The Distinctiveness of
Comparative Social Science
'Thinking without comparison is unthinkable. And, in the absence of
comparison, so is all scientific thought and scientific research" (Swanson
1971:145). Most social scientists today would agree with this observation,
although some might be tempted to substitute the phrase variables and relationships
for the word comparison. Virtually all empirical social research
involves comparison of some sort. Researchers compare cases to each other;
they use statistical methods to construct (and adjust) quantitative comparisons;
they compare cases to theoretically derived pure cases; and they
compare cases' values on relevant variables to average values in order to assess
covariation. Comparison provides a basis for making statements about
empirical regularities and for evaluating and interpreting cases relative to
substantive and theoretical criteria. In this broad sense, comparison is central
to empirical social science as it is practiced today. Lieberson (1985 :44)
states simply that social research, "in one form or other, is comparative
research."
While virtually all social scientific methods are comparative in this broad
sense, in social science the term comparative method typically is used in a
narrow sense to refer to a specific kind of comparison—the comparison of
large macrosocial units. In fact, the comparative method traditionally has
been treated as the core method of comparative social science, the branch of
social science concerned with cross-societal differences and similarities
(Easthope 1974). Despite this tradition, there is substantial disagreement today
concerning the distinctiveness of comparative social science in general
1
1
C O M P A R A T I V E S O C I A L S C I E N C E
and the comparative method in particular. Several comparativists have objected
to the idea that comparative social science is distinctive in any important
respects from social science in general (Grimshaw 1973:18).
Smelser (1976:2-3), for example, claims that comparative social scientific
inquiry is not a "species of inquiry independent from the remainder of
social scientific inquiry" and that "the analysis of phenomena in evidently
dissimilar units (especially different societies or cultures) should have no
methodological problem unique to itself." According to Smelser (1976:5),
this continuity between comparative and noncomparative work exists because
their respective goals are identical—to explain social phenomena by
establishing controls over the conditions and causes of variation. (See also
Aimer 1973:50.) Any technique that furthers the goal of explaining variation,
according to this reasoning, is a comparative method. This includes virtually
all analytic methods used by social scientists (see Bailey 1982).
This position, that there is nothing truly distinctive about comparative
social science and that virtually all social scientific methods are comparative
methods, is sound, and it is attractive because it suggests that social science
subdisciplines are united by their methods. The argument is favored by
many comparativists, in fact, because the emphasis on continuities between
comparative and noncomparative work supports the idea that comparative
social science is as scientific as its siblings. This position overlooks the fact,
however, that there are important differences between the orientations of
most comparativists and most noncomparativists and these differences have
important methodological consequences. While it is true that the logic of
social science is continuous from one subdiscipline to another, the peculiarities
of comparative social science make it an ideal setting for an examination
of key issues in methodology. In fact, I argue that a lot can be gained
from exaggerating the distinctive aspects of comparative work and that these
lessons can be applied to other social science subdisciplines as well.
The most distinctive aspect of comparative social science is the wide gulf
between qualitative and quantitative work. It is wider in comparative social
science than in perhaps any other social science subdiscipline. In part this is
because its qualitative tradition is dominant, the opposite of the situation in
most other fields. Over the last twenty years, some of the most celebrated
works in the social sciences (from Moore's Social Origins of Dictatorship and
Democracy to Wallerstein's Modern World System) have come out of this
tradition, making it appear continuous with the grand theorizing of such
classical scholars as Durkheim and Weber.
2
C O M P A R A T I V E S O C I A L S C I E N C E 3
More fundamental to the gulf, however, is the fact that several other divisions
coincide with the qualitative/quantitative split in comparative social
science and reinforce it. Qualitative researchers tend to look at cases as
wholes, and they compare whole cases with each other. While cases may be
analyzed in terms of variables (for example, the presence or absence of a
certain institution might be an important variable), cases are viewed as configurations—as
combinations of characteristics. Comparison in the qualitative
tradition thus involves comparing configurations. This holism contradicts
the radically analytic approach of most quantitative work.
Not only is the qualitative tradition oriented toward cases as wholes, as
configurations, but it also tends to be historically interpretive. The term interpretive
is used in a restricted sense here. Often, the term is used to describe
a type of social science that is only remotely empirical and concerned
primarily with problems of meaning or hermeneutics. In this book, interpretive
work is treated as a type of empirical social science: historically oriented
interpretive work attempts to account for specific historical outcomes
or sets of comparable outcomes or processes chosen for study because of
their significance for current institutional arrangements or for social life in
general. Typically, such work seeks to make sense out of different cases by
piecing evidence together in a manner sensitive to chronology and by offering
limited historical generalizations that are both objectively possible and
cognizant of enabling conditions and limiting means—of context. This definition
of interpretive work leans heavily on Weber (1949, 1975, 1977) but
makes more allowance for the possibility of historical generalization based
on examination of comparable cases. In this chapter I discuss these distinctive
characteristics and sketch the implications of these features for comparative
methodology. I begin by delineating the field.
