European Journal of Political Research 26: 193-219, 1994.

01994 Kluwer Academic Publishers. Printed in the Netherlands.

Comparing political systems: Establishing similarities and

dissimilarities

GISELEDE MEUR’ & DIRK BERG-SCHLOSSER~

’Universitk Libre, Bruxelles, Belgium; ’ Philipps-Universitat, Marburg, Germany

Abstract. Comparative political analysis at the macro-level of political systems can reduce the
inevitably high complexity of such comparisons by the systematic matching or contrasting of
cases, depending on the particular problem. Such ‘most similar systems’ or ‘most different
systems’ designs, in Przeworski & Teune’s terminology, thus constitute one of the major ways
out of the usual ‘small N - many variables’ dilemma. This paper proposes a detailed and
comprehensive method to establish such similarities and dissimilarities in a systematic and at all
stages transparent way. The examples chosen refer to an analysis of the conditions of survival
or breakdown of democratic systems in the inter-war period in Europe.

Introduction

Since the time of Aristotle comparative politics and the comparative method
have been considered by many authors to be the ‘royal road’ of political
science. (For an assessment of the venerable history of this field see Eckstein
1963). Comparative politics was to provide the discipline with a method and
a perspective which would lead to scientifically valid, testable propositions
with a high explanatory power in both space and time. Yet, as one major
analyst has noted, much of what has been written under this heading has
remained ‘essentially noncomparative, essentially descriptive, essentially par-
ochial, essentially static, and essentially monographic’ (Macridis 1955: 7). To
be sure, our substantive body of knowledge has expanded considerably dur-
ing the past decades and now encompasses much important information
relating to practically all countries and regions of the world. (The various
editions of the World Handbook of Political and Social Indicators, for exam-
ple, constitute a major effort in this regard; see Taylor & Jodice 1983.) Still,
in actual performance, the ‘revolution in comparative politics’ which began
in the 1950s has not lived up to its original promise in terms of the collection
of world-wide data and the development of new concepts and approaches.
(For recent assessments of this topic see Mayer 1989; Collier 1993).
On the one hand, configurative studies, dealing with the complex interac-
tion of a wide variety of variables in a single system, have remained mostly
descriptive. Their potential breadth and historical depth have often been
achieved at the price of a lack of systematic argument and scientific rigour.
On the other hand, macro-quantitative studies testing the relationships of a
few variables across a great number of cases have often been too narrow in
194

perspective, too unspecific in the operationalisation of concepts and too

indiscriminate in their selection of cases. In consequence, their results have
often remained spurious or superficial. Such studies can rightly be termed
‘scientistic’ in as much as they arrive at false or irrelevant substantive results
by the use of (presumably impressive) quantitative techniques (see also Ragin
1981).
In order to assess developments in the field of comparative politics, the
contents of the two leading specialized journals, Comparative Politics and
Comparative Political Studies, were analysed from their inception until 1981.
This study showed that, in actual fact, not so much’ had changed since
Macridis remarks: ‘Our inventory. . . reveals that in some respects the new
comparative politics is remarkably like the old comparative government. In
particular, the single country study - the mainstay of the field as it was
defined traditionally - has proved to be extremely durable, and still holds
firmly to its longstanding pre-eminence . . . masquerading under the compara-
tive label’ (Sigelman & Gadbois 1983: 301).
Adam Przeworski, co-author of the most influential methodological study
in this area (Przeworski & Teune 1970), also notes in his review of more
recent developments that ‘by the early nineteen seventies the field of com-
parative methodology had run out of steam’ (Przeworski 1987: 34). In parti-
cular, the ‘quasi-experimental design’ advocated as the central concern of
the comparative method by him and his predecessors, most notably John
Stuart Mill, hardly ever seems to have been put into practice: ‘I do not
know one single study which has successfully applied Mill’s canon of only
difference’ (i.e. ‘most similar systems design’ in the terminology of Przewor-
ski & Teune 1970; ibid, 19). As he further laments, ‘Methodologists are at
times listened to, always acclaimed, but rarely followed. Their canons are
often impossible to observe and their advice often turns out to be impractical’
(Przeworski 1987: 31).
The crux of comparative politics lies in the central dilemma that, on the
one hand, we are always dealing with very complex systems and a large
number of variables and that, on the other hand, the number of cases to be
analysed at the country level, even in global terms, remains relatively small
(see also Blalock 1984). The various emphases which can be undertaken in
this regard are illustrated in Figure 1.
The ‘most similar systems design’ refers to what is listed above as ‘painvise
comparisons’ (C2V,) and the ‘comparative method’ ( C x V y ) in the narrower
sense of the term. Here, as Przeworski and Teune state, ‘the assumption is
that we can find a pair of (or more) countries which differ in all but two
characteristics and that we will be able to confirm a hypothesis that X i s a
cause of Y under conditions under which Ceteris Puribus holds in the real
world’ (Przeworski & Teune: 17). However, they also note that ‘There are
no two countries in the world . . . which differ in only two characteristics and
in practice there are always numerous competing hypotheses’ (ibid.).
Before abandoning all systematic efforts in this direction, we would like
 195

1 2
CASES (C)
'x' lsniall) 'y' llargc)
I
I n 1
I----- T
- - - - - -r I
I I ynlnul.w I cnnip.uatlvc I I
I I comparison I I
I D I I I
I E l 2% I I
I I (Ilnkkan) I
I
I S I I I
A I I I I
R C I
I I I I I
I I I
I It I I
A I I I stallsllcal
B I
L I I method I

single
CLASS1 F l C A l l O N
I world
system C I y (Wallersleinl

Fig. 1. Types of comparisomSource: Adapted from Aarebrot & Bakka (1992: 59).

to propose a method which attempts to operationalise some of Przeworski's

and Teune's original ideas and to indicate at least certain approximations by
which such a research programme can be carried out with a reduced regional
and historical scope. In so doing, we consciously abandon the possibility of
immediately arriving at 'universal' and ahistorical propositions and focus our
efforts instead on a clearly-chosen 'medium range'. Only at a later stage, or
in parallel attempts, might similar comparisons on an intra- and inter-regional
or asynchronic scale prove helpful for more global considerations.
The inter-war period in Europe provides a particularly favourable setting
for such an undertaking since the cases to be considered share many common
socio-economic and political-cultural characteristics. Not only is their history
relatively well researched and documented, but the period concerned is also
clearly demarcated by common events, the two World Wars. Both wars
significantly altered the internal and external political landscapes, setting the
time between them apart from earlier and later developments. All of the
cases under consideration can be defined as having been parliamentary demo-
cracies initially, some well established and of relatively long standing, others
196

M I 9, M
I I
I I
I I

Fig. 2. Simplified system model.

of more recent origin and, in many instances, more democratic in form than
in substance. They were all subsequently affected by a common external
stimulus, the world economic crisis of the late 1920s and early 1930s.
The consequences of this ‘quasi-experiment’ can be examined with the
help of a ‘most similar systems’ - ‘most different systems’ research design.
For this purpose, we shall first sketch a general but ‘historically-informed’
systems model which is capable of accommodating the high level of com-
plexity of each case. On this basis a method will then be elaborated by which
the respective similarities or dissimilarities of systems can be specified more
closely. This procedure will be illustrated by concrete examples taken from
an international comparative research project in which the factors conducive
to the breakdown or survival of inter-war European democracies are ana-
lysed.

