the picture, looking at questions of case selection
and whether, when, and how the three variants of process-tracing can
be embedded in mixed-method research designs. We discuss case selection
for each of the variants, showing why existing prescriptions do not always
apply. The chapter argues that the theory-building and -testing variants of
process-tracing can be combined with other methods in mixed-method designs,
whereas explaining-outcome designs cannot be meaningfully combined
with other research methods. The key difference is that the former
variants focus on systematic mechanisms, enabling their theories to communicate
with those used in other methods, whereas the latter includes nonsys-
8 Process-Tracing Methods
tematic, case-specific parts, the inclusion of which limits the generalizability
of results.
Finally, the appendix presents a practical checklist for the use of the three
different variants of process-tracing, walking through each step of the research
process to offer guidelines and questions that can be used to structure
a process-tracing analysis.
9
Chapter 2
The Three DifferentVariants of
Process-Tracing and Their Uses
This chapter develops the argument that there are three different research
situations in which process-tracing methods can be used, resulting in three
distinct variants of process-tracing. In contrast, the state of the art treats
process-tracing as a singular method, resulting in murky methodological
guidelines. Whereas most case studies that use process-tracing employ a
case-centric variant that we term the explaining-outcome process-tracing,
most methodological works prescribe a theory-centric version of processtracing
that involves the deductive testing of whether a generalizable mechanism
is present in a single case. The dissonance between what we practice
and what we preach has resulted in considerable confusion about what good
process-tracing is. We contend that clearer prescriptions can be developed
when we differentiate process-tracing into three distinct variants.
We do not suggest this differentiation for its own sake. These differences
have important methodological implications for research design that are
masked when we treat process-tracing as a single method. We explore these
implications throughout the rest of this book. For example, the three variants
differ on key questions such as how causal mechanisms are understood,
whether the purpose is to make inferences about whether a mechanism is
present in a case or to account for a particular outcome, and whether they
can be nested into mixed-method designs.
We first summarize the state of the art, showing that existing work on
process-tracing treats it as a singular method. We then illustrate that there
are three distinct research situations that call for different methodological
tools, implying the need to differentiate the method into three distinct variants
that reflect these different purposes. We conclude by briefly illustrating
10 Process-Tracing Methods
each of the three variants, showing what we are tracing in each of them and
how analysis proceeds.
2.1.The State of the Art—One Method
In their chapter-length presentation of process-tracing, George and Bennett
(2005) mention the range of different forms of process-tracing as they have
been used in practice. The authors argue that process-tracing has been used
in a variety of ways, including both detailed narratives and case studies,
where “at least parts of the narrative are accompanied with explicit causal
hypotheses highly specific to the case without, however, employing theoretical
variables for this purpose or attempting to extrapolate the case’s explanation
into a generalization” (210–11). In other varieties of process-tracing, “the
investigator constructs a general explanation rather than a detailed tracing
of a causal process” (211). Yet in the rest of their chapter, George and Bennett
treat process-tracing as a singular method, masking the differences that
relate to the different uses.
More recent accounts also treat process-tracing as a single method, often
defining it as a deductive tool to test whether causal mechanisms are present
and function as theorized. For example Gerring (2007a: 172–85) describes
a two-stage deductive research process where the analyst first clarifies the
theoretical argument and then empirically verifies each stage of this model.
Checkel describes process-tracing as the attempt to “trace the process in a
very specific, theoretically informed way. The researcher looks for a series of
theoretically predicted intermediate steps” (2008: 363). The end result is a
middle-range theory. Bennett describes process-tracing as a method that involves
“the examination of ‘diagnostic’ pieces of evidence within a case that
contribute to supporting or overturning alternative explanatory hypotheses.
A central concern is with sequences and mechanisms in the unfolding of
hypothesized causal processes. The research looks for the observable implications
of hypothesized explanations. . . . The goal is to establish whether the
events or processes within the case fit those predicted by alternative explanations”
(2010: 208).
