Two or three approaches to explanatory case study research?

Joachim Blatter (University of Lucerne, [email protected])

Markus Haverland (Erasmus University Rotterdam, [email protected])

Paper prepared for presentation at the Annual Meeting of the American Political Science
Association, New Orleans, August 30 – September 2, 2012. Panel 8-10 “Quantitative Qualitative
Research”.

Abstract
This paper substantiates the core message of our book Designing Case Studies:
Explanatory Approaches in Small-N Research (Palgrave 2012) that there are three rather
than two distinct approaches to explanatory case study research. Distinguishing three
approaches is of crucial importance for executing, supervising and reviewing case study
research, as each approach has its own affinities with regard to crucial elements of the
research process such as appropriate research questions, criteria for case and theory
selection, methods of causal inference, modes of generalization and format of
presentation.
1

Electronic copy available at: https://ssrn.com/abstract=2105542

Introduction
In recent years we have seen an explosion of methodological reflections on case study
research. These reflections have challenged the co-variational orthodoxy (comparative
case study designs) that dominated the literature on case study methodology in political
science since the 1970s. Alternative understandings have focused on within-case analysis
and have been presented mostly under the heading of “process tracing”. We argue that
the various accounts of process tracing are actually a mixed bag and that we gain a lot
when we realize that there are two alternatives to the co-variational template. In addition
to an approach that is based on the co-variational template which we dub “co-variational
analysis (COV)” and deals with cross-case comparisons, we make a distinction between
two variants of within-case analysis: “causal-process tracing (CPT)” and “congruence
analysis (CON)”. Distinguishing two approaches of within-case analysis broadens the
available toolkit for drawing causal inferences in small-N research and allows for
developing internally coherent research approaches (Blatter and Haverland 2012, see also
Blatter and Blume 2008a, 2008b).
The paper proceeds as follows. First we sketch the development of the methodological
debate of explanatory case study research that has initially revolved around the
comparative case study designs and only later has embraced within-case approaches. We
will then compare recent work that is presented under the banner of process tracing:
George and Bennett (2005), Hall (2008) and Bennett (2010). We will try to demonstrate
that there are two distinctive approaches of within-case analysis hided in their work. In
the remainder of the paper then present our three approaches: COV, CPT and CON,
demonstrating how they differ from each other in terms of research goals and questions,
case and theory selection, data generation and analysis, and generalization.

Methodological reflections on small N studies: a brief history

Comparative case study designs
Traditionally, explanatory case study research is associated with the comparison of a
small number of variables across a few cases. The relevant designs are discussed in

Electronic copy available at: https://ssrn.com/abstract=2105542

comparative sociology in terms of Mill’s methods of difference and of agreement, hence
based on the logic of reaching conclusions (Mill 1875, see e.g. Skocpol and Somers
1980). In political science, these designs are known as the Most Similar Systems Design
and the Most Different Systems Design, hence emphasizing the case selection strategy
(Przeworski and Teune 1970). Lijpharts “comparative method”, or “comparable cases
strategy” (Liphart 1971, 1975) can be taken as one variant of the Most Similar Systems
Design. This variant does also underlie our approach to comparative case studies:
“covariational analysis” (Blatter and Haverland 2012).
There has been a very intensive debate about comparative case study designs.
Discussions revolved around the validity of its assumption, case selection criteria and
selection bias, feasible variants and combinations, and more generally the limitations of
the conclusions reached. This debate occurred both in comparative sociology and in
political science in the 1980s and 1990s, although in isolation from each other (see e.g.
Faure 1994, Frendreis 1983, Geddes 1990, Lieberson 1991, 1994, Nichols 1986, Skocpol
1986).
Yet, this intense methodological debate did not reasonate well with the fact that many
comparative case studies drew their strength less from the way they compare cases but
more by the within-case analysis. Take the probably most often discussed work in the
comparative social sciences Theda Skocpol’s State and Social Revolutions (1979). Many
scholars argue that it is less her cross-case comparison that make her study compelling
but more her analysis of revolutionary processes (Mahoney 1999: 1157, Sewell 2005:
97); in other words, her focus on the causal conjunctures and causal chains that have led
to the revolutions in France, Russia and China (Mahoney 1999: 1164-8, see also Blatter
and Haverland 2012).
More generally, discussing why comparative case studies have been so influential in
comparative politics, despite widely acknowledged limitations of comparative case study
designs, Peter Hall argues that it is because many scholars in comparative politics have
not relied heavily on the comparative method, “instead the success of the field rests
largely on the ways it has moved beyond that method” (2004: 2). Scholars have, for
instance, moved “toward the view that explanation consists in identifying the causal
process that lie behind political outcome” (Hall 2004: 3).