The systems framework

Our simplified systems model was developed on the basis of well-known

studies by Deutsch (1963), Easton (1965), Almond & Powell (1978) and
others. However, it is used here solely in a pre-theoretical, classificatory
sense in order to locate different elements and possible interactions more
closely without necessarily implying distinct causal relationships (such as the
effectiveness of certain links and feedbacks or the stability of the system as
such). A preliminary outline of the model is provided in Figure 2.
With the help of this model, it is possible to distinguish and locate the
general social system, the intermediary structures on the input side, the
 197

central political system and the output structures together with the respective
international environment. Furthermore, with regard to each sub-system, we
can distinguish an ‘objective’ dimension consisting of its internal structures,
institutions and similar aspects of a more durable and ‘tangible’ nature,
and a ‘subjective’ dimension reflecting the respective perceptions and actual
behaviour of the individuals and groups concerned. (For a fuller exposition
of a model of this type see Berg-Schlosser/Siegler 1990; Berg-Schlosser/Stam-
men 1992.’)
Using this broad systems outline, we selected seven major categories and
a certain number of characteristic variables within each category on which
to base our comparisons. In so doing, we attempted to be as parsimonious
as possible without losing sight of the overall dimensions and their com-
plexity. The first category, overall geopolitical and historical background,
draws its main substance from Rokkan’s (1975) ‘Political Map of Europe’.
Here, particular consideration was given to the ‘seaward-corebelt-landward’
and ‘reformation-non-reformation’ dimensions, the overall size of the popula-
tion and the timing of both the formation of the state and the establishment
of a democratic political system.
The second category deals with general economic conditions and includes
both the level of development and the basic class structure of the societies
concerned. Among the indicators selected for this category are the national
product per capita; urbanisation; literacy; industrialisation; and data per-
taining to the main social classes. In the latter context, the rural structure
(a significant share of landlords and a rural proletariat vs. a dominance of
family farms), the extent of the middle classes and the size of the industrial
labour force (coinciding with the indicator for the level of industrialisation)
are of particular importance.
The third category is concerned with the particular ethnic, linguistic, reli-
gious and regional composition of each case together with the possible exis-
tence of overarching structures which bridge the gap between such cleavages
(for example, a pattern of verzuiling). The second and third categories thus
cover the major social structural dimensions which figure in the bottom
square (‘social system’) of the overall systems model.
The fourth category summarizes those aspects of political culture which
are most relevant to our concerns. These include: the overall (‘national’)
identity of the society in question; the existence of strong cultural sub-milieus
which characterise the ‘community system’; attributes such as the extent of
secularisation, egalitarianism, tolerance and the acceptance of violence in
the ‘socio-cultural’ sphere. They also include a number of more directly
political orientations such as the level of political interest and information;
political participation; the dominant patterns of conflict resolution (competi-
tive or consensual) and decision-making (authoritarian or participatory); the
extent of ‘parochial’ and ‘subject’ orientations (according to Almond &
Verba’s (1963) definition); and the resulting degree of overall democratic
legitimacy.
198

The fifth category groups together significant features of the intermediary

structures. These include the strength of the major interest groups (rural,
commercial, employers, trade unions); the existence of important social
movements, militias or anti-system parties; the overall fragmentation of the
party system; and the incidence of clientelistic or corporatist forms of interest
mediation.
The sixth category deals with specific features of the central political
system. Among these are the general system type; the vertical separation of
powers (e.g. independence of the judiciary); the horizontal separation of
powers (centralized or federal); the electoral system (proportional or majorit-
arian); the stability of governments; the strength of the bureaucracy and the
repressive apparatus; the social security system; the political role of the
military; and, as an important normative criterion, the guarantee and observ-
ance of civil rights and political liberties.
Finally, the seventh category treats the external environment and includes
such factors as economic or political interactions, cultural influences and
specific historical conditions (e.g. the consequences of World War I, the
possession of colonies) The complete list of variables together with their
respective operationalisations is provided in the appendix.

Establishing similarities and dissimilarities

This kind of operationalisation, of course, can only provide a certain approxi-

mation and is open to further modification. Still, despite the rather rudimen-
tary manner in which categories and variables were selected, it leaves us
with altogether 61 variables for the 18 cases included in our project. Each
case can thus be characterized within its particular configuration, thereby
replacing ‘proper names. . . by the relevant variables’ (Przeworski & Teune
1970: 30). However, the differences remain considerable nonetheless. Across
all variables the minimum variation is 14 for the two ‘most similar’ cases
Estonia and Finland. It is 17 for Belgium - the Netherlands or Germany -
Austria.
The particular research design chosen - ‘most similar’ or ‘most different
systems’ also depends on the dependent variable in a given ‘quasi-experi-
ment’. In our case, we were interested in the effects which the world econ-
omic crisis of the late 1920s and early 1930s had as the major stimulus for
the eventual survival or breakdown of democratic regimes. Accordingly, only
‘most similar systems with different outcomes’ (MSDO) or ‘most different
systems with the same outcome’ (MDSO) designs were set up for closer
analysis.
Furthermore, we distinguish between breakdowns which led to the estab-
lishment of more traditional ‘authoritarian’ regimes and those where inter-
 199

Survlvors Breakdowns

with Samc
Outcome
(MDSO) Breakdowns

Most Similar survivors Breakdowns (5)

with Different
Oulcomc
(MSDO) Fasclst

Fig. 3. Comparative research design.