THE BOUNDARIESAND GOALS OF
COMPARATIVE SOCIAL SCIENCE
There have been several attempts to delineate the boundaries of comparative
social science. Yet,there is still little agreement today concerning its domain.
Most attempts to delineate the field have emphasized its special data or its
special types of data. For reasons detailed below, this is a poor starting point.
I argue that comparative social science is better defined by its distinctive
goals.
It is common to define comparative research as research that uses compa-
3
4 C O M P A R A T I V E S O C I A L S C I E N C E
rable data from at least two societies. This definition emphasizes the fact that
the data of comparative social science are cross-societal. (See Andreski
1965:66; Armer 1973:49.) While this is an acceptable working definition
of comparative social science, most comparativists would find this definition
too restrictive. It excludes, for example, comparatively oriented case studies.
Tocqueville's Democracy in America is excluded, as is Durkheim's Elementary
Forms of the Religious Life. Many area specialists are thoroughly comparative
because they implicitly compare their chosen case to their own
country or to an imaginary but theoretically decisive ideal-typic case. Thus,
to define comparative social science in terms of its special data is a misleadingly
concrete way to delineate its boundaries.
Others have attempted to differentiate comparative social science by emphasizing
its multilevel character (as in Rokkan 1966:19-20). According to
Przeworski and Teune (1970:50-51), comparative work proceeds at two levels
simultaneously—at the level of systems (or macrosocial level) and at the
within-system level. According to their argument, any analysis that is based
only on macrosocial similarities and differences is not truly comparative,
even if this analysis includes an examination of aggregations of withinsystem
characteristics. For example, if an investigator uses system-level
variables (such as GNP per capita) to explain variation in a dependent variable
based on aggregations of individual-level data within each system (such
as literacy rates), the study would not qualify as a comparative study according
to Przeworski and Teune. Ideally, system-level variables should be used
to explain variation across systems in within-system relationships.
Alford's (1963) study of international variation in class voting qualifies as
a comparative study by these criteria because he uses system-level variables
(degree of industrialization and urbanization) to explain differences among
countries in within-system relationships (the strength of the relationship between
social class and party support). Walton's (1984) study of national revolts
in the Third World also conforms to this definition of comparative
work. He uses degree of incorporation into the world economy, a systemlevel
variable, to account for variation in the degree to which popularprotests
and state reactions to protest contributed to the coalescence of revolutionary
situations in six countries (seeespecially Walton 1984:188-197).
Few studies traditionally thought of as comparative, however, conform to
these strictures. Comparatively oriented case studies are excluded, as are
quantitative cross-national studies that use only aggregate, national-level
data. (Note that quantitative cross-national studies focus directly on cross-
C O M P A R A T I V E S O C I A L S C I E N C E 5
societal similarities and differences.) Przeworski and Teune's definition of
comparative inquiry as multilevel research is much more restrictive than
even the first definition considered here.
Both definitions are inadequate. Yet they suggest a tentative solution to
the problem of delineating comparative work. One level that invariablyplays
a big part in definitions of comparative work is the macrosocial level. It appears
in the first definition offered above in its emphasis on data from two
societies and in the second's emphasis on multilevel analyses, with one level
the macrosocial. The boundaries of comparative social science, therefore,
must be coterminous with a specific usage of macrosocial units.
It is not as a data category that macrosocial units are important to comparativists,
but as a metatheoretical category. What distinguishes comparative
social science is its use of attributes of macrosocial units in explanatory
statements. This special usage is intimately linked to the twin goalsof
comparative social science—both to explain and to interpret macrosocial
variation.
The importance of macrosocial units to explanation in comparative social
science is best understood by example. Consider an investigation which concludes
that a strong relationship between social class and party preference
exists in Great Britain because "Great Britain is an industrial society." This
conclusion concretizes the term society by providing an example (Great
Britain) and by implying that there are other societies, some of which are
industrial and some of which are not. If the investigator had concluded instead
that the relationship exists because "citizens vote their pocketbooks" or
because "the relations of production shape political consciousness," then he
or she would have avoided concretizing any macrosocial unit and thereby
would have avoided engaging in comparative social science.