Electronic copy available at: https://ssrn.com/abstract=2105542

Hence, ironically, the methodological debate was unevenly concentrated on methods that
mattered relatively less, while the within-case analysis that often did the trick received
comparatively less attention.

Within-case designs: the early days

To be sure, some scholar have covered within-case analysis, but only very briefly and not
from a critical, methodological perspective. In fact, while some ideas about historical
processes have been sketched in George’s approach of structured focused comparisons
(George 1979), it was to our knowledge only in 1985 that the term “process tracing” has
been introduced in the methodological debate and briefly elaborated upon (George and
McKeown 1985).
Parallel to this development have been some references to the fact that one case can
actually mean many observations. Famously in a special issue of Comparative Political
Studies in which also Lijphart clarified and elaborated on “his” comparative method,
Campbell, the eminent advocate and developer of quasi-experimental designs wrote an
article that contains this much cited quote:
“In a case study done by an alert social scientist who has thorough local
acquaintance, the theory he uses to explain the focal difference also generate
predictions or expectations on dozens of other aspects of that culture, and he does
not retain the theory unless most of these are also confirmed. In some sense, he
has tested the theory with some degree of freedom coming from the multiple
implications of any one theory. The process is a kind of pattern matching in which
there are many aspects of the patterns demanded by theory that are available for
matching with his observation s on the local setting (Campbell 1975: 182-3).

“Pattern matching” reappeared in Yin’s bestselling book on case studies, for which
Campbell wrote a pre-face (Yin 1984). In that book, Yin provides a couple of pages on
several variants of pattern-matching of which some were not immediately recognizable as
variant because they carried rather confusing names such as “logic models”.
Overall we think that this was very much what was there in terms of methodological
treatments of within-case analysis until the 1990s..

Electronic copy available at: https://ssrn.com/abstract=2105542

Current discussion on within-case analysis: Process tracing all over
A more comprehensive view on explanatory case studies was stimulated by the
publication of King, Keohane and Verba’s Designing Social Inquiry, which sought to
catholize case study scholars into doing small N research following a statistical template.
As a response qualitative scholars thought harder about the intrinsic merits of small N
research designs, and ways to improve them (e.g. Brady and Collier 2004, 2010, Goertz
and Mahoney forthcoming, Ragin 2008). In this context, the discussion about within-case
analysis seemed to have converged towards notion of process tracing. Yet, as stated in the
introduction, we argue that what counts as process tracing is rather a mixed bag.

George and Bennett Case Studies and Theory Development

George and Bennett’s monograph Case Studies and Theory Development (2005), a
restatement and elaboration of their earlier work, became a major reference point for the
current discussion. They treat process tracing as “an operational procedure for attempting
to identify and verify the observable within-case implications of causal mechanisms” and
define causal mechanisms “as ultimately unobservable physical, social, or psychological
processes through which agents with causal capacities operate, but only in specific
contexts or conditions, to transfer energy, information, or matter to other entities”
(George and Bennett 2005: 137-38). Furthermore, they stressed that explanations via
causal mechanisms have affinities to a scientific realist epistemology – which implies a
commitment to micro-foundations – fundamental assumptions about the behavior of
actors which form the basis for most basic social science theories; and that explanations
via causal mechanisms draws on spatial contiguity and temporal succession (George and
Bennett 2005: 137-145). Also important for our discussion is that they claim that “all
(emphasis by the GB) the intervening steps in a case must be predicted by a hypothesis…
or else the hypothesis must be amended…to explain the case” (George and Bennett 2005:
207) .

