Anda di halaman 1dari 14

Int. J. Social Research Methodology Vol. 10, No. 1, February 2007, pp.

4961

Selecting Cases in Cross-National Political Research


Dag Anckar
Received 15 November 2005; Accepted 23 November 2005
TSRM_A_221035.sgm Taylor and Francis Ltd

Four specific problems are discussed that relate to selecting comparable cases in crossnational political research. First, to manage the so-called Galtons problem, which is about the assumption of autonomous units, two strategies are discussed. The one is to abandon the assumption and study the mechanisms of diffusion, the second is to choose cases for comparison which can be assumed to have influenced each other to a minor extent only. Second, it is suggested that the method of paired comparisons be applied more than before to minimize the many variables, small N problem (Arend Lijphart). Third, the tendency of the comparable cases strategy to over-determine the dependent phenomenon throws an unfavourable light on binary comparisons, which are much too popular in comparative political science. Finally, the task of explaining by means of variances in several independent variables the variance in one dependent variable requires the use of several and not only one set of cases.
dag.anckar@abo.fi DagAnckar 0 100000February 10 Taylor 2007 & Francis Original Article 2007 1364-5579 (print)/1464-5300 International Journal of Social Research 10.1080/13645570701211159(online) Methodology

Introduction More than three decades ago, a wise recommendation was issued in a textbook for the benefit of comparative politics researchers: The key to the use of specification for the control of background factors rests in the proper selection of research sites (Holt & Turner, 1970: 12). Taking this recommendation as a point of departure, this article discusses four specific problem areas in the selection of cases, i.e. countries, in crossnational political research. First, implications are discussed of the so-called Galtons problem, which is about the autonomy of cases selected for comparison. Second, the
Dag Anckar is Professor of Political Science at the bo Akademi University, Finland. He is a former President of the Nordic and the Finnish Political Science Associations. His two most recent books, both in Swedish, are Paths Around and Through Political Parties (with Lauri Karvonen and Guy-Erik Isaksson, Stockholm: SNS Frlag, 2001) and Politics in Lilliput States (bo: bo Akademi University Press, 2004). Correspondence to: Dag Anckar, Professor of Political Science, bo Akademi University, Department of Political Science, Biskopsgatan 15, 20500 bo 50, Finland. Tel: (+358)-2-2154316 (office); Fax: (+358)-2-2154585; Email: dag.anckar@abo.fi ISSN 13645579 (print)/ISSN 14645300 (online) 2007 Taylor & Francis DOI: 10.1080/13645570701211159

50

D. Anckar

article argues that the small N-problem in cross-national research can be made manageable by introducing a comparison of pairs of pairs. Third, the article calls attention to the fact that much comparison takes place through binary cases, and suggests that weaknesses inherent in this approach may be overcome by means of triangulation, i.e. multiple measurement. Finally, the article makes a case for the use of many sets of units rather than one set in research that aims at exploring relations between several variables. By way of introduction, two limitations of the presentation must be briefly stated. First, for obvious reasons, the article does not dwell upon large-scale statistical studies that encompass most (e.g. Vanhanen, 2003) or, indeed, all countries in the world (e.g. C. Anckar, 2004). In such studies problems of case selection do not appear. Second, while dealing with case selection and selection biases in focused comparative research (Collier & Mahoney, 1996: 5659), the article does not take a stand in the potential controversy between traditional and more recent qualitative approaches to the handling of small N-research, the more recent categories including applications of the mathematical techniques of Boolean algebra (Ragin, 1987) and more lately applications also of fuzzy algebra (Smithson, 1987, Ragin, 2000). The controversy does not extend to the logic of case selection which remains essentially the same in both types of approaches. In like manner, the article is not primarily concerned with the controversy within traditional approaches between so-called most similar and most different research designs. The first design implies seeking for differences in the midst of similarities; the merits of this design, which is more usual in comparative political research, are analyzed at some length by Arend Lijphart in a by now classic article (1975). The second design implies seeking for similarities in the midst of differences; this design was identified and advocated by Adam Przeworski and Henry Teune in one early and most important methodological study (1970). Giovanni Sartori has in one of his many writings on the comparative method emphasized that seeking contrasts and seeking similarity are different approaches (1991: 250); on the other hand, however, recent research (C. Anckar, 2005) has demonstrated that the controversy between designs is less clear-cut and dramatic than was earlier assumed. Be this as it may, the article deals with problems that originate primarily, although by no means exclusively, from the frames of the most similar approach. Galtons Problem The logic of the scientific method builds upon a belief that the greater the ability of a proposition to survive repeated tests, the greater the truth that is assigned to the proposition. However, this will hold only if the units that are tested are independent. If a unit is not independent, no new information about the phenomenon under study is obtained by studying it twice, and no additional confirmation of a theory is obtained by counting it twice (Zelditch, 1971: 282283). Comparative research as a rule departs from the assumption that the units that are researched are indeed independent from each other in respect to the variables that are investigated. This condition, however, has always been problematic and is increasingly so in the field of politics, now permeated