This direct, empirical implementation of abstract, macrosocial units is a
metatheoretical act, and it separates comparativists from noncomparativists.
In order to compare societies or any other macrosocial unit, the comparativist
must identify them by name. The comparativist thus assumes, at least
implicitly, that macrosocial units are real and then defines them, sometimes
by default, in the course of research. The fact that the difference between
comparativists and noncomparativists is a metatheoretical difference based
on the special goals of comparative social science has been obscured by the
tendency of all social scientists to claim that they study societies or that social
science is the study of society. For the noncomparativists, however, macrosocial
units tend to remain abstractions. Noncomparativists can assure
C O M P A R A T I V E S O C I A L S C I E N C E
themselves that the patterns and processes they study exist in a society; the
concept need not be operationalized explicitly. For the comparativists, however,
macrosocial units impinge on their work in a fundamental manner.
Rarely are these large, encompassing units defined. (Parsons 1977 and
Marsh 1967 are exceptions.) In his discussion of the distinctiveness of comparative
work, for example, Grimshaw (1973:4) states, "I will defer discussion
of what constitutes a [macrosocial] system/' This reluctance is not uncommon;
most comparativists are more interested in making comparisons
than in defining the objects of their comparisons (see Andreski 1965:66).
The fact remains, however, that comparativists compare macrosocial units;
they must be operationalized in the course of comparative work.
At a very general level, comparativists are interested in identifying the
similarities and differences among macrosocial units. This knowledge provides
the key to understanding, explaining, and interpreting diverse historical
outcomes and processes and their significance for current institutional
arrangements. Cross-societal similarities and differences for many social
scientists constitute the most significant feature of the social landscape, and,
consequently, these researchers have an unmistakable preference for explanations
that cite macrosocial phenomena. This tendency is reinforced by the
fact that the goals of comparative social scientists typically extend beyond an
interest in simply cataloging and explaining cross-societal similarities and
differences. Most comparativists, especially those who are qualitatively oriented,
also seek to interpret specific experiences and trajectories of specific
countries (or categories of countries). That is, they are interested in the cases
themselves, their different historical experiences in particular, not simply in
relations between variables characterizing broad categories of cases. This interest
reinforces the tendency to use macrosocial attributes in explanatory
statements.
The decision to study macrosocial variation and to use explanatory statements
citing macrosocial properties is, of course, a conscious choice, shaped
in large part by the enduring reality of countries, nations, states, and other
large (and imposing) political entities. As long as social scientists continue to
be influenced by their social and historical contexts and continue to try to
interpret them, they will use macrosocial attributes in their explanations of
social phenomena. It is possible to imagine a social science devoid of explanatory
statements citing macrosocial phenomena. A totally psychologized social
science, for example, might attempt to disavow such explanations. It is
unlikely, however, that social scientists will lose interest in interpreting na-
6
C O M P A R A T I V E S O C I A L S C I E N C E
tional and international events and processes and thereby divorce themselves
from significant features of their social contexts. (In any event, to do so
would be to deny the social origins and bases of social science.) Thus, macrosocial
units are central to the practice of comparative social science because
they are an essential ingredient of the explanations comparativists offer.
A NOTE OF CAUTIONON UNITSOF ANALYSIS
It would be wrong at this point to conclude simply that comparativists differ
from noncomparativists in their "chosen unit of analysis." The example supplied
previously suggests that any data unit can be used in comparative research.
All that matters is how the results of research are understood. The
fact that the explanations of comparative social science tend to be crosssocietal
and cite macrosocial phenomena, however, implies that the question
of units is relevant.
Very little continuity exists, however, in discussions of units of analysis
offered by comparatively oriented social scientists. An important source of
this lack of continuity is the simple fact that the term unit of analysis is used
to describe two very distinct metatheoretical constructs. Sometimes unit
of analysis is used in reference to data categories. In a quantitative crossnational
study of economic dependency and economic development, for example,
an investigator might state that the unit of analysis is the nation-state
because the data are collected at that level. At other times, however, the term
unit of analysis is used in reference to theoretical categories. Wiener (1976),
for example, in a review of Harrington Moore's Social Origins of Dictatorship
and Democracy (1966), states that Moore's unit of analysis is "class."
Wallerstein (1974, 1979,1980, 1984) argues in various works that there is
only one valid unit of analysis in comparative social science: the "world system."