Electronic copy available at: https://ssrn.com/abstract=2105542

Hall Systematic Process Analysis
Another important contribution to within-case analysis stem from Peter Hall. Already in
2003 he has called upon political scientists “to align methodology with ontology”, by
which he means that the methods we use should reflect the causal complexity assumed by
our theories. This call found wide resonance. In 2008 he then presents his own take on
process tracing called “systematic process analysis”. He starts his account by
distinguishing three modes of positivist explanations. In addition to the “multi-variate
explanation” that informs statistical analyses, he distinguishes the “historically specific”
explanation” and the “theory-oriented explanation”. Regarding the historical specific
explanation Hall writes that “historians are unusually attentive to context, namely how
factors interact to generate an outcome and the spatial and temporal specificities affecting
the value of each factor” (Hall 2008: 305).The theory-oriented approach is characterized
by “ elucidating and testing a theory that identifies the main determinants of a broad class
of outcomes and attaches special importance to specifying the mechanisms whereby
those determinants bear on the outcome…. The focus is on elucidating the process
whereby the relevant variables has effect” (Hall 2008: 306). He also states that a theory-
oriented approach has affinity with critical realism (ibid.).
It is on the theory-oriented explanation that he builds his “systematic process analysis”.
In other words, he leaves the study of the interaction of variables in specific cases and of
spatial and temporal specificities to historically-oriented scholars. To be sure for theory
formation he advises to account for how principal causal variables “interact in that causal
chain that leads to the outcome” and the process whereby the variables have an effect
(2008: 309).
However when discussing the way in which conclusions are drawn the emphasis is on the
congruence between predictions and observations and on whether congruence concerns
core proposition of the theory rather than on interactions and on causal chains.
This view is distinct from process tracing as expressed in the monograph of George and
Bennett. As stated above George and Bennett emphasize that each element of the causal
chain needs to be documented, Hall seems to be less strict about that. Systematic process
tracing is a “three corner fight” (Hall). The theory “wins” that reveals more about “its”
assumed causal chain. This is different form identifying the causal process resulting in a

Electronic copy available at: https://ssrn.com/abstract=2105542

particular outcome (or class of outcomes). In our view Hall’s emphasis is much more on
deduction and predictions than palatable for a scientific realist. His approach is more in
line with a deductive-nomological model of explanation. To be it bluntly, there is nothing
related to “process” that is strictly necessary for his “systematic process analysis”.
In fact what Hall describes as “systematic process analysis” comes much closer to what
George and Bennett discuss as an alternative for process tracing, the “congruence
method”. Here “ the investigator begins with a theory and then attempts to assess its
ability to explain or predict an outcome in a particular case” George and Bennett 2005:
181) They also mention that one way to increase the validity of the findings is to establish
whether the theory fits the case better than another theory George and Bennett 2005:
184). In other words Hall’s “three corner fight” does here belong to the alternative of
process tracing.

Bennett Process Tracing and Causal Inference

In a more recent piece it seems that Bennett has departed from the scientific realist
foundation of process tracing with its “naturalistic” emphasis on spatial contiguity and
temporality that has characterized his monograph with George (Bennett 2010). One could
argue that he even goes a step further than Hall in his emphasis on theory testing and
deduction. Although he emphasizes upfront that “a central concern is with sequences and
mechanisms in the unfolding of hypothesized causal processes” (2010: 208) he continues
by stating that “process tracing involves several kinds of empirical tests, focusing on
evidence with different kind of probative value”. He then builds on the well know
typology of tests developed by Van Evera, such as the “straw in the wind test” and the
“hoop” test. Our point is that these tests can do with single observations, or several
observations that do not necessarily have to capture a process. The tests do not
necessarily depend on timing and sequences. Note that Van Evera prefaces his discussion
of various tests with an elaboration of the “covering law” explanation, in other words the

Electronic copy available at: https://ssrn.com/abstract=2105542

 deductive-nomological model of explanation, which is different from scientific realism
(Van Evera 1997: 74-75).1

Towards systematization: Causal-process tracing versus congruence analysis

We argue that there are two approaches hided in the recent contributions to within-case
analysis. We dub the first approach “Causal-process tracing”. This approach “saves”, so
to speak, process tracing as envisaged in George and Bennett’s monograph (2005). We
take up the notion of George and Bennett that causal mechanisms only produce effects
under “specific contexts or conditions”. We further elaborate and specify the contextual
nature of mechanism-based explanations by using “configurational causation” as
ontological presuppositions. Furthermore, our understanding shares George and Bennett’s
affinity to “scientific realism” as an epistemological position that implies that the spatio-
temporal continuity and contiguity of social processes serves an important “natural basis”
for drawing causal inferences (George and Bennett 127-149). Researchers who advocate
a “scientific realist” epistemological stance emphasize that “causation is a relation in
nature, not in logic” (Wendt 1999: 81). Moreover as process tracing has to deal with
processes we have to put “timing” or “temporality” into the centre of this method of
drawing causal inference.2
Our second approach “congruence analysis” is more in line with what George and
Bennett (2005) call the congruence method or Hall has dubbed “systematic process
observation” (2008). Congruence analysis is a strongly deductive endeavor in which we
develop competing theories, and deduce propositions before we enter the empirical field
in order to test, compare and refine these competing theories. Bayesian reasoning plays a
role when we reflect on the status of theories in the scientific discourse and when we are
looking for “crucial cases”.