Yet treating process-tracing as a singular method results in a large discrepancy
between our prescriptions for good process-tracing (which rely on
a relatively deductive variant of process-tracing) and what we do in practice
(where many scholars want to use the method either to build theories or to
account for particularly puzzling outcomes). The result of treating process-
Three Different Variants of Process-Tracing and Their Uses 11
tracing as one method is a set of murky methodological guidelines, along
with confused students and practitioners.
2.2.The Three Different Uses of Process-Tracing Methods
Process-tracing methods have three distinct research purposes. As illustrated
in figure 2.1, distinctions exist among having the research goal of testing
whether a causal mechanism is present in a case, building a theoretical mechanism,
and crafting an explanation that accounts for a particular outcome.
There is a clear bifurcation overall between theory-centric and case-centric
process-tracing, reflecting a choice between building/testing (relatively) parsimonious
causal mechanisms that can be generalized across a bounded context
of cases and focusing on explaining particular outcomes through the
pragmatic use of mechanismic explanations to account for the important
aspects of the case.
In theory-testing process-tracing, a causal mechanism is hypothesized to
be present in a population of cases of a phenomenon. The researcher selects
a single case where both X and Y are present, and the context allows the
mechanism to operate. Here the goal is to evaluate whether evidence shows
that the hypothesized causal mechanism linking X and Y was present and
that it functioned as theorized. The ambition is to go beyond correlations
and associations between X and Y, opening up the black box of causality to
study more directly the causal mechanism whereby X contributes to producing
Y (see section 3.3).
Theory-building process-tracing involves building a theory about a
causal mechanism between X and Y that can be generalized to a population
of a given phenomenon, starting from a situation where we are in the dark
regarding the mechanism.
Third, and most common in practice, is the situation where we want to
explain a particularly puzzling historical outcome. Here the ambition is not
the theory-centric one of building or testing a generalizable theorized mechanism;
instead, the aim is to craft a sufficient explanation of the outcome.
Instead of studying mechanisms that cause war (Y), the analysis would focus
on explaining a particular outcome such as World War I.
The bifurcation into case- and theory-centric variants of process-tracing
captures a core ontological and epistemological divide within the social sciences.
On the theory-centric side are both neopositivist and critical realist
positions, where the understanding is that the social world can be split
12 Process-Tracing Methods
into manageable parts that can be studied empirically (Jackson 2011). The
ambition here is to build generalizable theories, irrespective of whether we
have the more narrow ambition of working with midrange theories that are
bound within specific contexts or the (perhaps unattainable) ambition to
find law-like generalizations. As chapter 3 discusses, causal mechanisms in
theory-centric studies are understood to be systematic factors, meaning that
they can be generalized across cases that are within the context in which they
are expected to operate (Falleti and Lynch 2009). Here, causal mechanisms
are understood as relatively simple, parsimonious pathways whereby X contributes
to producing Y, but they are not theorized as sufficient causes of Y
by themselves.
Case-centric process-tracing methods operate with a different ontological
understanding of the world. The philosophy of science offers many different
paths to the case-centric position. One path is described by Jackson,
who illustrates the difference between what he terms a dualistic ontology of
mind-world relations where the world exists independent of its human observers
and a monist ontology where “the objects of scientific investigation
are not inert and meaningless entities that impress themselves on our (natural
or augmented) senses or on our theory-informed awareness” (2011: 114).
The monist ontology implies that instead of attempting what is perceived to
be the mission impossible of building and testing law-like generalizations
(theory-centric research), we should instead adopt a form of instrumentalism
aimed at accounting for outcomes in particular cases.
Irrespective of the philosophical path to this position, case-centric reFig.
2.1.Three different uses of process-tracing methods
Three Different Variants of Process-Tracing and Their Uses 13
searchers agree that the social world is very complex, multifactored, and
extremely context-specific. This complexity makes the ambition of producing
knowledge that can be generalized across many cases difficult, if not
impossible. Instead, the ambition is to account for particularly puzzling
outcomes.
Theories are used here in a much more pragmatic fashion—that is, as
heuristic instruments that have analytical utility in providing the best possible
explanation of a given phenomenon (Peirce 1955). Case-centric research
scholars contend that it makes little sense to distinguish between systematic
and case-specific parts, given the impossibility of generalization in the
complex social world. Further, theories that are developed are much more
eclectic, often including conglomerates of different mechanisms along with
more case-specific mechanisms.