International Journal of Social Science Methodology 51

by processes of internationalization and globalization. More often than not, the phenomena that are investigated by political scientists tend to move from one system to another, and the implication of this diffusion is that the relations between phenomena become non-functional. The Nordic region may be referred to as an example. Recent efforts to advance the pursuit of comparative research on, as well as in, the Nordic countries (D. Anckar, 1993b) have to a large extent been inspired by the belief that due to their many similarities, the Nordic countries are especially inviting as objects for comparative research. Thus, even in the 1960s the similarities in the social, economic and political structures of the Nordic countries were seen as a ready-made field of comparative study (Ruin, 1969: 175). However, while certainly appealing, this emphasis on similarity also sets a trap. Probably more than any other family of nations, the Nordic countries are imbued with diffusion. The countries are socially and culturally homogenous, they maintain contacts on an everyday basis, and these contacts are largely institutionalized and integrated. There are obvious structural similarities between the countries; the decision-makers may be expected to confront similar problems and to search for solutions in similar contexts. They can therefore also be expected to turn to the neighboring countries for models and inspiration. Indeed, this happens frequently. It has, for instance, been demonstrated in empirical research that Finnish politics and legislation have a propensity to reflect trends in Sweden, and that elements of Swedish policies have been consciously introduced in Finland (Karvonen, 1981). There is an abundance of other similar instances of spreading across the Nordic region. In fact, there is so much diffusion between the countries that if, in some case, diffusion cannot be found, this arouses curiosity and calls for research into the mechanisms that prevent diffusion (for example Isaksson, 1989). This problem of securing the integrity of cases selected for comparison is the socalled Galtons problem. It is named after Francis Galton, who in 1889, at a meeting of the Royal Anthropological Institute, introduced an objection against the then new cross-cultural survey method. Responding to a presentation of this method, Galton pointed out that traits identified in surveys often spread by diffusion, and that this rendered difficult the assessment of their independence (Galton, 1889, cited here from Naroll, 1965: 428429). In an essay that describes the history and the nature of this problem, Raoul Naroll states that Galtons problem is one of the most important problems of scientific method which face social scientists (1965: 428). Naroll also asserts that Whenever sociologists or political scientists test grand theories about society or culture by correlations, they ought to be aware of Galtons problem (1965: 429). In like manner, other scholars have expressed concern that the problem does not merit enough attention among comparative political scientists. In one of the more sophisticated treatments of the corollaries of the problem, William Mould (1974: 149) concludes that comparative politics here faces a crucial problem in the statistical study of political development. Mould also (1974: 149) takes care to point out that the closed-system-assumption that is embedded in any neglect of the problem needs to be thoroughly researched not only to increase the explanatory power of a particular model of political change, but more importantly, to eliminate an extremely plausible