1
In his contribution to the Oxford Handbook of Political Methodology, Bennett compares process tracing
with Bayesian logic and concludes that they have very much in common (Bennett 2008). In our view,
Bayesian logic is difficult to square with scientific realism.
2
In order to highlight this stance we bind the two terms “causal” and “process” together by a hyphen. This
indicates that we are actually tracing “causal-processes” which implies a multiplicity of observations which
provide information about the social development at different points in time. When we refer to usages of
this term in the literature that are not in correspondence to our understanding, we apply the term without a
hyphen (causal process tracing).

Electronic copy available at: https://ssrn.com/abstract=2105542

We want to emphasis that this distinction is not made l’art pour l’art. We argue that
together with “co-variational” analysis we actually have three approaches to explanatory
case study research which, as already stated above, have distinctive affinities with regard
to crucial elements of the research process such as appropriate research questions, criteria
for case and theory selection, methods of causal inference, modes of generalization and
format of presentation. This makes the distinction highly relevant in particular for those
who actually conduct explanatory case studies because it allows them to execute
internally consistent case studies and sensitize them to possibilities and limitation
regarding the combination of different approaches.

The three approaches to explanatory case study research

In this section we will compare our three approaches to explanatory case studies research
regarding research goals and questions, case and theory selection, data generation and
analysis and directions of generalization.

Research goals and research questions

The co-variational (COV) approach, to case study research typically aims to investigate
whether a specific factor makes a difference. For example: Does government
reorganization reduce public spending? As this kind of research is interested in the effect
of a specific causal factor, or independent variable, this research can be labelled X-
centred research. But the focus on “independent variables” has a further, deeper meaning
because the COV approach assumes that the causal factors function independently of
each other; this approach is based on the ontological assumption that each factor has
autonomous causal power, in other words, interaction effects are absent.
The causal-process tracing (CPT) approach, instead, start with an interest in a specific
(kind of) outcome. The investigator asks what factors lead to a concrete outcome or
which preconditions are necessary and sufficient in order to make a specific kind of
outcome possible. For example: Which conditions lead to social revolution? Because the
researcher is interested in the (various) causes of an effect rather than in the effect of a
specific cause (independent variable), this approach can be called Y-centered research.

Electronic copy available at: https://ssrn.com/abstract=2105542

For an Y-centred research project it makes sense to start with the ontological assumption
that a plurality of factors work together to produce the outcome of interest. Such a
holistic ontological starting point leads to the search for configurations of causal
conditions or social mechanisms and it tries to reveal the details of the “causal pathways”
that lead to the outcome of interest.
Other case studies are conducted primarily with the aim of contributing to the theoretical
debate in a discipline or field of research. Typical research questions read as follows: Is
Liberal Intergouvernmentalism the best explanation for European Integration?
(Moravcsik 1998). Such research questions recognize that paradigms and theories have
an important function in the process of knowledge generation because they provide the
anchor points for research programs and structure the scientific discourse. The
congruence analysis approach (CON) is theory-oriented, but note that theories are not
reduced to single independent variables (as in the COV approach) but are treated as
comprehensive explanatory frameworks that are specified through a set of constitutive
and causal propositions. Case studies are used to elucidate and to compare the
explanatory merits of competing or complementary theories.

Case and theory selection

For the COV approach, case selection is crucial to demonstrate that it was indeed
variation in X and not another factor that caused the effect (variation in Y). In other
words, case selection is crucial to making valid causal inferences. A plurality of cases is
selected according to the experimental template. This means that the cases must express
strong differences with respect to the main independent variable of interest, and they
must be as similar as possible with regard to variables associated with other potential
explanations. As discussed above, this design is described using the term “most similar
system design” (Pzeworski and Teune 1970) or, alternatively, the “method of difference”
(Mill 1875); this design also corresponds to the “comparative method” Lijphart (1975).
The accessibility of a case is the primary precondition for investigation when it comes to
the CPT approach. This is because a CPT approach depends on the ability to provide
quite comprehensive storylines on the temporal unfolding of the causal process, to
provide a dense description of critical moments, and on opportunities to gain deep
10