Upon closer examination, however, one finds that Moore's cases are
different countries and Wallerstein's discussion of the modern world system
is rife with references to nation-states and comparisons of, for example, core
countries and peripheral countries.
The fact that the term unit of analysis has been used in reference to both
data categories and theoretical categories has created a great deal of confusion
in the field of comparative social science. Some followers of Wallerstein,
for example, have attacked those who use the nation-state as a unit of analysis
in the data category sense, arguing that this practice violates worldsystems
theory and results in meaningless tests of its propositions. (See, for
7
C O M P A R A T I V E S O C I A L S C I E N C E
example, Bach 1977.) Other researchers have attempted to use the modern
world system as a unit of analysis in the data category sense and have examined
cycles and trends in the world economy as a whole. (See, for example,
Bergesen 1980 and McGowan 1985.) It is clear from Wallerstein's discussion
and from his actual analyses of the world system, however, that his argument
is that the world system is the only valid explanatory unit, not the
only valid data unit.
This tension between the two meanings of unit of analysis has bedevilled
the comparative social science literature at least since the early 1960s. Issues
associated with the aggregation problemhavecompounded the terminological
difficulties and confusion. Allardt (1966:339-341), for example, attempted
to draw a distinction between "data units" and "analytical units/' arguing
that the latter are more theoretically relevant. In a similar vein, Scheuch
(1966:164) argued that comparativists should distinguish between "unitsof
observation" (see also Walton 1973:176) and "units ofinference/' In an early
attempt to formulate a methodological position, Hopkins and Wallerstein
(1970:183) contrasted "research sites" and "theoretical units." Several researchers
attempted to clarify the situation by limiting their comments to
"units of comparison" (Eisenstadt 1966:86; Etzioni and Dubow 1970:7;
Czudnowski 1976:27). Finally, Przeworski and Teune (1970:8, 49-50)
attempted to distinguish between "levels of observation" and "levels of
analysis."
Most of these discussions were stimulated by the ambiguity associated
with the term unit of analysis. For most noncomparative social scientists,
the term presents no specialproblems. Their analyses and their explanations
typically proceed at one level, the individual or organizational level. This
is rarely the case in comparative social science, where the analysis often
proceeds at one level (perhaps the individual level, as in the preceding example)
and the explanation is couched at another level (usually the macrosocial
level). Of course, this duality exists in other types of social science,
and the methodological issues raised here apply to these areas as well. The
duality is most pronounced,however, in comparativesocial science, which is
one of the features that makesit an idealarena for methodological discussion.
To clarify the unit of analysis question in comparative social science, it is
necessary to distinguish between observational units and explanatory units.
This distinction follows my discussion concerning the two meanings of unit
of analysis—as a data category and as a theoretical category. Observational
8
C O M P A R A T I V E S O C I A L S C I E N C E 9
unit refers to the unit used in data collection and data analysis; explanatory
unit refers to the unit that is used to account for the pattern of results obtained.
In the class voting example mentioned above, the observational unit
is the individual (the relationship is based on individual-level data) and the
explanatory unit is societal.
METHODOLOGICALCONSEQUENCES
The explanation that there is a strong relationship between social class and
party preference in a sample of British voters because "Great Britain is an
industrial society" implies that societies can be identified, that they can be
classified as either industrial or not industrial, and that in industrial societies
there is a strong relationship between social class and party preference,
while in nonindustrial societies there is no such relationship. Because societies
are (at least apparently)identifiable,an investigator conceivablycould
draw up a list of them, classify them as industrial and not industrial (or at
least measure the degree to which each society is industrial), and then examine
the degree to which the more industrial societies agree in manifesting
a consistent relationship between social class and party choice and also the
degree to which the less industrial societies agree in manifesting a weaker
relationship. If these two patterns of agreement can be established, then the
general statement (that in industrial societies there is a strong relationship
between social class and party preference) used to explain the particular instance
(the relationship observed in Great Britain) is supported.
Unfortunately, social scientific investigation is rarely this simple. There
are many practical problems associated with establishing cross-societal demonstrations
such as the one described above. Most of these practicalproblems
concern the comparability of relatively dissimilar societies. This concern for
comparability derives ultimately from the fact that the cases (say,countries)
which comparativists study have known histories and identities. They are
not anonymous, disembodied observations. In the preceding investigation,
for instance, a researcher familiar with the relevant cases might have doubts
about the cross-societal comparability of measures of class positions or
about the identification of parties with social classes. An investigator might
also have doubts about the classificationof societies as industrial and not industrial
or about ordinal and interval measures of degree of industrialization.