52

D. Anckar

threat to its empirical validity. Lauri Karvonen (1994: 110), while addressing the very nature of the problem, voices concern that by disqualifying the basic assumption of autonomous units, diffusion may constitute a very difficult problem in comparative research. Although the problem probably cannot be fully resolved, two methods for coping with it at least in part may be considered. First, it is possible to regard the problem not as a problem but rather as an asset, meaning that comparativists may choose precisely to study diffusion, which of course necessarily entails comparison (Hague, Harrop & Breslin, 1998: 275). In other words, the very factor that renders many comparisons dubious may become subjected to rigorous comparative research. While holding the accepted rules of comparative logic and, indeed, of case selection in full respect, comparativists may aim at finding out, for instance, what social, economic and other factors weaken or promote the emergence of diffusion, or what differences there are between policy sectors in terms of susceptibility to diffusion. In short, comparativists may embark upon a process of learning more about who learns what from whom, to quote a relevant title from the literature (Dolowitz & Marsh, 1996). Examples of investigations in this category are studies that approach the spread of strategies and norms concerning constitutional law from a diffusion perspective (e.g. D. Anckar, 2004; Go, 2003). Such studies, then, aim at generalizations about circumstances that enhance or put limits to colonial and similar heritages in the construction and reconstruction of constitutions. The second solution is to choose cases for comparison which can be assumed to have influenced each other to a minor extent only. For instance, one may try to find cases that are at a sufficient distance from each other in terms of time and space (Manheim & Rich, 1991: 221). One rather extreme application of this strategy is of course the searching for asynchronic cases, i.e. cases that are even far apart in terms of time but still display powerful analogies in historical context (Forrest, 1994). Anyhow, while commendable in terms of logic, the recommendation to select cases that are at great geographical distance from each other entails the obvious risk that the countries also represent structural and cultural differences among societies and communities. The risk is, in other words, that the countries are not similar enough to constitute comparable cases. Sometimes, of course, the risk can be evaded. A comparison of Sweden and New Zealand, to pick one example from the research literature (Davidson, 1989), embraces two countries that are at maximal geographical distance from each other, but are still similar in terms of many potential explanatory factors. However, as a rule geographical distance correlates with factors that serve to weaken the extent of similarity: moving from one continent to another often implies moving from one culture to another, from one religion to another, from one social system to another and from one political climate to another. The travelling problem in comparative research, as described by Sartori (1991), Peters (1998: 8693), and others (e.g. Denk, 2002: 7982) indeed needs to be thoroughly considered when selecting cases for analysis. Although much can be said for the principle of avoiding countries with which the comparativist has poor substantive knowledge (Sanders, 1994), it still makes good sense to secure by case selection that the countries which are to be compared represent

International Journal of Social Science Methodology 53

some evident variation in terms of geographical and mental distance. The quite popular families of nations approach (e.g. Castles, 1993; Peters, 1998: 7477) should therefore be applied with discretion and caution, and the view that a comparable case strategy implies a justification of area studies (Landman, 2002: 40) should be accepted only with reservations. For any researcher who confronts Galtons problem, perhaps the best recommendation is also the simplest one. The researcher must be aware of the existence of this problem, and must, on the basis of this knowledge, try to identify and soften the consequences of the problem. It needs also to be said that Galtons problem should not obstruct the pursuit of comparative research. True, Galtons problem is general in nature, but Galtons ghost does still not necessarily appear everywhere. For instance, it may well be the case that two nations are structurally similar, moreover, precisely because of this similarity they will have to face similar problems and eventually end up autonomously with similar solutions (Karvonen, 1994: 111). It is also conceivable that approaches and theoretical assumptions that emphasize the impact of structural factors or diffusion are in fact complementary. Anyhow, they are both justified and must both be accommodated in the practices of research. The Need for Paired Comparisons In customary usage, comparative politics excludes within-nation comparisons and refers to the study of foreign countries only (Rose, 1991: 446447). The ultimate goal of comparative politics, a textbook states, is simply to encompass the major political similarities and differences between countries (Hague, Harrop & Breslin, 1998: 272). In managing this task, however, comparative politics needs to come to terms with the fact that the number of independent states in the world is fairly restricted, the states being, on the other hand, characterized by an abundance of variables that depict similar experiences of aspects of history, culture, economy and society (Peters, 1998: 65). In short, comparative politics faces the many variables, small N problem (Lijphart, 1971: 686). In some early writings, Arend Lijphart, one of the great figures of contemporary comparative politics, developed a classification of approaches and strategies, and suggested several methods for minimizing the above problem (Lijphart, 1971: 686691; 1975: 159). These methods include: (1) Increasing the number of cases as much as possible (2) Reducing the property-space of the analysis by, for instance, combining two or more variables that express an essentially similar underlying characteristic into a single variable (3) Focusing the analysis on comparable cases, i.e. cases that are similar in a large number of important characteristics but dissimilar with regard to the variables between which a relationship is hypothesized (4) Restricting the analysis to key variables While certainly recognizing that statistical methods as well as case studies may be used in comparative analyses, Lijphart still wanted to reserve the term comparative method for the comparable-cases strategy. He defined, in consequence, the comparative