Electronic copy available at: https://ssrn.com/abstract=2105542

insights into the perceptions and motivations of important actors. Therefore, accessibility
is the primer criterion for case-selection. Causal-process tracing is a within-case
analytical technique; therefore, we need not select more than one case, although we do
have the option to do so. In the ideal typical form of the CPT approach, those cases are
selected that show a strong “positive” result with respect to the outcome of interest. In a
second step, further “possible” cases can be selected to test the relevance of specific
factors that have been identified as necessary for the outcome in the first study (see
Mahoney and Goertz 2004).
Within the CON approach, the selection of theories has to be done more explicitly than in
the other approaches. Ideally, this step precedes the selection of cases. We advocate
selecting more than one theory and avoiding the ex-ante integration of those theories in a
synthetic explanatory approach. The researcher should consider a plurality of theories and
should reflect on the status of these theories in the scientific discourse. This allows for
selecting potentially “crucial cases”: cases which are “most-likely” to show high levels of
congruence with the expectations deduced from the dominant theory in the scientific
discourse and “least-likely” to conform to alternative theories. The ex-ante likeliness of a
case in respect to be conform to theoretical expectations, in turn, depends on some prior
knowledge about specific features of the case, such as context or antecedent conditions
(see also Eckstein 1975)

Data generation and data analysis

While it is a defining characteristic of all case study approaches that a large number of
(diverse) empirical observations are collected per case and that there is an intensive
reflection on the relationship between concrete empirical observations and abstract
theoretical concepts, there are strong differences in the ways in which observations are
transformed into data and in the ways the data are analysed to draw causal inferences.
Despite these differences, all case study approaches share one feature: in case study
research, it is the first step, data generation, that is most crucial; case study researchers
invest much more time and intellectual energy in this first step in comparison to the time
and energy they invest into the second step of data analysis, and the cogency of case

11

Electronic copy available at: https://ssrn.com/abstract=2105542

studies depends almost fully on this first step (whereas in large-N studies the result is
often evaluated on the basis of the technical skills applied in statistical data-analysis).
In table 1, we present the processes of data generation and data analysis separately to
present clearly the functional equivalents in each approach. Whereas for the rather
deductive approaches COV and CON, this neat separation represents the way we conduct
case studies (or at least, it corresponds to the way we present the findings), this is not the
case with the rather inductive CPT approach, in which the separation of data generation
from data analysis is less clear-cut.3
In the COV approach, indicators that scholars have selected for operationalizing variables
into observable entities define which empirical information is seen as relevant and which
information must be collected for each case. The relevant empirical information is used to
determine the scores for each of the variables; therefore, we call the corresponding
information “variable-scoring observations”. Researchers invest significantly in making
sure that each score is valid, and they typically employ a large number of empirical
observations for this task. As a result, a crucial step in this research approach is the
process of transforming the information that we find “out there” in the social world into
scores for individual variables. Compared with large-N studies, the COV approach makes
it much easier to apply indicators in a context-sensitive way, which means that nominally
different states of the social world (for example, number of parties in a parliament) can be
treated as functionally equivalent (for example, for the concept of “competition”), and
nominally equal states can be scored differently. Data analysis takes place in a second
distinct step after we have transferred all scores of all cases for all variables into a
rectangular data sheet. Through visual inspection, we discover whether there is co-
variation among the scores of the dependent variable of interest (Y) and the scores of the
independent variable (X). If so, we can conclude that X has a causal effect on Y. A
necessary condition for this inference is that there exists no other theoretically plausible
co-variation among scores of other independent variables and the dependent variable.

3
The term “rather” indicates in its first usage that in small-N studies deductive approaches in reality are
never as deductive as large-N studies where the operationalization has to be finished before we start to
search for information. “Rather” in the second part of the sentence points to the fact that CPT is an
inductive approach only in comparison to the other approaches. As we will see, the method of CPT
can be applied in more inductive research designs and more deductive ones.