The ambition is not to prove that a theory is correct but instead to prove
that it has utility in providing the best possible explanation. Explanations are
case-specific and cannot be detached from the particular case (Humphreys
2010: 269–70) (see chapter 5).
2.3.The ThreeVariants of Process-Tracing
What are the core elements of each of the three variants of process-tracing?
A number of commonalities exist across the three variants. For example, all
variants share the goal of studying causal mechanisms. Ontological assumptions
about the nature of causal relationships are also shared. These include
the use of deterministic theorization and a mechanismic understanding of
causation that focuses on the process whereby causal forces are transmitted
through a series of interlocking parts of a mechanism to produce an outcome
(see chapter 3). The three variants of process-tracing share a theoretical
understanding of mechanisms as invariant; they are either present or not (see
chapter 4). In addition, all three methods draw on a Bayesian logic of inference
to make within-case inferences about the presence/absence of causal
mechanisms (see chapter 5).
What differentiates the three variants is
•	 whether they are theory-centric or case-centric designs
•	 aim to test or build theorized causal mechanisms
•	 their understanding of the generality of causal mechanisms (from
systematic mechanisms expected to be present in a set of cases
[population] to case-specific mechanisms)
14 Process-Tracing Methods
•	 the types of inferences being made, where theory-testing
or -building variants make inferences about the presence/absence of
a mechanism, whereas explaining-outcome process-tracing enables
inferences about the sufficiency of the explanation to be made.
We now turn to a presentation of what each variant is actually tracing, illustrating
a typical research process for each variant.
Theory-Testing Process-Tracing
In theory-testing process-tracing, we know both X and Y and we either have
existing conjectures about a plausible mechanism or are able to use logical
reasoning to formulate a causal mechanism from existing theorization.
Figure 2.2 illustrates a simple abstract example of a theory-testing case
study. The first step in testing whether a hypothesized causal mechanism
was present in the case is to conceptualize a causal mechanism between X
and Y based on existing theorization along with making explicit the context
within which it functions. In this example, a two-part mechanism between
X and Y is deduced, with each part composed of entities engaging in activities.
This theorized causal mechanism then needs to be operationalized (step
2), translating theoretical expectations into case-specific predictions of what
observable manifestations each of the parts of the mechanism should have if
the mechanism is present in the case. In practice, theory-testing has inductive
elements, especially regarding the operationalization of empirical tests, where
we draw on existing empirical work to make case-specific empirical predictions
about what evidence we should see if the theory is valid (see chapter 6).
Once the mechanism and context are conceptualized and operationalized,
the analyst proceeds to step 3, where she collects empirical evidence
that can be used to make causal inferences, updating our confidence in
(1) whether the hypothesized mechanism was present in the case and (2)
whether the mechanism functioned as predicted or only some parts of the
mechanism were present. The bold lines in figure 2.2 illustrate the inferences
made in theory-testing process-tracing, where we infer from the empirical
evidence collected that a causal mechanism was present in the case.
The empirical analysis in step 3 proceeds stepwise, testing whether evidence
indicates that each part of the mechanism was present. Most important,
the evidence necessary to test whether the different parts are present
can be very different, making evidence for the parts noncomparable with
each other. Therefore, a case study usually does not read like an analytical
Three Different Variants of Process-Tracing and Their Uses 15
narrative, in that while evidence in the form of events can be an observable
manifestation of one part of a causal mechanism (depending on the type of
observable implications that are predicted), other types of evidence, such
as pattern evidence (e.g., the number of documents produced by different
agencies) can be equally relevant (see section 3.3.).
What, then, are we actually tracing when we engage in theory-testing
process-tracing? What is being traced is not a series of empirical events or
narratives but instead the underlying theorized causal mechanism itself, by
observing whether the expected case-specific implications of its existence are
present in a case (see chapter 3).