54

D. Anckar

method as the method of testing hypothesized empirical relationships among variables on the basis of the same logic that guides the statistical method, but in which the cases are selected in such a way as to maximize the variance of the independent variables and to minimize the variance of the control variables (1975: 164). In this method, known as the most similar system design, one may find the logic of what John Stuart Mill described as the method of difference and the method of concomitant variation (Lijphart, 1971: 687688; Faure, 1994: 316320; Denk, 2002: 5761). The point of departure for this design is simply that features that are common to a group of states cannot explain differences between the states, these differences being instead caused by features that are different to the group of states. It follows, therefore, that similar systems, given that they also display some interesting differences, emerge as rewarding foci for research in making focused comparisons. They allow for the establishment of quasi-experimental settings, where the impact of operative variables is approached through dissimilarity, whereas the impact of contesting variables is controlled through similarity. The central recommendation of the most similar system approach is, in the words of Giovanni Sartori (1991: 250): choose entities that are similar, if possible, in all variables, with the exception of the phenomenon to be investigated. Intriguing as Lijpharts recommendations for overcoming the many variables, small N problem may seem, they are to some extent logically incompatible. The focus on comparable cases, as Lijphart himself readily admits (1975: 163), will by necessity limit the number of cases that are available for analysis. Countries that are fairly similar in many aspects but differ in the one that is researched do not come in large quantities; cases for comparison must therefore be identified from a fairly limited group of candidates. However, it is possible to avoid this problem to some extent at least. The solution lies in the application of the method of paired comparison. Although a basic tool in qualitative studies (Collier & Mahoney, 1996: 66), this method is, surprisingly enough, seldom used in political research, the latter tending to neglect the ineluctable thesis that selection without reflection may lead to serious problems of inference (Landman, 2000: 44). The method requires the arranging of pairs of units, which are selected to address the requirements of a comparable-cases-strategy. Table 1 provides an imaginary and simplified example. In this example, the research task is to explain differences in political stability between countries, the guiding hypothesis being that regime form (V4) emerges as an important determinant. In examining the validity of this hypothesis, the researcher wants to hold constant the impact of some contesting variables, i.e. variables that may also influence stability. One such variable is country size (V1), the rationale being that because of the closeness between decision-makers and citizens that is characteristic of small units (Dahl & Tufte, 1973: 1215), such units are likely to maintain a high degree of political stability. Another variable is colonial heritage (V2), the rationale being that colonial masters differ in the extent to which they familiarized local elites with pluralistic and peaceful modes of government (e.g. Smith, 1978, Hadenius, 1992: 128133), and a third variable (V3) is per-capita GDP, the rationale being the much-quoted view of Seymour Lipset that the more well-to-do a nation, the greater the chances that it will sustain democracy (1959: 75).

International Journal of Social Science Methodology 55 Table 1 Matched Pairs of States. An Imaginary Example
Variables States S1 S3 S2 S4 V1 (population) 1,1 m 1,7 m 13,3 m 14,2 m V2 (colonial heritage) Britain Britain France France V3 (per-capita GDP) USD 3.200 USD 3.750 USD 24.900 USD 27.000 V4 (regime form) parliamentary presidential parliamentary presidential

Table 1 shows two pairs of states, namely S1 and S3 and S2 and S4 respectively, and the relevant attributes of these states also appear in the Table. As is evident in the table, S1 and S3 are similar or fairly similar in respect of three variables, namely size, colonial heritage and per-capita GNP, but differ in respect of the fourth variable, namely political regime. These states therefore form a frame of reference that enables an assessment of the impact of the variable V4 while controlling at the same time for the impact of variables V1V3. The pair of states, S2 and S4, differ much on variable values from the state pair S1 and S3, but is internally coherent in very much the same manner as S1 and S3. In consequence, also this second pair provides a valid frame of reference for measuring the impact of V4. By this method of arranging pairs, it becomes possible to expand considerably the field of research. For example, there is no principal impediment to pairs of countries such as Finland-Sweden and Nigeria-Uganda appearing simultaneously as test cases in a study that investigates empirically a theoretical proposition. The Pitfall of Binary Comparison Social reality is indeed complex, and any description of this reality entails of necessity a large number of variables and relations. Therefore, the question of selecting variables to be controlled for in the search for comparable cases is essential as well as delicate and demanding. The guiding rule, of course, is that correct conclusions concerning the relationships between variables will follow only if the variables that are left out of consideration do not affect systematically the variables under study (e.g. Nurmi, 1974: 82). The very possibility of paying full attention in the comparable cases method to this rule is, however, often disputed. The leading argument is given by Przeworski and Teune (1970: 34) in their presentation and defense of the most dissimilar method. The argument is that the most similar method tends to overlook the fact that while the number of differences among seemingly similar countries may be limited, they are still invariably sufficiently large to over-determine the dependent phenomenon. In short, to pursue a most similar strategy sufficiently similar cases can never be found. The scholarly response to this criticism has been varied. Lijphart issues (1971: 688) a warning against the application of a too exacting scientific standard, but he also admits (1975: 172) that the criticism is valid, although he argues that the problem of over-determination can be alleviated by an imaginative selection of additional cases,