ventions of strong fascist forces oceurred in the process of breakdown. The

resulting pattern of research designs is illustrated in Figure 3.
Thus, six different types of comparison emerge: the most diferent cases
(1) among all survivors as well as (2) among all breakdowns and, more
specifically, (3) among all fascist and (4) among all authoritarian breakdowns;
and the most similar cases ( 5 ) among all survivors versus all breakdowns and
(6) among fascist versus authoritarian breakdowns. We face the difficulty of
having to measure the closeness or remoteness of any given pair of cases in
a heterogeneous, multi-dimensional space and then of finding the ‘most
different’ and the ‘most similar’ cases. Accordingly, several problems had to
be addressed. These included in particular the choice of a distance with
which to measure proximity and the weighting of the variables.
As a measure of distance we opted for ‘Boolean’ distance. This measures
the number of Boolean (i.e. dichotomised) variables by which two selected
countries differ from one another. In itself, of course, the Booleanization of
variables implies a certain loss of information when compared to more finely
graded measures or scales. However, a number of the variables for our cases
were in a rather crude form anyhow, for example those relating to the levels
‘low’ and ‘high’ (especially when the variables in question were of the ‘softer’
judgemental type) and to the absence or presence of certain factors. Even
where ‘harder’ and more differentiated independent data from standard
sources (levels of GNP, urbanisation, etc.) were available, it turned out upon
closer inspection that many of these data were not really comparable because
of differences in definition or coverage, or even because of gross insufficiencies
on the part of statistical bureaus in the countries concerned. Thus, for the
present purpose, we adopted a fairly straightforward measure of Boolean
distance, seeing that its relative crudeness is offset to a certain extent by
other advantages (see Berg-Schlosser/De Meur 1994 for another application).
The second consideration concerned the weighting of the variables and
the possible effects of intercorrelations of certain factors. Here, we decided
not to consider all variables at once and give them an equal weight because
 SW FI BE NL FR C B CS IR PU ge it hu ro er pr sp gr pi
sw
FI 2
BE 5 5
NL 3 5 2 Zone 1
FR 5 1 1 4 4
CB 3 5 6 4 2
cs 4 2 3 5 5 1 7 1
IR 6 4 3 5 3 5 2
PU 4 2 3 5 5 7 l o 2
ge 6 4 3 3 5 5 4 4 4 Zone 2
it 7 5 2 4 4 6 3 3 I Zone3
hu 4 2 3 5 5 7 2 0 4 3
ro 4 3 5 5 7 2 0 4 3 0
CS b 5 5 7 4 2 4 I s 2 2
Pr 5 4 4 2 4 Y--I 3 5 4 3 3 5
*l, 6 6 5 5 I 3 4 2 4 4 3 4 4 6**1
R‘ 3 1 4 6 6 6 1 3 5 4 1 I I 1 4 ~
PI 5 3 4 6 4 6 1 3 3 2 I II 3 4 3 2
Zone 5 Zone 6 Zone 4

0minimahaxima of zones I , 3,4,S and 6 maximum of zone 2

Fig. 4. Distance matrix for ‘general background’ (8 variables). Countries: SW - Switzerland;

FI - Finland; BE - Belgium; NL -The Netherlands; GB - Great Britain; CS - Czechoslovakia;
IR - Ireland; au - austria; ge - Germany; it - Italy; hu - Hungary; ro - Romania; es - Estonia;
pr - Portugal; sp - Spain; gr - Greece; pl - Poland.

the number of variables can differ widely from one category to the next.
Rather, we proceded first by establishing similarities and dissimilarities cate-
gory by category and then, in a second step, by aggregating these somewhat
further. In this way, the qualitative importance of each (systemically derived)
category was retained. Also, it was now more justifiable to give the variables
an equal weight within each category (any other weighting would have been
just as arbitrary, but at least gross distortions could be avoided). On this
basis we proceeded in three main consecutive steps:
- the composition and synthesis of distance matrices;
- the design and synthesis of similarity and dissimilarity graphs; and
- the selection of the most striking configurations for comparison.

Distance matrices

Within each of the seven categories of Boolean variables (see Table 1 in the
Appendix for the full data), we computed the Boolean distance between
pairs of countries - this distance is defined above - and listed them in a
triangular distance matrix. An example of the first category is given in Figure
4.
In this matrix we have grouped the survivor cases (indicated by capital
letters) and the breakdown cases (indicated by small letters) separately so
that the comparative designs mentioned above emerge. On the one hand,
 201

we consider the most difference systems among all survivors (zone l), among
all breakdowns (zone 2), among all fascist cases (zone 3) and among all
authoritarian cases (zone 4). On the other hand, we look at most similar
systems when survivors and breakdowns (zone 5 ) or fascist and authoritarian
cases (zone 6) are contrasted.
For each of the seven distance matrices, we considered the minimum
distance obtained for different outcomes (MSDO) and the maximum distance
for countries with the same outcome (MDSO). For the first category, the
most different cases among the survivors are FinlandFrance and Czechoslo-
vakia/Great Britain (distance 7). The most different among the breakdowns
are Spain and Estonia (distance 6). The most similar pairs when comparing
survivors and breakdowns are Czechoslovakia/Austria Czechoslovakia/
Hungary, Czechoslovakia/Romania and Finland/Estonia (distance 0). Each
MSDO or MDSO pair is indicated in Figure 4 by its respective distance
which has been highlighted by a box.
The additional distinction between fascist and authoritarian breakdowns is
covered in zones 3, 4 and 6. The most dissimilar pair among fascist cases is
Estonia/Italy (with distance 5 ) and among authoritarian cases Greece/Spain
(with distance 5 ) . Contrasting fascist and authoritarian breakdowns, the
most similar pairs are Austria/Greece, Austria/Poland, HungaryIGreece,
Hungary/Poland, Romania/Greece, Romania/Poland, and EstoniajGreece
(with distance 1).
Somewhat different results were obtained, as was to be expected, across
all seven categories. We thus aggregated the results by category in a further
step. For this purpose, we first juxtaposed the results within each category
in a comprehensive triangle in which each cell was composed as follows:

Bl Here, Ci = the distance (between a pair

of countries) for the ith category.

This resulted in Table 2 in the Appendix.

Then, for each of the six zones of comparison, we marked the resulting
matrix at several levels. We started (at level 0) by selecting for each pair of
countries the categories in which the distance was either equal to the mini-
mum (for countries with a different outcome) or to the maximum (for coun-
tries with the same outcome).
The minimal and maximal values for each category actually obtained in
each zone (as contrasted to the hypothetically possible minimal and maximal
values) are portrayed in Figure 5.
Subsequently, level 1 was obtained by considering the categories in which
the distance was s min + 1 (resp. 2 max - 1). In this way, we continued to
mark all further levels until we reached the threshold level k which separates
similarity from dissimilarity. This borderline was established for each cate-
gory by dividing the total number of variables by 2.
202

Fig. 5. MSDO minima and MDSO maxima.