12

Electronic copy available at: https://ssrn.com/abstract=2105542

In the ideal-typical CPT approach, the search for relevant empirical information proceeds
in a much more inductive fashion. The researcher has to search for all kinds of
information about the temporal unfolding of the causal-process that allows her to present
a comprehensive story line with a sequence of causal steps. For decisive situations and
phases of transformation, the researcher searches for information that gives her a more
detailed picture of the “scene” and a denser description of the temporal unfolding of
events during these critical times. Finally, she has to dig deeper and collect information
about the perceptions and motivations of major actors. The data generation process in the
CPT approach is not only more inductive in comparison to the COV approach, but the
separation between data generation and data analysis is also less clear-cut. Nevertheless,
the functional equivalents to scores for the variables in the COV approach are
“comprehensive story lines”, “smoking guns”, and “confessions”. From the
comprehensive story lines, the scholars extract “causal chains” and “causal
conjunctions”; detailed descriptions of critical situations lead to strong evidence for a
dense connection between a cause and an effect (corresponding to the observation of a
“smoking gun”), and “confessions” provide deep insights into the perceptions and
motivations of major actors. These kinds of condensed empirical information have to be
combined with counterfactual thought experiments and/or with theoretical reflection on
the working of causal mechanisms and process dynamics in order to identify those
configurations of conditions and/or mechanisms that are individually necessary and
jointly sufficient for making the outcome possible.
In the CON approach, the sort of information required is delineated by expectations
(propositions, hypotheses and predictions) deduced from the theories that have been
selected and specified ex-ante. This is to some extent similar to the COV approach.
Nevertheless, in this approach, the information is not transformed into variable scores but
is used to determine whether the formulated expectations are confirmed or contradicted.
As a result, the investigator obtains a set of confirmations and/or contradictions for each
of the theories. As a second analytical step, he uses the differences among the theories
with respect to the level of congruence between expectations and observations either for
drawing conclusions about the relative importance of the selected theories in explaining
the case(s) or for combining the theories into a comprehensive explanation.

13

Electronic copy available at: https://ssrn.com/abstract=2105542

Generalization

Conclusions beyond the cases under investigation are usually discussed under the
heading of “generalization” – we follow this practice, although one of the main messages
of our book is that “generalization” means something quite different within the three case
study approaches.
Drawing conclusions within the COV approach is similar to the understanding of
generalization in large-N studies; we therefore call it “statistical generalization”. The
researcher draws conclusions from the selected and investigated cases to a wider
population of cases.
It is important to realize that the CPT approach does not strive for this kind of
generalization but for something that we call “possibilistic generalization”. The findings
of a CPT case study lead to knowledge about the causal configurations (combinations of
causal conditions or social mechanisms) that make specific outcomes possible. The
configurations of conditions and/or mechanisms that the researcher identifies as
necessary and sufficient for an outcome within the cases under investigation are used to
elucidate the set of “potential” configurations (all logically possible combinations of the
identified conditions and mechanisms) and/or the set of “proven” causal configurations.
The first set is helpful for developing “typological theories” inductively; the second set
includes all those configurations that have been shown to lead to the outcome of interest.
Within the CON approach, the researcher uses the insights gained in the case study for
the debate on the relevance of theoretical approaches in the broader scientific discourse.
The impact that the case study might have on this theoretical discourse depends on how
“crucial” the selected case is for the theories that “populate” the scientific discourse.