56

D. Anckar

particularly at the sub-national level, where similarity is plentiful. In a similar vein, other scholars have pointed at the possibilities inherent in geographical manipulation. One relevant argument is that a focus on less than hemispheric-wide data may prove useful: instead of applying a theory throughout the Americas, sub-regional emphases on, for instance, the Central American context or the region of Grancolombia should be applied (Martz, 1994: 254255). Still other scholars (Dogan & Kazancigil, 1994: 8 9) take a more critical view of the Przeworski and Teune argument and find that the criticism is not justified because similar countries should not be chosen on simple characteristics but rather on the criteria of basic analogies such as the sociological context or the socio-economic level. However, it is not the ambition of this article to solve this complex and difficult problem, which becomes further aggravated by the risk that similarities perceived by one researcher may be differences for another (Landman, 2000: 41). Rather, the point to be made here is that the problem of over-determination throws a very unfavorable light indeed on binary comparisons. As obvious, for instance, from the popular saying that one can greatly improve insights into one country by immersing oneself in the study of at least one other country (Daalder, 1987: 16), binary comparisons are often advocated in the comparative politics discourse. Among studies that are regarded as classics in comparative political research are several binary enterprises (e.g. Heclo, 1974; Lipset, 1990). Binary comparisons are also quite frequent in the practice of comparative research. For instance, a study of the position of comparative politics within political science has reported that of 132 articles in some leading European journals, 40 compared only two countries (Page, 1990: 47). This is, however, an unfortunate practice, and a rather obvious indication that the selection of cases for comparison often remains a superficial exercise, that is guided less by theoretical insight than by an interest in the detail of particular countries or perhaps even by familiarity and convenience (Peters, 1998: 66). An illustration of the pitfalls of binary comparison as well as the gains of adding at least one more case to the research design is given below in Table 2. Let us first assume a binary comparison between the two states A and B, which differ markedly in terms of the outcome in the dependent variable. This difference may have come about as a result of any of the five independent variables (IDV 15) in the table, all of which are presumed here to be theoretically relevant and all of which take differTable 2 Expanding Binary Comparisons. An Imaginary Example
Units Independent Variables IDV1 IDV2 IDV3 IDV4 IDV5 Dependent Variable A 213 low 1,3 high 60 53 B 375 high 2,7 medium 80 75 C 408 high 2,3 high 65 51

International Journal of Social Science Methodology 57

ing values for A and B. In consequence, a variety of possibilities must be considered, and in this process the method cannot in itself justify conclusive statements. As far as the method is concerned, all possibilities are equally relevant (Meckstroth, 1975: 134). To reduce this amount of over-determination, for purposes of control, at least a third state must be added to the research design. This state is named C in the table and is here chosen to match state A in terms of some independent variables as well as state B in terms of some other independent variables. Given the outcome on the dependent variable that is inserted in table 2, as evident now from the table, the independent variables 13 may be removed from the list of potential explanatory factors. These variables have similar or fairly similar values in states B and C and are, therefore, unable to explain the difference between B and C. From similarities, of course, only similarities can follow. Also, the three variables have different values in states A and C, which report similar outcomes in the dependent variable. In this example, then, the replacement of a binary comparison with a triangular one has contributed to eliminating some possible causes, and, thereby, to refining the explanatory model. In that sense, a move from binary comparison to triangular comparison implies much more than adding just another case to the research population. Indeed, the step from two cases to three is much more decisive and rewarding than the step from one case to two or from, say, four cases to five. This is a message that needs also to be well received and applied in Nordic political science communities, which are too much occupied with within-region comparisons on a binary basis. A mapping of the content during the 1980s in four Nordic political science journals indicates (D. Anckar, 1993a: 116117) that a good quarter of all comparative politics writings were about binary comparisons, in which a Nordic country was often compared to another Nordic country. This practice, in order to overcome over-determination as well as Galtons problem, should be replaced by a practice that adds to the binary Nordic comparison at least one non-Nordic case. Many Sets of Units, Not One Quite often in political analysis the research design is one of explaining by means of variances in several independent variables the variance in one dependent variable. One example of this approach is an important study by Hadenius (1992: 73142), who inserts the Third World states of the world on a scale that measures the democratic performance of the states and then explains the differences between the nations by reference to ten different explanatory factors. In this example the number of cases is large, and statistical methods were therefore employed. When the number of cases is clearly smaller the method of comparable cases comes to use; in applying this method, however, the researcher confronts a problem that renders necessary the operation with many and not only one set of states. The problem is illustrated here in Table 3. In this table the states A and B are very similar and in fact almost identical in regards to values on four independent variables (IDV 14) that relate to democratization. These variables are about social capital, ethnic fragmentation, literacy, and international exposition. In contrast, the fifth independent variable (IDV 5), which is about