In order to reduce the heterogeneity of the number of variables per cate-

gory, we decided to retain four levels of (dis)similarity (and no more). The
lower levels are more restrictive than the higher which, for their part, provide
more comprehensive information involving a larger set of categories. If the
threshold level in a particular category is obtained at an earlier stage, this
value is retained for the subsequent levels as shown below:

Levels of dissimilarity:

DO Mean

Here, for example, all levels for category 7 are clearly differentiated. For
category 3, in contrast, the threshold is reached at level I and thus remains
the same for levels 2 and 3 (see numbers marked with an asterisk). Category
6 does not produce any dissimilarities at all since its starting value of 4
already lies at the threshold (mean).
Figure 6 shows markings at four levels (Do to D3) within the MDSO zone
1 which compares survivors. Level 0 is symbolized by Do, level 1 includes
both Do and D1, level 2 Do, D 1 ,D2 and so on. For each pair, the individual
symbols are located in the position of the respective category as indicated in
Figure 6. For example, for the pair Belgium/Sweden, D2 is located in the
position of category 1 and D3 is located in the position of category 7.
 203

- Switzerland; FI - Finland; BE -
Fig. 6. Dissimilarities among survivors. Countries: SW
Belgium; NL - The Netherlands; GB - Great Britain; CS - Czechoslovakia.

Overall similarity and dissimilarity graphs

Next, we counted the number of marked categories for each level. Here, for
example, the pair NL/FI has the highest number (5). The highest value
obtained (h) between a pair of countries was then reproduced in a ‘similarity’
or ‘dissimilarity’ graph. An additional procedure, providing a more complex
set of information takes the value (h - 1) into account as well - those pairs
which, in this example, show levels of difference for four categories, namely
BE/FI, GB/FI and CS/GB. The results of both procedures are illustrated in
Figure 7 with regard to the most different systems among survivors. Continu-
ous lines indicate the maximal value (h), dotted lines indicate the sub-
maximal value (h - 1). The specific categories concerned are indicated in
parentheses for each pair.
Finally, we synthesised all this information into aggregated graphs like the
one for survivors shown in Figure 8. The four levels of (dis)similarity are
superimposed; in the cases where the value h alone determines the pairs,
the procedure stops here. Alternatively, when h - 1 is also taken into
account, we decided to keep each (h) pair and those (h - 1) pairs which
occur at least twice.
In our example, the final graph for survivors contains such pairs as SW-
204

*
0 0
1 sw /sw

CY
NL
0 0
BE BE
cs-Sw:wII(S)
CS. CB.Do=Z(G.Fl
FI-FR:wII[C.)
FI - C BDoc 2IE.I)
CS S W Dl= 2 (S.1)
-
~

CS CB:D1= 3 (G.S.Fl
-
FI C B Dl= 3 (E.1)
CS I R Do= I IS] -
PI I R Do= I (P)
-
~

PI NL: DO= 1 (S)

IcrcL1;_ lmuL
0 0
SW sw

&
00 0
FR
cs

BE VF1 -
CS - PR: D2- S (C.S.p) FI GO: D2- 3 (G.E.1)
-
FI I R D2- 3 (P.1.F)
Fl - N L D2= 4 1C.E.S.I)
-
FI BE: D2= 4 1G.E.P.PI
2S
NL /
of
BE
- C B D3= 4 (G.S.1.F) FI - G B DS= 4 1C.E.I.R
FI - M; D3- 6 1G.E.S.I.P)
Fl - B E D3=4 1G.E.P.F)

Fig. 7. Dissimilarity graph (at all levels) for survivors. Abbreviutiom for the categories: G -
Category 1 (General backcround); E - Category 2 (Socio-economic conditions); S - Category
3 (Social composition); P -Category 4 (Political-culturaltraditions) I - Category 5 (Intermediary
structures); C - Category 6 (Central poltical system); F - Category 7 (External factors).

BE (corresponding to the maximal value h at level 1) and SW-IR (corre-

sponding to the sub-maximal value h - 1 at levels 0 and level 2); however,
it does not contain SW-FI which occurs only once (at level 2) with a weak
(h - 1) relation. When underlined, Di indicates the strong (h) relation.
In summary, the process described above is designed to reveal similarities
(or proximities) and dissimilarities (or remoteness) between countries in a
number of complementary ways. On the one hand, the construction of
different levels (of which we retained four) establishes proximities of different
kinds: the lower the level, the higher the demands placed on the criterion
of likeness (or of dissemblance) within each category. Accordingly, few cate-
 205

cs C S --SSWW
DO=l (Sl
D l = 2 (S.1)

CS - GB:
DO12 (0.F)
D l r 3 (0,S.F)
D2 D 2 4 (G.S.I.p1
D3=4 (G.S.1.P)

sw

FI - GB: F1- I R FI - NL: FI - B E

D0=2 (E.1) DO=] (PI D k l IS) D2=4 (G.E.P.FI
D1=2 (E.1,) D2=3 (P,l,F) D2=4 (G.E.S.1) D3=4 (G.E.P.F)
D2=3fG.E.1) D3=51C.E.S.I.Fl

Fig. 8. Most different pairs among survivors.

gories will fit this requirement. To be ‘alike at level zero’ thus means to be
as close as possible within the (few) categories where this high proximity
occurs.
Moving to higher levels implies a progressive relaxation of the likeness
(or dissemblance) requirement within each category and the simultaneous
demand that it should appear in a growing number of categories. Here,
proximity is weaker and less specific but concerns a greater number of
dimensions; it is locally less,but globally more demanding.
On the other hand, considering the maximal value (h) of the number of
categories which satisfy the proximity criterion at a given level corresponds
to focusing upon the closest pairs for that criterion. The additional consider-
ation of the next closest value (h - 1) means that pairs just a bit less proxi-
mate will also be taken into account. Such a weakening of the proximity
requirement provides for a wider range of information.

Selection of most important configurations

The final graphs obtained for the six possibilities (zones) in our overall
research design (see Figure 1) resulted in the following six patterns which
are illustrated in Figures 9a to 9f.
For the purpose of clarity, we isolated the configurations which were
 Fig. 9a. MDSO among survivors.

chosen to be studied as a whole. When ‘poor’ configurations (consisting of

an isolated pair) emerged, we decided to enrich the possible comparisons by
adding a case with weaker links (symbolised by dotted lines). Greece, for
example, was added to the second configuration in Figure 9b and placed in
relationship to Spain. In general, we considered multiple comparisons to be
more interesting than pairwise comparisons: they determine a more restricted
subset of variables and thus allow us to focus on those which might contribute
more decisively to the outcome.
At first, we considered the most dissimilar systems designs.
(a) Survivors (as demonstrated above). The greatest differences can be
found between Great Britain and Finland on the one hand and Great Britain
and Czechoslovakia on the other. If we add the weaker links, the configur-
ations highlighted in Figure 9a (a triple comparison of Great Britain and
Sweden versus Czechoslovakia on the one hand and a fivefold comparison
of Great Britain, Belgium, the Netherlands and Ireland versus Finland on
the other) emerge.
(b) Breakdowns. Here, the greatest differences are between Germany and
the triad Portugal, Romania and Greece on the one hand and between Spain
and Estonia on the other. To the latter constellation one can add the case
of Greece as contrasted with Spain.
(c) Fascist breakdowns. For the fascist breakdowns two major constel-
lations emerged: Germany versus Romania on the one hand and Estonia
versus Italy on the other. Weak links between Germany and the pair of Italy
and Estonia can be added.
(d) Authoritarian breakdowns. Here, Greece and Spain are the most dif-
 207

ge es gr

SQ -es: sp - gr:
A"oly(& g&!;,F)
D1=4 (G.E.1.F) D1=4 (G.E.1.F)
f%=r&.F)
D2=5 (l3.S.I.C.F)
DO=3 (G.P.F) D1=3 (G.1.F)
D1=3 (C.P.F)
D2=5 (G.E.1.C.P)

Fig. 96. MDSO for all breakdowns.

d.. . .. .. ..
. . . . . . . . . D?. . . . . . . .