14

Electronic copy available at: https://ssrn.com/abstract=2105542

Table 1: Three explanatory approaches in case study research
Co-Variational Analysis Causal-Process Tracing Congruence Analysis
(COV) (CPT) (CON)
Research Does variable X make a difference? What makes the outcome (Y) possible? Which explanatory approach
questions Testing whether different values of X Revealing the temporal interplay among provides more/new insights?
and lead to different outcomes conditions or mechanisms that lead to Comparing the descriptive and
research goals specific outcomes explanatory merits of different theories
Focus Independent variables as: Causal configurations as: Comprehensive theories
- factors of influence sequential and situational combinations that compete with
- factors that have an autonomous of causal conditions or social mechanisms and/or complement
influence each other
Selection of Select multiple cases Select one or more cases Select multiple theories
cases and according to: according to: according to:
theories - strong differences in respect to the - their accessibility, AND - their place in the scientific discourse,
independent variable of interest, - the practical or theoretical relevance of AND
AND the outcome - the researcher’s theoretical aspiration
- high similarity in respect to control Selection of one or more cases
variables Selection of one or more cases according to the ex-ante “likeliness”
sequentially: of cases in respect to the selected
1. “positive” case(s) theories
2. “possible” case(s)
Data generation Observations: Observations: Observations:
Information corresponding to the - Information on the temporal unfolding Information corresponding to the
indicators specified for the variables of the process; expectations
- information on spatial-temporal (propositions, hypotheses, predictions)
proximity between causes and deduced from theories
consequences;
- information on perceptions and
Resulting data: motivations of actors Resulting data:
Scores of each variable for all cases Resulting data: A set of confirmations and/or
- Causal chains + conjunctions contradictions for each theory
- Smoking guns; - Confessions
Data analysis Necessary content of data: Necessary content of data: Necessary content of data:
= drawing Co-variation among scores of the - Comprehensive story line Differences among the theories in
causal dependent variable (Y) and scores of - Smoking guns respect to the level of congruence
inferences for the independent variable of interest - Confessions between expectations and observations
the cases under (X) Conclusion:
investigation Conclusion: Conclusion: - Relative importance
X has a causal effect on Y Identification of configurations of of selected theories
conditions and/or mechanisms that are - Comprehensive explanation
necessary and together sufficient for the through a combin. of theories
Further necessary conditions for outcome
conclusions: Further necessary tools for drawing
No theoretically plausible co-variation conclusions:
among scores of the dependent Counterfactuals AND/OR
variable and scores of other Theoretical concepts of mechanisms and
independent (control) variables process dynamics
Generalization Statistical generalization Possibilistic generalization Theoretical generalization
= drawing Drawing conclusions about the causal Drawing conclusions from the identified Drawing conclusions from the
conclusions effect of X on Y from the selected causal configuration(s) and mechanisms explanatory power of theories in more
beyond the cases to a population of cases that are for an outcome to the set of potential or less ‘crucial’ cases to the relevance
cases under similar in respect to all control configurations and/or to the set of proven of theories in the scientific discourse
investigation variables causal configurations and mechanisms

15

Electronic copy available at: https://ssrn.com/abstract=2105542

Conclusion
This paper deals with explanatory case study research. It argues that there are two rather
than one alternative to the much discussed comparative case study designs. Both
alternatives deal with within-case analysis but have until now been discussed under the
same label, process tracing. This has disguised the fact that they are birds of different
feathers. Hence we arrive at three approaches to explanatory case study research and we
elucidate the distinctive affinities of each of the approaches in terms of research goals and
questions, case and theory selection, data generation and analysis and directions of
generalization. We believe that our three-fold typology helps to systematize and
therewith to clarify the methodological debate on explanatory case study research, but
perhaps even more importantly helps case study research to conduct more compelling,
because internally more consistent, explanatory case studies.

Bibliography
Bennett, A. (2008). Process Tracing: A Bayesian Perspective. In Box-Steffensmeier, J., H. E.
Brady and Collier, D. (eds.), The Oxford Handbook of Political Methodology. Oxford:
Oxford University Press, 702–21.
Bennett, A. (2010). Process Tracing and Causal Inference. In Brady, H. E. and D. Collier (eds.)
Rethinking Social Inquiry. Diverse Tools, Shared Standards, 2nd ed. Lanham: Rowman
& Littlefield, 207-19.
Blatter, J. and T. Blume (2008a). “In Search of Co-Variance, Causal Mechanisms or Congruence?
Towards a Plural Understanding of Case Studies.” Swiss Political Science Review
14(2): 115–56.
Blatter, J. and T. Blume (2008b). Co-Variation and Causal Process Tracing Revisited:
Clarifying New Directions for Causal Inference and Generalization in Case Study
Methodology. Qualitative Methods – Newsletter of the American Political Science
Association Organized Section for Qualitative and Multi-Method Research 6(1): 29–34.
Blatter, J. and M. Haverland (2012). Designing Case Studies. Explanatory approaches in
Small-N Research. Houndsmills Basingstoke: Palgrave Macmillan
Brady, H. E. and D. Collier, eds (2004). Rethinking Social Inquiry: Diverse Tools, Shared
Standards. Lanham: Rowman & Littlefield.
Brady, H. E. and D. Collier, eds (2010). Rethinking Social Inquiry: Diverse Tools, Shared
Standards. Lanham: Rowman & Littlefield (2nd edition).
Eckstein, H. (1975). “Case Study and Theory in Political Science.” In Handbook of Political
Science, eds F. Greenstein and N. Polsby. Reading: Addison-Wesley, 79–138.
Faure, A. (1994). “Some Methodological Problems in Comparative Politics.” Journal of
Theoretical Politics 6(3): 307–22.
Frendreis, J. (1983). “Explanation of Variation and Detection of Covariation: The Purpose and
Logic of Comparative Analysis.” Comparative Political Studies 16(2): 255–72.