58

D. Anckar

Table 3 The Inadequacy of One Set of Cases. An Imaginary Example


Independent Variables IDV 1 (civil society prevalence) IDV 2 (fragmentation) IDV 3 (literacy rate) IDV 4 (international exposition) IDV 5 (per-capita national income) State A wide 0.76 100% high high State B wide 0.74 98% high middle

economic performance, receives different values in the units. Let us now assume that we are, like Hadenius, interested in explaining differences between nations in terms of democratic standard, and that we wish to test a theory that points at economic development as an essential determinant of democracy. Since the states at hand differ in terms of economy, but not in terms of other democracy determinants, the setting is adequate for the purpose of theory testing. Now, if an inspection in A and B of the values of the dependent variable (democracy) suggests that the units are different in this respect as well, it follows that this variation may be referred to the variation in IDV5. The theory is therefore correct: there is a connection between economic development and democracy and economy variation may be assumed to be the cause of democracy variation. On the other hand, if no or very little variation is at hand in the dependent variable, it follows that IDV5 does not possess explanatory power. Economic variation has not produced democratic variation, and since democracy cannot, in consequence, be explained by economy, the research must be extended to cover other explanatory factors as well. At the bottom of it is now that the set of states that appears in the example cannot be used for any other purpose than testing the explanatory power of IDV5. If the economic development that is measured by this variable has failed in explaining democracy variance, one may be interested in probing, instead, the impact of ethnic fragmentation on democracy. Assuming that IDV2 describes in terms of a fragmentation index (e.g. Anckar, Eriksson & Leskinen, 2002) a central characteristic of the ethnic setting, one now confronts and recognizes the futility of advancing further without changing the set of nations. The impact of ethnicity is namely controlled for by means of case selection in the example, meaning that there is no significant difference between A and B in terms of ethnicity. The potential explanatory power of this factor can therefore not be observed, as it has been removed from the research design. When and if one chooses to regard IDV2 as an explanatory variable, this factor needs to receive different values for A and B while the other factors, namely IDV1 and IDV35, must receive values that satisfy the demand for similarity. The selection of cases is therefore useless and one needs to find by means of comparison and a systematic search a new and different set of nations. Although these remarks may seem trivial, they are still necessary. Very few focused comparative studies in political science employ several parallel sets of nations. Rather, the testing of different questions through the use of the one and same set of nations