These measurement problems are very important, and they have ab-
1 0 C O M P A R A T I V E S O C I A L S C I E N C E
sorbed the attention of comparative social scientists for some time. In fact,
many discussions of comparative methods have concerned these issues almost
exclusively.
At a more basic level, it is difficult to evaluate explanatory statements of
comparative social science because the number of relevant units available for
such assessments is often limited by empirical constraints. Even the investigator
who claims that he or she is interested in all societies, and defines
societies as all contemporary nation-states, encounters serious statistical
problems if a quantitative analysis of these cases is attempted. A seemingly
large set of more than one hundred nation-states can be reduced by half if
there are problems with missing data. Often, the remaining cases are not
representative of the original hundred-plus nation-states, much less of all
societies (or all macrosocial systems). This problem is apparent in the hypothetical
research described above. There are many societies, both industrial
and nonindustrial, that are not democratic. Thus, any attempt to assess the
strength of the relationship between social class and party preference in
these countries would be questionable, if not misguided. Furthermore, the
definition of democratic society is problematic and ideologically charged.
Theoretical strictures also may reduce the number of relevant cases. In
the hypothetical analysis of more and less industrial societies discussed
above, for example, it is possible that the general statement (that social class
shapes party preference only in industrial societies) is theoretically meaningful
only when applied to democratic countries with a feudal past. If this
were the case, then the investigator would first draw up a list of democracies
with a feudal past and then distinguish between more and less industrial
countries within this set. Generally speaking, the greater the theoretical or
empirical specificity, the smaller the number of cases relevant to the investigation.
The smaller the number of relevant cases, the greater the likelihood
that the investigator will find it difficult to evaluate an explanatory statement
in a way that conforms to the standards of mainstream social science, especially
its quantitative branch.
Sometimes there are more explanations of a certain phenomenon than
there are examples of it because these strictures reduce the number of relevant
cases to a mere handful. In such investigations it is impossible to adjudicate
among competing explanations. In the language of the statistical
method, the use of societies in explanatory statements often presents serious
degrees-of-freedom problems, for the number of relevant explanatory variables
may far exceed the number of cases. From the perspective of main-
C O M P A R A T I V E S O C I A L S C I E N C E 1 1
stream social science, therefore, comparative social science is severely deficient
in the opportunities it presents for testing theory.
But many comparativists, especially those who are qualitatively oriented,
are not often involved in "testing" theories per se. Rather, they apply theory
to cases in order to interpret them. Because the explanatory statements of
comparative social science cite attributes of macrosocial units, objects with
known identities and histories figure prominently in the conduct of inquiry.
Thus, it is very difficult to treat these units simply as the undifferentiated
raw material of empirical social science. There is an ever present pressure to
take into account and to explain the particularity of specific cases, which in
turn requires the use of case-oriented methods sensitive to time, place,
agency, and process.
Recall also that one of the distinctive goals of comparative social science is
to interpret significant historical outcomes. From the perspective of mainstream
social science this goal imposes very restrictive boundaries on social
research, dramatically reducing the number of relevant observations. In essence,
when a comparativist interprets significant historical outcomes, he or
she selects extreme values on a more general dependent variable (for instance,
social revolution is an extreme value on a general measure of social
turmoil) and studies the cases with these extreme values exclusively. This
practice is justified by the qualitative break that exists between extreme values
and lesser values on what might be viewed by some as a continuum and
also by the cultural importance and historical significanceof these extreme
cases. Thus, the problem of having too few societies on which to test theory
is compounded by the fact that the interests and goals of comparative social
science (and scientists) often dictate the design of studies with a small number
of cases—too few to permit the application of any technique of statistical
comparison.
Most comparativists, in fact, are interested in questions that are limited,
substantively and historically. The questions they ask usually are much
more circumscribed than the abstract research question posed above concerning
the effect of industrialization on the strength of the relationship between
social class and party preference. In the typical comparative study,
only a small set of cases may provide the basis for empirical generalization.
Instances of social revolution, at least as defined by Skocpol (1979), for example,
are few. There are also only a few instances of successful antineocolonial
revolt. There are more cases of dependent industrial development
in the Third World today, but not so many that they can be studied
1 2 C O M P A R A T I V E S O C I A L S C I E N C E
easily with quantitative cross-national techniques. Yetthese and related topics
demand the attention of comparative social scientists. The fact that there
are few relevant instances of each phenomenon and that these instances have
known identities and histories (that is, known particularity) has a powerful
impact on the character of the research process.