Theory-testing process-tracing enables inferences to be made about
whether a causal mechanism was present in a single case along with whether
the mechanism functioned as expected. However, theory-testing processtracing
does not enable us to test the relative explanatory power of competing
mechanisms against each other except in the rare situation where two
competing mechanisms can be conceptualized so that they are composed
of the same number of diametrically opposite parts with observable implications
that rule each other out (see chapter 5). Further, given that we can
make inferences only about whether a mechanism was present in the single
Fig. 2.2.Theory-testing process-tracing
16 Process-Tracing Methods
case, no claims about the necessity of the mechanism can be logically made.
To do so requires cross-case analysis (see chapter 8).
Theory-Building Process-Tracing
The second identifiable variant of process-tracing also has theoretical ambitions
beyond the confines of the single case. In its purest form, theorybuilding
process-tracing starts with empirical material and uses a structured
analysis of this material to detect a plausible hypothetical causal mechanism
whereby X is linked with Y. While it is mentioned as a possibility in the
literature, this inductive, theory-building variant of process-tracing is surprisingly
neglected. To our knowledge, the literature contains no attempts
to show how it is done in practice.
Theory-building process-tracing is utilized in two different research situations:
(1) when we know that a correlation exists between X and Y but we
are in the dark regarding potential mechanisms linking the two (X-Y-centric
theory building) as we have no theory to guide us; or (2) when we know an
outcome (Y) but are unsure about the causes (Y-centric theory building). In
the second instance, the analysis first traces backward from Y to undercover
a plausible X, turning the study into an X-Y-centric analysis.
What is also being traced here is a theoretical causal mechanism that
is expected to be present across a population of cases (i.e., it is a systematic
mechanism). The core difference between theory-testing and -building
process-tracing involves theory before fact versus fact before theory. In
theory-building process-tracing, empirical material is used to build a hypothesized
theory, inferring first that what is found reflects the observable
implications of an underlying causal mechanism. A second leap is then made
by inferring from these observable implications that they reflected an underlying
causal mechanism. However, both variants share a focus on tracing a
generalizable causal mechanism by detecting its empirical manifestations.
While theory-building process-tracing as an inductive method has some
elements that overlap with explaining-outcome process-tracing, the key
difference between the two is that theory-building process-tracing seeks to
build a midrange theory describing a causal mechanism that is generalizable
outside of the individual case to a bounded context (e.g., spatially or temporally
bounded), whereas explaining-outcome process-tracing focuses on
building a minimally sufficient explanation of the outcome in an individual
case. Theory-building process-tracing studies do not claim that the detected
causal mechanism is sufficient to explain the outcome.
Three Different Variants of Process-Tracing and Their Uses 17
Figure 2.3 illustrates the basic framework of a theory-building processtracing
case study. After the key theoretical concepts (X and Y) are defined,
theory-building proceeds to investigate the empirical material in the case
(step 1), using evidence as clues about the possible empirical manifestations
of an underlying causal mechanism between X and Y that fulfills the guidelines
for a properly conceptualized causal mechanism (see chapters 3 and 5).
This process involves an intensive and wide-ranging search of the empirical
record.
Step 2 involves inferring from the observable empirical evidence that
these manifestations reflect an underlying causal mechanism that was present
in the case. Evidence does not speak for itself. Theory-building often has
a deductive element in that scholars seek inspiration from existing theoretical
work and previous observations. For example, an analyst investigating
socialization of administrative officials within international organizations
could seek inspiration in theories of domestic public administration or in
psychological theories of small group dynamics while also reading more descriptive
accounts of the workings of international organizations for plauFig.
2.3.Theory-building process-tracing. (Bold lines = direct inferences; shaded
lines = indirect (secondary) inferences; shaded area = what is being traced.)
18 Process-Tracing Methods
sible causal mechanisms. Here, existing theory can be conceived as a form
of grid to detect systematic patterns in empirical material, enabling inferences
about observable manifestations. In other situations, the search for
mechanisms is based on hunches drawn from puzzles that are unresolved by
existing work. In step 3, the secondary leap is made from observable manifestations
to infer that they reflect an underlying causal mechanism.