Fig. 9c. MDSO among fascist breakdowns.

ferent cases. Some additional information can be gained by looking at Portu-

gal (weakly linked to Spain) within the context of a triple comparison.
Finally, the most similar systems designs for different outcomes can be
listed as follows.
(e) Survivors versus breakdowns. Here, three constellations are especially
noteworthy. Finland and Estonia are the pair with by far the strongest
similarities but with a different outcome; to this we can add Sweden and
Ireland on the one hand and Hungary and Germany on the other. Another
208

FI

FR-S FR-gr:
SI=3&,s.o SO=2 (S.P)
S2=5 IG.S.I.P.FI

IR

SF-ge: F1-es: sw - cs:

-1 (I) SO=] (G) SO=2 (SJ)
S1=3 (P.1.F) S1=4 (G.E.S.C) S2=5 (G.S.P.C.v
S2=6 (G.E.S.I.C,F) S3-6 (G.E.SJ'.C.F)
53=6 (C.E.S.1.C.F)

-
SW hu: IR - CS: -
IR hu:
SO=1 (S) sO=2(ES) SO=1 (S)
s3=5(E.S.I.C.Fl S3=6(G.E.S.1.C.F)

Fig. 9d. MSDO among authoritarian breakdowns.

Fig. 9e. MDSO contrasting survivors and breakdowns.

very similar pair is Czechoslovakia and Austria which, together with Hun-
gary, make an interesting triple comparison. France and Spain also have
much in common but a different outcome; here, a further comparison with
Greece and Great Britain can provide additional information.
(f) Fascist versus authoritarian breakdowns. Three interesting constel-
lations can be recognised here: Germany and Italy versus Spain and Portugal;
 Fig. 9f: MSDO compaiing fascist versus authoritarian breakdowns.

Greece versus Austria, Romania and Estonia; and, finally, Hungary versus
Poland and Portugal.
All major patterns have thus been identified and can now be discussed in
greater detail. As a shortcut to identifying additional relationships, similarit-
ies and dissimilarities can also be approximated by a more 'down and dirty'
method. This consists of calculating the sum of the number of identical
variables divided by the maximum number of variables obtained in each
category; the total is then divided by 7 - the total number of categories. In
this way, the principle of giving each variable the same weight within each
category and similarly of giving each category an equal weight is maintained.
The results of this procedure are depicted in Figure 10. As cutting points we
can choose, for example >0.5 for the most different and <0.35 for the most
similar pairs. The latter threshold is somewhat more demanding because,
given the regional and historical focus of our analysis, the overall similarities
in our universe of cases are relatively greater than the differences. It must
be born in mind, however, that this procedure is much less accurate and
reliable than the method demonstrated above.

Example

Among the various constellations considered, Finland and Estonia (MSDO

contrasting survivors and breakdowns) constitute a particularly striking exam-
ple. They are by far the most similar cases, having evidenced the highest
number of similar categories across all four levels (SO. . . S3). When the
actual differences are listed across all categories (see Table 3 in the Appen-
210

SW FI BE NL FR GB CS IR mu go it hu m cs pr sp gr

I
1.25
1.35 0.33
1.28 0.49 0.37
1.44 0.59 0.38 0.22 Zonc3
1.33 0.41 0.48 0.34 0.37
1.46 0.58 0.34 0.33 0.32 0.51
3.37 0.34 0.25 0.46 0.43 0.54 0.36
1.42 0.53 0.40 0.37 0.31 0.27 0.31 0.46
3.38 0.38 0.40 0.38 0.34 0.38 0.44 0.38 0.38
Zone 6 Zone 4

Fig. 10. ‘Down and dirty’ distance matrix. (For abbreviations of the countries, see Figure 4.)

dix), the remaining variables are highlighted and these indicate where the
reasons for the difference in outcome may lie. In the case of Finland and
Estonia, 14 variables out of the original 61 were left over as ‘different’. Such
a table can direct the researcher’s attention to those areas in which the most
significant differences - those which may explain the different oucome in
spite of all similarities - are located. In our example, the greatest discrepanc-
ies between the two cases clearly occur within the area of political culture,
indicating that one should investigate this aspect and its different components
(including their historical roots) more closely.
It is also possible to add a further very similar case - Sweden recommends
itself in our example - and to inspect the similarities and dissimilarities within
the resulting triangular constellation (see Table 3b in the Appendix). The
overall differences between the two survivor cases on the one hand and the
breakdown case on the other are thereby reduced to 8 variables. With
further additions of very similar cases it may be possible to eliminate more
idiosyncrasies and thus to reduce the range of relevant variables even further.
The categories and variables discussed thus far describe the general histori-
cal, social structural, political cultural, and institutional background of the
cases under investigation. Some of the important ‘structural’ reasons for the
observed outcome (which can be identified by our method) can be considered
to lie among these variables. Now, it is possible to introduce a further
element into our quasi-experiment: the specific outside stimulus (in our
example, the effects of the world economic crisis - see category 8 in Table
3). It can be shown that the effects of both the immediate post-war crisis of
1918/19 and the general economic crisis after 1929 were particularly severe
in both of our ‘most similar’ systems. While the first crisis was overcome by
 211

the existing parliamentary regimes in both Finland and Estonia, the second
crisis led to the collapse of parliamentary democracy in the latter case.
At this point the specific actions and reactions of the major political
groups and actors become crucially important (see category 9). In the case of
Finland, the threat posed by the Lapua movement was countered by the
energetic intervention of President Svinhufvud on behalf of the democratic
system which led to the formation of a broad-based ‘red-green’ (socialist-
agrarian) coalition (see Karvonen 1988). In Estonia, it was also the elected
President who intervened, but this time in what he claimed to be a pre-
emptive coup against the similarly strong ‘veterans’ movement’ which led to
the establishment of an authoritarian regime (see Varrack, forthcoming).
With the help of the method outlined here, both structure- and actor-
related approaches can be brought to bear on any particular constellation of
factors and their respective outcome. In Jon Elster’s (1984: 13ff.) somewhat
different terminology, both the specific ‘opportunity set’ and the particular
choices of relevant actors, including their interdependencies and interactions,
can be highlighted and more specifically analysed. It is thus actually possible
to conduct a ‘quasi-experiment’ in which most factors are controlled and
crucial differences arising in a critical situation are emphasized.