ENTER THE COMPARATIVE METHOD
As the number of relevant observations decreases, the possibility of subjecting
arguments to rigorous statistical testing diminishes. Other methods
must be used. Smelser (1976:157) argues that the method of "systematic
comparative illustration" (a method he portrays as a crude approximation of
more sophisticated statistical methods) must be used when the number of
relevant cases is small: "This method is most often required in the comparative
analysis of national units or cultures." Smelser provides as one example
of the method of systematic comparative illustration Tocqueville's
three-way comparison of American, French, and English customs. Tocqueville
argued simply that the conditions these collectivities share (such as language
in the case of the English and the Americans) could not be used to
explain their differences and that differences could not be used to explain
similarities (Smelser 1976:158). In general, the technique of systematic
comparative illustration involves applications of Mill's (1843) method of
agreement and his indirect method of difference. (These case-oriented techniques
are discussed in detail in Chapter 3.)
In an earlier work, Smelser (1973) called this systematic analysis of similarities
and differences the comparative method and contrasted it with the
statistical method. In his more recent Comparative Methods in the Social
Sciences (1976), however, Smelser argues that, broadly speaking, virtually
all social scientific methods are comparative and that the method of systematic
comparative illustration is inferior to the statistical method as a comparative
method. It is inferior, according to Smelser, because it must be used
when the number of relevant cases is small and the possibility of establishing
systematic control over the sources of variation in social phenomena is reduced.
The possibilities for social scientific generalization are reduced.
In fact, the method that Smelser calls "the method of systematic comparative
illustration" is what social scientists traditionally have called the
comparative method. It forms the core of the case-oriented strategy and is
quite different from correlational methods which form the core of the vari-
able-oriented strategy (see Chapters 3 and 4). It is proper to call this method
the comparative method because it follows directly from asking questions
about empirically defined, historically concrete, large-scale social entities
and processes—the kinds of questions that comparative social scientists tend
to ask. Questions that necessarily lead to detailed analyses of relatively
small numbers of cases are asked in other types of social science, as well, but
this type of investigation is most common in comparative social science.
Once it is admitted that the comparative method derives its distinctiveness
from the special goals of comparative social science and that it is
most often a direct consequence of engaging in this enterprise, the special
features of the comparative method can be delineated.
THE LOGIC OF THE COMPARATIVE METHOD
"It is surprising, for all that has been said about the value of comparison,
that a rigorous comparative methodology has not emerged. The reason for
this lack may be the great difficulties that a rigorous comparative methodology
would impose" (Porter 1970:144). Smelser might argue that a rigorous
comparative method is a contradiction in terms because, by definition,
the comparative method is used only when the number of relevant cases is
too small to allow the investigator to establish statistical control over the
conditions and causes of variation in social phenomena. While the number
of cases relevant to an analysis certainly imposes constraints on rigor, often
it is the combinatorial nature of the explanations of comparative social science
and the holistic character of the comparative method that militate
against this kind of rigor.
Most comparativists, especially those who are qualitatively oriented, are
interested in specific historical sequences or outcomes and their causes across
a set of similar cases. Historical outcomes often require complex, combinatorial
explanations, and such explanations are very difficult to prove in a
manner consistent with the norms of mainstream quantitative social science.
When causal arguments are combinatorial, it is not the number of
cases but their limited variety that imposes constraints on rigor.
When qualitatively oriented comparativists compare, they study howdifferent
conditions or causes fit together in one setting and contrast that with
how they fit together in another setting (or with how they might fit together
in some ideal-typic setting). That is, they tend to analyze each observational
entity as an interpretable combination of parts—as a whole. Thus, the ex-
13COMPARATIVESOCIALSCIENCE
1 4 C O M P A R A T I V E S O C I A L S C I E N C E
planations of comparative social science typically cite convergent causal conditions,
causes that fit together or combine in a certain manner.
A simple example illustrates this practice. A comparativist might argue
that social class and party preference are strongly related to each other in a
sample of British voters not simply because Great Britain is an industrial
society but also because it has a long history of class mobilization and conflict
which coincided with the development of its current political system. In
effect, this explanation cites three convergent conditions: (1) a history of
class struggle (2) coinciding with polity maturation (3) in a country that has
been industrialized for a long time. It is their combined effect that explains
the enduring individual-level relationship between social class and party
preference. The argument would be that this configuration of causes explains
the observed a
To evaluate this argument rigorously, it would be necessary to find instances
(among democratic countries) of all the logically possible combinations
of the three conditions and then to assess the relationship between social
class and party preference in each combination. Each logically possible
combination should be examined because the argument is that it is the coincidence
of these three conditions that explains the association. If the expected
relationship is obtained only when these three conditions coincide,
and if all instances of such concurrence manifest the predicted relationship,
then the general statement would be supported.