16

Electronic copy available at: https://ssrn.com/abstract=2105542

Geddes, B. (1990). “How the Cases You Choose Affect the Answer You Get: Selection Bias in
Comparative Politics.” Political Analysis 2(1): 131–50.
George, A. and A. Bennett (2005). Case Studies and Theory Development in the Social Sciences.
Cambridge (MA): MIT Press.
George, A. L. (1979). Case studies and theory development: The method of structured, focused
comparison. In Diplomacy: New approaches in history, theory and policy., ed. Paul
Gordon Lauren, New York: Free Press, 43/68.
George, A. T.J. McKeown (1985). “Case studies and theories of organizational decision making”.
Advances in Information Processing in Organizations 2 (1): 21-58
Goertz, G. and J. Mahoney (forthcoming). A Tale of Two Cultures: Contrasting Qualitative and
Quantitative Paradigms. Princeton: Princeton University Press.
Hall, P. (2003). “Aligning Ontology and Methodology in Comparative Politics.” In Comparative
Historical Analysis in the Social Sciences, eds J. Mahoney and D. Rueschemeyer.
Cambridge: Cambridge University Press, 373–404.
Hall, P. (2004). “Letter from the President. Beyond the Comparative Method”,
Newsletter of the Organized Section in Comparative Politics of the American
Political Science Association 15 (2): 1-4.
Hall P. (2008) “Systematic Process Analysis: When and How To Use It, European Political
Science, 7(3) 304-317, also published in European Management Review, 2006, 3(1):
24–31.
King, G., R. O. Keohane, and S. Verba (1994). Designing Social Inquiry: Scientific Inference in
Qualitative Research. Princeton: Princeton University Press.
Lieberson, S. (1991). “Small N's and Big Conclusions: An Examination of the Reasoning Based
on a Small Number of Cases.” Social Forces 70(2): 307–20.
Lieberson, S. (1994). “More on the Uneasy Case for Using Mill-Type Methods in Small-N
Comparative Studies.” Social Forces 72(4): 1225–37.
Lijphart, A. (1971). “Comparative Politics and the Comparative Method.” American Political
Science Review 65(3): 682–93.
Lijphart, A. (1975). “The Comparable-Cases Strategy in Comparative Research.” Comparative
Political Studies 8(2): 158–77.
Mahoney, J. (1999). “Nominal, Ordinal, and Narrative Appraisal in Macrocausal Analysis.” The
American Journal of Sociology 104(4): 1154–96.
Mill, J. (1875). A System of Logic, Ratiocinative and Inductive: Being a Connected View of the
Principles of Evidence and the Methods of Scientific Investigation. London: Longmans,
Green, Reader, and Dyer.
Moravcsik, A. (1998). The Choice for Europe: Social Purpose and State Power from Messina to
Maastricht. London: Routledge.
Nichols, E. (1986) Skocpol on revolution: Comparative analysis versus historical conjuncture,
Comparative Social Research, 9, 163-186
Przeworski, A. and H. Teune (1970). The Logic of Comparative Social Inquiry. New York: Wiley
& Sons.
Ragin, C. (2008). Redesigning Social Inquiry: Fuzzy Sets and Beyond. Chicago: University of
Chicago Press.
Sewell, W. (2005). The Logics of History: Social Theory and Social Transformation. Chicago:
Chicago University Press.
Skocpol, T. (1979). States and Social Revolutions: A Comparative Analysis of France, Russia,
and China. Cambridge: Cambridge University Press.
Skocpol, Theda and Margaret Somers. 1980. The Uses of History in Macro-social Inquiry.
Comparative Studies in Society and History, 22: 174-197

17

Electronic copy available at: https://ssrn.com/abstract=2105542

Skocpol, Theda 1986. Analyzing causal configurations in history: a rejoinder to Nichols T
Skocpol Comparative Social Research, 1986, 9, 187-194
van Evera, S. (1997). Guide to Methods for Students of Political Science. Ithaca (NY): Cornell
University Press.
Wendt, A. (1999). Social Theory of International Politics. Cambridge: Cambridge University
Press.
Yin, R. (1984). Case Study Research. Thousand Oaks: Sage.

18

Electronic copy available at: https://ssrn.com/abstract=2105542