Conclusions

The preceding presentation has attempted to demonstrate an approximate

but, in our opinion, feasible way out of the central dilemma of comparative
social research which Przeworski & Teune have formulated as follows: ‘Since
the number of the relevant determinants of any kind of social behavior is
likely to exceed the number of accessible social systems, the objective of a
theory free of all proper names will not be easily reached, and thus proce-
dures must be formulated to maximize this objective’ (1oc.cit.: 31). This was
done with the help of an explicit systems model which, for the example
chosen, was given a more concrete ‘filling’ in the form of conceptual and
historical considerations concerning the cases and the period under investi-
gation.
It was then shown how the respective similarities and dissimilarities for
different possible research designs can be established in a systematic manner
while taking into account the difficulties of reducing such complexity within
a multi-dimensional space. The example given, the pairwise and triple com-
parisons of the Estonian, Finnish and Swedish cases, illustrated the reductive
power inherent in the process of ‘matching’ these cases more closely in the
manner described. Against this background, the effects of the crisis and the
reactions of the major actors could be demonstrated.
It must be noted, however, that such an approach should not be employed
mechanically in order to ‘distill’ out any single causal factor. The cases and
the problems investigated remain much too complex for an analysis of this
212

kind. Rather, the complexity reduction and the similarities and dissimilarities
arrived at should be utilised to guide the researcher and focus hidher atten-
tion on certain key categories and variables with the help of which a more
qualitative and theoretically founded explanation might be attempted. For
such purposes, even the approximations and procedures outlined above
which make use of Boolean variables may, although in some ways still
relatively crude, prove to be sufficient. With the help of this method, then,
quantitative and qualitative procedures can be meaningfully combined. Suf-
ficiently operationalised comparisons of this type may not lead to a ‘royal
road’, but they do provide indispensable elements for all kinds of empirically-
based theory. As Arthur Stinchcombe put it: ‘By the simple act of asserting
that two instances are alike . . . a class, a concept, is created, a generalization
about it is offered, some evidence is brought forth, and we are embarked
on a scientific enterprise’ (1978: 123). Thus, in the manner described, it may
be possible to generate approximations for a ‘medium range’ of proposition-
testing and theory-building which are more conceptually guided and histori-
cally informed.
Such an approach, with appropriate modifications, can be applied to other
regions, periods and theoretical problems as well. As Theda Skocpol notes:
‘The practice of analytic historical sociology forces a more intimate dialogue
with historical evidence than either interpretative historical sociology or the
application of a model to a historical case’ (Skocpol 1984: 385; emphasis in
the original). Or, to use Arthur Stinchcornbe’s metaphor, ‘the theory is built
as a carpenter builds, adjusting the measurement as he goes along, rather
than as an architect builds, drawing first, building later’ (Stinchcombe 1978:
122). The results of systematic operationalised comparisons are indispensable
building blocks for any such purpose.

Appendix: Definition of variables (Boolean version)

(thresholds indicated in parentheses)
0: no, low, weak, below threshold, etc.
1: yes, high, strong, above threshold, etc.
1. General background
POPULATION population (20 million)
SEAWARD seaward periphery
COREBELT core belt
LANDWARD landward periphery
NONREF non-reformed or counter-reformation
REFORM reformation
EARLYSTATE early state-building (before 1800)
PREWARDEM consolidated pre-WWI democracy

2. Socio-economic conditions
NATPRODCAP national productlcap. (<ZOO US$)
 213

URBANIZATI urbanization (50% ; population in towns with more than

20,000 inhabitants)
LITERACY literacy (75%)
LANDLORD significant share of landownership by landlords (100 ha)
FAMFARMS family farmers (50% of agrarian population) (Vanhanen
1984)
AGRPROL agrarian proletariat (20% of agrarian population)
INDLAB industrial labour force (30% of labour force)
MIDDLE old and new middle classes

3 . Social composition
ETHNLINGCL ethno-linguistic cleavage@) (weaklstrong)
RELIGCL religious cleavage(s)
REGIONALCL regional cleavage(s)
OVERVERZUI overarching structures (‘verzuiling’)

4. Political-cultural traditions
NATIDENTIT ‘national identity’
SUBMILIEUS sub-milieus (class, religion, regional, or ethnic; at least one
of these milieus ‘strong’)
VIOLACC acceptance of violence
EGALITAR egalitarianism
POLINFORM political information
POLITIPART political particpation
STATISM statism
PAROCHIAL parochialism
DEMLEGITIM democratic legitimacy
CONSENSlCONFL dominant pattern of conflict resolution (0 conflict/ 1 consen-
sual)
TOLERANCE social and political tolerance
AUTHlPART authoritarianlparticipatorystyle of decision making
SECULAR secularization
SUBJECT subject orientation

5 . Intermediate structures
INTRURAL rural interest groups (weaklstrong)
INTCOMMERC small commercial interest groups
INTUNIONS trade unions (weaklstrong)
INTEMPLOYE employers’ organizations (weaklstrong)
CLIENTELISM clientelism
MOVEMENTS 0 social movements of more recent orign (stronglweak)
MILITIAS armed militias (weakktrong)
PARTFRAG fragmentation of party system (Rae’s F 0.8)
ANTISYSP share of votes of right and left antisystem parties (15%)
CORPORATISM corporatism (weakktrong)

6 . Central political system

POLI’ITYPE political system (constitutional monarchy/republic)
ELECTSYSPR electoral system (majoritarianlproportional)
214

STABGOVERN stability of governments

ROLEBUREUA political role of bureaucracy (weak/strong)
MILITARY 0 1 political role of military (weak/strong)
SOCIALSEC social security system (weaklstrong)
CIVRIGHT Freedom House Index of civil rights (0 if score 3 and above)
POLRIGHT Freedom House Index of political rights (0 if score 2 and
above)

7 . External factors
WWlWINNER winner of WWI
ECONDEPEND economic dependence
CULTANGLO cultural links: Anglo-Saxon (weaklstrong)
CULTGERM cultural links: Germanic
CULTROMAN cultural links: Romanic
CULTSLAVIC cultural links: Slavic
IDEOLCATH ideological links: Catholicism
IDEOLMARX ideological links: Marxism
COLONIES colonies

8. Crisis
POSTWARCRI impact of post-war crisis (weakktrong)
WORLDECON impact of world economic crisis
INTERNREACT internal reactions (strikes, demonstrations, violence)
ELECTANTI significant strengthening of right and left wing antisystem
parties.