Figure 2.3 illustrates that theory-building process-tracing is examining
an underlying theoretical causal mechanism, depicted as the shaded area
that forms the backdrop of the theoretical level (X, causal mechanism, Y).
In contrast to theory-testing process building, the empirical analysis itself,
understood as the collection of the “facts” of the case, is two inferential leaps
removed from the theorized causal mechanism (i.e., the inferences are indirect).
This is illustrated by the bold lines linking the “facts” with observable
manifestations (direct inferences) and the subsequent secondary inferential
leap from these observable implications to the inference that parts of an
underlying causal mechanism existed.
In reality, theory-building process-tracing is usually an iterative and creative
process. Hunches about what to look for that are inspired by existing
theoretical and empirical work are investigated systematically, with the results
of this search forming the background for further searches. This means
that steps 1 and 2 are often repeated before step 3 is reached.
Explaining-Outcome Process-Tracing
The goal of many (if not most) process-tracing studies is to explain a particular
interesting and puzzling outcome. While existing prescriptions for
process-tracing speak almost exclusively about what we understand as the
theory-centric variants, what most scholars are actually using is explainingoutcome
process-tracing.
This type of process-tracing can be thought of as a single-outcome study,
defined as seeking the causes of a specific outcome in a single case (Gerring
2006).1
Here the ambition is to craft a minimally sufficient explanation of a
particular outcome, with sufficiency defined as an explanation that accounts
for all of the important aspects of an outcome with no redundant parts being
present (Mackie 1965). This approach marks a significant departure from the
two theory-centric variants. For example, in theory-testing process-tracing,
no claims are made about whether the mechanism is sufficient; rather, inferences
are made only about whether the postulated mechanism is present or
absent in the single case.
Three Different Variants of Process-Tracing and Their Uses 19
While explaining-outcome process-tracing studies sometimes more
closely resemble historical scholarship, this type of process-tracing is in our
opinion still social science research, as the ultimate explanation usually involves
more generalized theoretical claims than historians feel comfortable
with. In addition, explaining-outcome studies often have theoretical ambitions
that reach beyond the single case.
It is vital to note that the term causal mechanism is used in a much
broader sense in explaining-outcome process-tracing than in the two
theory-centric variants. First, whereas theory-testing and -building variants
of process-tracing aim to test/build mechanisms that are applicable across
a range of cases, crafting a minimally sufficient explanation almost always
requires combining mechanisms into an eclectic conglomerate mechanism
to account for a historical outcome (see chapter 3). Second, given that the
ambition is case-centric and seeks to craft a minimally sufficient explanation
of a particular outcome, it is usually necessary to include nonsystematic
parts in the causal mechanism, defined as a mechanism that is case-specific.
Explaining-outcome process-tracing is an iterative research strategy that
aims to trace the complex conglomerate of systematic and case-specific
causal mechanisms that produced the outcome in question. The explanation
cannot be detached from the particular case. Theorized mechanisms are
therefore seen as heuristic instruments whose function is to help build the
best possible explanation of a particular outcome (Humphreys 2010; Jackson
2011).
While explaining-outcome process-tracing as an iterative strategy most
closely resembles abduction, which is a dialectic combination of deduction
and induction (Peirce 1955), for our purposes it is more helpful to disaggregate
two alternative paths that can be chosen when building the best possible
explanation of an outcome—deductive and inductive paths, as shown in figure
2.4. This figure does not split the mechanism into parts, as the previous
figures do, because of the complexity of a pictorial depiction of the parts of
an overlapping, conglomerate mechanism.
The deductive path follows the steps described previously under theorytesting,
where an existing mechanism is tested to see whether it can account
for the outcome. This process is illustrated using black arrows for each of the
three steps. The first arrow is where a theory is conceptualized as a mechanism.
In the second step, empirical tests are developed that are then evaluated
against the empirical record. Finally, the third arrow illustrates the stage
where the analyst assesses whether a sufficient explanation has been crafted.