It would be difficult to evaluate this argument because instances of all
logically possible combinations of conditions are not available.A completely
rigorous assessment would require the identification of democratic countries
with eight different combinations of characteristics. (There are eight different
logically possible combinations of three dichotomies.) Each different
combination is conceived as a different situation, a different totality, not
simply as a different collection of values on three variables. Some of these
combinations, however, while logically possible, do not exist. At best, the
investigator would be able to examine the combinations that do exist
and assess the relationship between class and party within each of these
configurations.
While this simple example shows the limitations placed on the comparative
method as a consequence of its holistic nature, it also illustrates key
features of the method. As already noted, the comparative method
to configurations of conditions; it is used to determine the different
binations of conditions associated with specific outcomes or processes. More-
C O M P A R A T I V E S O C I A L S C I E N C E 1 5
over, the comparative method is based on "logical methods" (see Gee 1950);
it uses two of Mill's methods of inductive inquiry: the method of agreement
and the indirect method of difference (Mill 1843; see also Skocpol 1979 :36;
Skocpol and Somers 1980; Zelditch 1971; Ragin and Zaret 1983). These
methods use all available and pertinent data concerning the preconditions of
a specific outcome and, by examining the similarities and differences among
relevant instances, elucidate its causes.
Because the comparative method has this character, statistical criteria are
less important to this approach. This means that the comparative method
does not work with samples or populations but with all relevant instances of
the phenomenon of interest and, further, that the explanations which result
from applications of the comparative method are not conceived in probabilistic
terms because every instance of a phenomenon is examined and accounted
for if possible. Consequently, the comparative method is relatively
insensitive to the relative frequency of different types of cases. For example,
if there are many instances of a certain phenomenon and two combinations
of conditions that produce it, both combinations are considered equally valid
accounts of the phenomenon regardless of their relative frequency. If one is
relatively infrequent, an application of the statistical method to this same set
of data might obscure its existence. The comparative method would consider
both configurations of conditions relevant since both result in the phenomenon
of interest.
Smelser's argument implies that the comparative method is inferior to the
statistical method. Is it? The comparative method is superior to the statistical
method in several important respects. First, the statistical method is not
combinatorial; each relevant condition typically is examined in a piecemeal
manner. Thus, for example, the statistical method can answer the question:
what is the effect of having a history of class struggle net of the effect of
industrialization? But it is difficult to use this method to address questions
concerning the consequences of different combinations of conditions (that is,
to investigate situations as wholes). To investigate combinations of conditions,
the user of the statistical method must examine statistical interactions.
The examination of a large number of statistical interactions in variableoriented
studies is complicated by collinearity and by problems with scarce
degrees of freedom, especially in comparative research where the number of
relevant cases is often small. An exhaustive examination of different combinations
of seven preconditions, for example, would require a statistical
analysis of the effects of more than one hundred different interaction terms.
1 6 C O M P A R A T I V E S O C I A L S C I E N C E
Second, applications of the comparative method produce explanations
that account for every instance of a certain phenomenon. True, these explanations
may contain interpretive accounts of the particularity of one or more
deviating cases, but at least the comparative method automatically highlights
these irregularities and requires the investigator to propose explanations of
them. This concern makes the comparative method more consistent with the
goal of interpreting specific cases and addressing historical specificity. This
feature of the comparative method also makes it especially well suited for
the task of building new theories and synthesizing existing theories.
Third, the comparative method does not require the investigator to pretend
that he or she has a sample of societies drawn from a particular population
so that tests of statistical significance can be used. The boundaries of a
comparative examination are set by the investigator (see Walton 1973:
174-175); they are not coterminous with the boundaries of an arbitrarily
defined or (more typically) undefined population of societies or points in
time or events in societies.
Finally, the comparative method forces the investigator to become familiar
with the cases relevant to the analysis. Tomake meaningful comparisons
of cases as wholes, the investigator must examine each case directly and
compare each case with all other relevant cases. The statistical method, by
contrast, requires the investigator only to disaggregate cases into variables
and then to examine relationships among variables, not to conduct a direct
examination of the differences and similarities among cases considered as
configurations of characteristics (that is, as meaningful wholes).
In short, the comparative method is not a bastard cousin of the statistical
method. It is qualitatively different from the statistical method, and it is
uniquely suited to the kinds of questions that many comparativists ask.