9. Major interventions and moves

KEYMDEMCOA broader democratic coalition
KEYMECOREF economic reforms
KEYMCHURCH pro-democratic intervention of church
KEYMMILIT anti-democratic intervention of military
KEYMAUTHOR anti-democratic intervention of authoritarian (upper class
based) forces
KEYMFASCIS fascist intervention
USEOFEMERG use of emergency powers
EXTERNALIN external influences (weaklstrong)

10. Outcome
OUTCOME breakdownlsurvival of democracy
 215

Table 1. Boolean data

sw n ns NL IR GI ca DI AU cs IT HU no ELI rn a CR PL
1. General Backgnmnd (C)
popumno O O O O 1 1 O O O 1 l O O O O l O l
SEAWARD 0 0 0 0 1 1 0 1 0 0 0 0 0 0 1 1 0 0
COREBELT 0 0 1 1 0 0 0 0 0 1 1 0 0 0 0 0 0 0
LANDWARD 1 1 0 0 0 0 1 0 1 0 0 1 1 1 0 0 1 1
NONREF 0 0 1 1 1 0 1 1 1 1 1 1 1 0 1 1 0 1
REFORM l l O l O l O 0 O l O O O l O O O O
EARLYSTATE 1 0 0 1 1 1 0 0 0 0 0 0 0 0 1 1 0 0
PREWARDEM 1 0 1 1 1 1 0 0 0 0 0 0 0 0 0 0 0 0
.2 Socio-EconomicConditions@)
NATPRODCAP 1 0 1 1 1 1 1 0 1 1 0 0 0 0 0 0 0 0
URBANUATI 0 0 1 1 1 1 1 0 0 1 0 0 0 0 0 0 0 0
UTERACY 1 1 1 1 1 1 1 1 1 1 0 1 0 1 0 0 0 1
LANDUlRD 0 0 0 0 0 1 0 0 1 1 1 1 1 0 1 1 0 0
FAMFARMS 1 1 1 1 1 0 1 1 1 1 0 1 1 1 0 1 1 1
AORPROL 1 1 0 0 0 0 0 0 1 1 1 1 1 0 1 1 0 1
INDLAB 0 0 1 1 1 1 1 0 1 1 0 0 0 0 0 0 0 0
MIDDLE 1 0 1 1 1 1 1 0 0 1 0 0 0 0 0 1 0 0
3. Social Composition (S)
El3MLMOCL 0 1 1 0 0 0 1 0 0 0 0 0 1 0 0 1 0 1
REUOCL o o o l o o l o o l o o o o o o o l
RMIONALCL 0 0 1 1 1 1 1 0 1 1 1 0 0 0 1 1 1 1
OVERVERZUI 0 0 I I 0 0 - 1 0 0 0 0 0 0 0 0 0 0 0
4. Political-Culturrl Traditions (P)
NATIDENTIT 1 1 0 1 1 1 1 1 0 1 1 1 1 1 1 0 1 1
SUUMIUEUS 0 1 1 1 0 1 1 1 1 1 1 0 0 0 1 1 1 1
VIOLACC 0 l 0 0 0 0 0 0 0 l 1 0 1 0 0 1 0 0
EOAUTAR 1 1 1 1 0 0 0 1 0 0 0 0 0 0 0 0 0 0
POLINFORM o o l o l o l l l l o o o l o l l l
POLmPART 0 0 1 1 0 1 1 1 1 1 1 0 0 1 0 1 0 0
STATISM 1 1 0 1 1 0 1 0 1 1 1 1 1 1 1 0 0 0
PAROCHIAL 0 0 0 0 0 0 0 1 0 0 1 1 1 0 1 1 0 1
DEMLUlmM 1 0 1 1 1 1 1 1 0 0 0 0 0 0 0 0 1 0
CONSENWXNFL I 0 I I 0 0 0 0 0 0 0 0 0 I 0 0 0 I
TULERANCE 1 0 1 1 1 1 1 1 0 0 0 0 0 0 0 0 0 0
AvrwPART I I 1 1 1 1 1 0 0 0 0 0 0 0 0 0 0 0
SECULAR 1 1 1 1 1 1 1 0 0 1 0 0 0 1 0 0 1 0
SUBJECT 0 0 0 0 0 0 0 1 1 1 0 1 1 0 1 1 0 1
5. IntcnncdiatcStructures 0
MTRURAL 0 1 1 0 1 0 1 1 1 1 1 1 0 0 0 1 0 1
INTCOMMERC 0 0 0 0 1 1 1 1 0 0 0 0 0 0 0 0 0 0
IN"I0NS 1 0 1 1 1 1 0 1 1 1 1 0 0 0 0 1 0 0
INlFMPLOYE 1 1 1 1 1 0 1 1 1 1 1 1 0 0 1 1 0 0
CLIENTELISM 0 0 0 0 0 0 1 1 0 0 1 1 1 0 1 0 1 1
MOVEMENTS0 I I l o o o o l l l l l l o o l o I
MlUTlAS 0 1 1 0 1 0 0 0 1 1 1 0 0 1 0 1 0 1
PARTFRAO 0 1 1 1 1 0 1 0 0 1 1 0 0 1 0 1 0 1
ANnSYSP 0 1 0 0 0 0 1 0 1 1 1 0 1 1 0 1 1 1
CORPORATISM 0 0 0 1 0 0 0 0 1 0 0 0 0 0 1 0 0 0
6. Ccntral Political Syrtcm Q
POurplPE 0 1 0 0 1 0 1 1 1 1 0 0 0 1 1 1 1 1
ELeCrSYSPR I I l l 0 0 0 l l l l 0 0 l 0 0 0 l
STABGOVERN 0 0 0 0 0 1 1 1 1 0 0 1 0 0 0 0 0 0
ROLEBUREUA 1 1 0 1 0 1 1 1 1 1 1 1 1 1 0 1 1 1
MILITARY01 0 0 0 0 0 0 0 0 0 1 1 0 0 0 1 1 1 I
SOCIALSU: 0 0 0 0 0 1 1 0 1 1 0 0 1 1 0 0 0 1
CIWUOIIT 1 1 1 1 1 1 1 1 1 1 1 1 0 1 0 1 0 0
poLRI0HT 1 1 1 1 1 1 1 1 1 1 1 0 0 1 0 1 1 0
7. External Factors (p)
WWlWlNNER 0 0 1 0 1 1 0 1 0 0 1 0 1 0 0 0 1 0
ECONDEPEND 1 0 1 1 0 0 1 1 1 0 1 1 1 1 1 0 1 1
CULTANGLO 0 0 0 1 0 1 0 1 0 0 0 0 0 0 0 0 0 0
CULTOERM 1 1 0 1 0 0 1 0 1 1 0 1 0 1 0 0 0 1
CULTROMAN 0 0 1 0 1 0 0 0 0 0 1 0 1 0 1 1 0 1
CULTSLAVIC 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 1
IDMLCATH 0 0 1 1 1 0 1 1 1 1 1 1 0 0 0 1 0 1
IDEOLMARX 0 1 0 0 1 0 1 0 1 1 l 0 0 0 0 l 0 0
COLONIES 0 0 1 1 1 1 0 0 0 0 1 0 0 0 1 1 0 0