However, in most explaining-outcome studies, existing theorization cannot
provide a sufficient explanation, resulting in a second stage of research
20 Process-Tracing Methods
where either a deductive or inductive path can be chosen, informed by the
results of the first empirical analysis. If the deductive path is chosen again,
alternative theories must be tested to see whether they provide a sufficient
explanation. Alternatively, the inductive path can be chosen in the second
iteration, using empirical evidence to build a better explanation.
The inductive path is often used when we are examining a little-studied
outcome. This path is depicted in figure 2.4 as gray arrows, starting from the
empirical level. Here, the analyst can proceed in a manner more analogous
with historical methodology or classic detective work (Roberts 1996)—for
example, working backward from the outcome by sifting through the evidence
in an attempt to uncover a plausible sufficient causal mechanism that
produced the outcome. This is a bottom-up type of analysis, using empirical
material as the basis for building a plausible explanation of causal mechanisms
whereby X (or multiple Xs) produced the outcome.
The important question is then when should we stop this process—that
is, How do we know a minimally sufficient explanation when we see it?
There is no foolproof answer to this question; instead, the decision that we
have a minimally sufficient explanation is based on an assessment of whether
all of the relevant facets of the outcome have been accounted for adequately
Fig. 2.4. Explaining-outcome process-tracing
Three Different Variants of Process-Tracing and Their Uses 21
while ensuring that the evidence is best explained by the developed explanation
instead of plausible alternative explanations. This is an iterative process
where we update the model until it provides what can be thought of as the
best possible explanation (Day and Kincaid 1994). We can never confirm
a theory with 100 percent certainty; instead, we stop when we are satisfied
that the found explanation accounts for the most important aspects of the
outcome (see chapter 5).
2.4. Conclusions: A New Understanding of Process-Tracing
We need to differentiate process-tracing methods into three distinct variants
to bring alignment between what we practice and what we preach. Common
to all three variants is the ambition to trace causal mechanisms, although
TABLE 2.1. Summary of the Main Differences between theThreeVariants of Process-Tracing
Theory-Testing Theory-Building Explaining-Outcome
Purpose of Situation one Situation two Situation three
analysis— Correlation has been Build a plausible Explain particularly
research found between X causal mechanism puzzling historical
situation and Y, but is there linking X:Y based outcome by
evidence that there on evidence in case building minimally
exists a causal sufficient explanation
mechanism linking in case study
X and Y?
Ambitions of Theory-centric Theory-centric Case-centric
study
Understanding Systematic Systematic Systematic,
of causal (generalizable (generalizable nonsystematic
mechanisms within context) within context) (case-specific)
mechanisms and
case-specific
conglomerates
What are we Single, Single, Case-specific,
actually generalizable generalizable composite
tracing? mechanism mechanism mechanism that
explains the case
Types of (1) Parts of Observable Minimal
inferences causal mechanism manifestations sufficiency of
made present/absent reflect explanation
(2) Causal underlying
mechanism is mechanism
present/absent
in case
22 Process-Tracing Methods
the term causal mechanism as used in theory-testing and theory-building
variants refers to relatively parsimonious mechanisms that are generalizable
to a bounded population of cases, whereas in explaining-outcome processtracing,
mechanisms refer to systematic mechanisms, case-specific, nonsystematic
mechanisms (events leading to an outcome), and eclectic casespecific
conglomerates of different mechanisms.
Table 2.1 summarizes the main points of difference across the three variants
of process-tracing. There are three different purposes of process-tracing
methods: (1) testing whether a generalizable causal mechanism exists in a
case and functions as expected; (2) building a generalizable mechanism from
evidence in a case; and (3) explaining a particular outcome. The methods differ
regarding whether they are theory- or case-centric, along with what they
are actually tracing and the types of inferences they enable.
The rest of this book addresses the commonalities and differences across
the three variants of process-tracing with regard to their ontological and
epistemological foundations (chapter 3), the practical guidelines for each
stage of the research process from working with theories (chapter 4), and
the types of inferences being made (chapter 5). The book also explores developing
empirical tests (chapter 6), working with evidence (chapter 7), and
answering questions of case selection and nesting case studies in broader,
mixed-method research designs (chapter 8).