THE QUALITATIVE/QUANTITATIVE SPLIT IN COMPARATIVE
SOCIAL SCIENCE
As outlined here, the comparative method is essentially a case-oriented
strategy of comparative research (see Chapter 3). The focus is on comparing
cases, and cases are examined as wholes—as combinations of characteristics
(Ragin and Zaret 1983). This orientation distinguishes it from mainstream
statistical methodology. Of course, not all social scientists who call themselves
comparativists use the comparative method as presented in this chapter.
Many use a variable-oriented strategy which conforms to the methodological
norms of mainstream social science with its emphasis on variables
C O M P A R A T I V E S O C I A L S C I E N C E 1 7
and their interrelationships. The usual goal of variable-oriented investigations
is to produce generalizations about relationships among variables, not
to understand or interpret specific historical outcomes in a small number of
cases or in an empirically defined set of cases (see Chapter 4). Combined
strategies also exist, but close examination usually shows that studies using
combined strategies tend to fall into one of the two camps (see Chapter 5).
Examples of combined strategies include variable-oriented analyses supplemented
with case studies (as in Paige 1975 and Stephens 1979) and case
studies reinforced with quantitative analyses (as in Shorter and Tilly 1974).
The dichotomized nature of comparative work (case-oriented comparative
study versus variable-oriented analysis) makes it an ideal setting for examining
methodological issues—especially the gap between qualitative and
quantitative orientations and how this gap might be bridged. Comparative
work is the one branch of contemporary American social science that accords
high status to the qualitative analysis of a small number of cases. In
comparative social science, the variable-oriented strategy poses a challenge
to traditional qualitative approaches. In other social science research areas,
by contrast, the opposite is true. Thus, in comparative social science there is
an established case-oriented tradition that can be directly contrasted with a
growing variable-oriented tradition.
In comparative social science the qualitative tradition is strong because
other methodological divisions coincide with the qualitative/quantitative
split. As the preceding discussion of the logic of the comparative method
shows, qualitative researchers tend to ask historically and empirically defined
questions and typically answer these questions historically, in terms of
origins. Thus, qualitative comparative researchers are both holistic and interpretive
in their approach to comparative materials.
The split between qualitative and quantitative work in comparative social
science is further aggravated by the fact that all comparativists are concerned
with questions of direct relevance to macrosocial units with meaningful social
identities (nation-states, for example). These identities are crucial to
qualitative researchers, whereas they sometimes confound the workof those
who do quantitative cross-national work. (For example, Kuwait is always a
troublesome outlier in studies of economic dependence and development.)
This aspect of comparative social science magnifies its value as an arena for
addressing methodological issues. Contrasts between research strategies are
exaggerated and the (often political) implications of methodological decisions
are readily apparent.
Development, for example, is an outcome that has attracted the attention
1 8 C O M P A R A T I V E S O C I A L S C I E N C E
of social scientists for some time. Yet it can be defined in a variety of ways.
To define it in terms of gross national product per capita makes Western Europe,
the United States, and a few oil-rich countries appear to be the most
developed. Defining it in terms of satisfaction of basic human needs, however,
shuffles the development hierarchy and Eastern European countries occupy
more of the prominent positions. Alternatively, development can be
defined politically and qualitatively in terms of the emergence of a national
political culture supported by a stable central government which, in turn, is
acknowledged as legitimate by its subjects. This third definition reshuffles
the hierarchy (Mexico, for example, is among the more advanced countries
according to this definition) and suggests a complete rethinking of issues
surrounding the causes of development.
Thus, methodological decisions that might seem minor in other research
areas have unavoidablypoliticalimplications in comparativework. These implications
are especially salient to researchers who do qualitative work.
LOOKING AHEAD
Before contrasting the two major strategies of comparative research (in
Chapters 3 and 4), I address the issues of heterogeneity and causal complexity,
especially multiple conjuncturalcausation, in Chapter 2. The latter issue
is important for two reasons. First, many comparativists are especially interested
in historical outcomes, and their explanations often cite combinations
of causal conditions. The assessment of causal complexity, therefore, is of
major importance to comparative social science. Second, the two major research
strategies differ dramatically in their approach to causation. In the
case-oriented approach, causal complexity is easier to examine (and to assert)
because usually only a small number of cases are examined. In the
variable-oriented approach, by contrast, causal complexity poses difficult
specification issues. Thus, the examination of causal complexity provides an
important backdrop for contrasting the two major strategies.