Countries: SW - Switzerland; FI - Finland; BE - Belgium; NL - The Netherlands; FR - France;

GB - Great Britain; CS - Czechoslovakia; IR - Ireland; AU - Austria; GE - Germany; IT -
Italy; HU - Hungary; RO - Romania; ES - Estonia; PR - Portugal; SP - Spain; GR - Greece;
PL - Poland.
216

Table 2. Synthetic distance matrix for all categories

BE

m
ki,
2
1

5
2
5

3 5 2 8

3 5
3 2 1 5 2 3
5 2 0

Counrries: SW - Switzerland;FI - Finland; BE - Belgium; NL -The Netherlands; FR -France;

GB - Great Britain; CS - Czechoslovakia; IR - Ireland; AU - Austria; ge - Germany; it -
Italy; hu - Hungary; ro - Romania; es - Estonia; pr - Portugal; sp - Spain; gr - Greece.
217
218

References

Aarebrot, F. & Bakka, P.H. (1992). Die vergleichende Methode in der Politikwissenschaft, pp.
51-69 in D. Berg-Schlosser & F. Miiller-Rommel (eds.), Vergleichende Politikwissenschafi.
Opladen: Leske & Budrich.
Almond, G.A. & Verba, S. (1963). The civic culture: Political attitudes and democracy in five
Western nations. Princeton, NJ: Princeton University Press.
Almond, G.A. & Powell, G.B. (1978). Comparativepolitics:system, process, andpolib. Boston:
Little, Brown.
Berg-Schlosser, D. & Siegler, R. (1989). Political stability and development. Boulder, CO:
Lynne Rienner.
Berg-Schlosser, D. & Stamrnen, Th. (1992). Einfiihrung in die Politikwissenschafi, 5th ed.
Miinchen: Beck.
Berg-Schlosser, D. & De Meur,G. (1994). Conditions of democracy in inter-war Europe: A
Boolean test of major hypotheses, Comparative Politics.
Blalock, H.M. (1984). Basic dilemmas in the social sciences. Beverly Hills: Sage.
Collier, D. (1993). The comparative method, pp. 105-119 in A.W.Finifter (ed.), Political
science: The state of the discipline 11. Washington, DC: American Political Science Associ-
ation.
Deutsch, K.W. (1963). The nerves of government. New York: The Free Press.
Easton, D. (1965). A systems analysis of political life. Chicago: University of Chicago Press.
Eckstein, H. (1963). A perspective on comparative politics, past and present, pp. 3-32, in H.
Eckstein & D.E. Apter (eds.), Comparative Politics. New York: The Free Press.
Elster, J. (1984). Nuts and bolts for the social sciences. Cambridge: Cambridge University Press.
Flora, P. et al., eds. (1983/1987). State, economy and society in Western Europe 1815-1975. A
data handbook. 2 vols. London: Macmillan.
Karvonen, L. (1988). From white to blue-and-black: Finnish fascism in the inter-war era.
He1sinki:Societas Scientarum Fennica.
Larsen, S.U., Hagtvet, B. & Myklebust, J.P., eds. (1980). Who were the fascists? Bergen:
Universitets Forlaget.
Lijphart, A. (1971). Comparative politics and the comparative method, American Political
Science Review 65: 682-693.
Lijphart, A. (1975). The comparable cases strategy in comparative research, ComparativePoliti-
cal Studies 157-175.
Lijphart, A. (1977). Democracy in plural societies. New Haven: Yale University Press.
Linz, J.J. & Stepan, A. (1978). The breakdown of democratic regimes. Baltimore: Johns Hopkins
University Press.
Linz, J.J. (1975). Totalitarian and authoritarian regimes, pp. 175-411, in F.J. Greenstein &
N.W.Polsby (eds.), Handbook of Political Science, Vol. 3: Macropolitical Theory. Reading,
MA: Addison-Wesley Publishing Company.
Lipset, S.M. & Rokkan, S. (1967). Introduction, in S.M. Lipset & S.Rokkan (eds.), Party
Systems and Voter Alignments. New York: The Free Press.
Macridis, R.C. (1955). The study of comparative government. New York: Random House.
Mayer, L.C. (1989). Redefining comparative politics. Newbury Park: Sage Publications.
Mills, J.S. (1843). A system of logic. Quoted from the eighth edition 1904. New York.
Mitchell, B.R. (1981). European historical statistics 175&1975. London: Macmillan.
Nunnally, J.C. (1978). Psychometric theory. New York: McGraw-Hill.
Przeworski, A. & Teune, H. (1970). The logic of comparutive social inquiry. New York: Wiley.
Przeworski, A. (1987). Methods of cross-national research, 1970-1983, pp. 31-49, in M. Di-
erkes, H.N. Weiler & A.B. Antal, Comparative policy research: Learning’from experience.
Aldershot: Gower Publishing Company.
Ragin, Ch.C. (1981). Comparative sociology and the comparative method, International Journal
of Comparative Sociology 22: 102-120.
 219
Ragin, Ch.C. (1987). The comparative method. Berkeley: University of California Press.
Rokkan, S. (1975). Dimensions of state formation and nation-building, pp. 562-600, in Ch.
Tilly (ed.), The formation of national states in Western Europe. Princeton, NJ: Princeton
University Press.
Sigelman, L. & Gadbois, G.H. (1983). Contemporary comparative politics: An inventory and
assessment, Comparative Political Studies 16: 275-305.
Skocpol, Th., ed. (1984). Vision and method in historical sociology. Cambridge: Cambridge
University Press.
Stinchcombe, A.L. (1978). Theoretical methods in social history. New York: Academic Press.
Taylor, Ch.L. & Jodice, D . (1983). World handbook of political and social indicators, 3rd ed.
New Haven: Yale University Press.
Tilly, Ch. (1984). Big structures, large processes, huge comparisons. New York: Russel Sage
Foundation.
Varrak, T. (forthcoming). The Collapse of Democracy in Inter-War Estonia, in D. Berg-
Schlosser & J. Mitchell (ed.), Crisis, compromise, collapse - Conditions of authoritarianism,
fascism and democracy in inter-war Europe, Vol. 1 . Oxford: Oxford University Press.
Wiarda, H.H., ed. (1985). New directions in comparative politics. Boulder, CO: Westview Press.
Winer, B.J. (1971). Statistical principles in experiment design. New York: McGraw-Hill.

Address for correspondence: Dirk BergSchlosser, Institut fur Politikwissenschaft, Philipps-

UniversitLt, D-35032 Marburg, Germany
Phone: 6421-284397 ; Fax: 6421-288913