International Journal of Social Science Methodology 59

appears a rule. One recent example, picked at random from a multitude of similar cases, is an otherwise imaginative and thorough study by Hans Agn (2004), which unfortunately makes use of the same set of units (France, Sweden, United Kingdom and the European Union) for answering a row of different questions that concern the relation of democratic theory and internationalization. The comparable cases strategy does not, however, accommodate this oversimplified method of selection which is in conflict with the very nature and logic of the strategy. It may be added that the weakness of the selection method of course prevails in research endeavors that do not apply or apply only implicitly the comparable cases strategy. The main problem is one of disparity between questions and research cases from which to extract answers, and while making its presence felt in a very straightforward way in the comparable cases strategy, the problem does not by any means disappear in other and less rigorous settings. References
Agn, H. (2004). Democracy Reconsidered. Stockholm: University of Stockholm, Stockholm Studies in Politics, 104. Anckar, C. (2004). Determinants of the Death Penalty. A Comparative Study of the World. London and New York: Routledge. Anckar, C. (2005) Some Reflections on the Applicability of the Most Similar Systems Design and the Most Different Systems Design in Comparative Research. Paper, XIV Nordic Political Science Congress, Reykjavik, Iceland, August 1113.2005. Anckar, C., Eriksson, M. & Leskinen, J. (2002). Measuring Ethnic, Linguistic and Religious Fragmentation in the World. bo Akademi University, Department of Political Science, Occasional Papers Series, 18, 714. Anckar, D. (1993a). Comparative Research in the Nordic Countries: Overcoming Ethnocentrism? Scandinavian Political Studies, 16, 107126. Anckar, D. (1993b). Nordiska frbundet fr statskunskap som komparativistisk arena och aktr. Politiikka, 35, 173186. Anckar, D. (2004). Regime Choices in Microstates: The Cultural Constraint. Journal of Commonwealth and Comparative Politics, 42, 206223. Castles, F. G. (Ed.) (1993). Families of Nations: Patterns of Public Policy in Western Democracies. Aldershot: Dartmouth. Collier, D. & Mahoney, J. (1996). Insights and Pitfalls. Selection Bias in Qualitative Research. World Politics, 49, 5691. Daalder, H. (1987). Countries in Comparative European Politics. European Journal of Political Research, 15, 321. Dahl, R. A. & Tufte, E. (1973). Size and Democracy. Stanford: Stanford University Press. Davidson, A. (1989). Two Models of Welfare. The Origins and Development of the Welfare State in Sweden and New Zealand, 18881988. Uppsala: Publications of the Political Science Association in Uppsala, Nr 108. Denk, T. (2002). Komparativ metod frstelse genom jmfrelse. Lund: Studentlitteratur. Dogan, M. & Kazancigil, A. (1994). Introduction. In M. Dogan & A. Kazancigil (Ed.s), Comparing Nations. Concepts, Strategies, Substance, pp. 113. Oxford: Basil Blackwell. Dolowitz, D. & Marsh, D. (1996). Who Learns What From Whom: A Review of the Policy Transfer Literature. Political Studies, 44, 343357. Faure, A. M. (1994). Some Methodological Problems in Comparative Politics. Journal of Theoretical Politics, 6, 307322.

60

D. Anckar

Forrest, J. B. (1994). Asynchronic Comparisons. In M. Dogan & A. Kazancigil (Ed.s), Comparing Nations. Concepts, Strategies, Substance, pp. 260296. Oxford: Basil Blackwell. Galton, F. (1889). On a Method of Investigating the Development of Institutions Applied to the Laws of Marriage and Descent. Journal of the Royal Anthropological Institute, 18, 272. Go, J. (2003). A Globalizing Constitutionalism? Views From the Postcolony, 19452000. International Sociology, 18, 7195. Hadenius, A. (1992). Democracy and Development. Cambridge: Cambridge University Press. Hague, R., Harrop, M. & Breslin, S. (1998). Comparative Government and Politics, 4th edition. London: Macmillan Press. Heclo, H. (1974). Modern Social Policies in Britain and Sweden. New Haven, Connecticut: Yale University Press. Holt, R. T. & Turner, J. E. (1970). The Methodology of Comparative Research. In R. T. Holt & J. E. Turner (Ed.s), The Methodology of Comparative Research, pp. 120. New York: The Free Press. Isaksson, G-E. (1989). Resultatls omlokalisering. bo: bo Akademis Frlag. Karvonen, L. (1981). Med vrt vstra grannland som frebild. bo: Meddelanden frn Stiftelsens fr bo Akademi forskningsinstitut, Nr 62. Karvonen, L. (1994). Diffusionsproblemet i studiet av komparativ politik. Politiikka, 36, 109115. Keman, H. (2002). Comparing Democracies: Theories and Evidence. In H. Keman (Ed.), Comparative Democratic Politics, pp. 3261. London: Sage Publications. Landman, T. (2000). Issues and Methods in Comparative Politics. London: Routledge. Lijphart, A. (1971). Comparative Politics and the Comparative Method. The American Political Science Review, 65, 682693. Lijphart, A. (1975). The Comparable-Cases Strategy in Comparative Research. Comparative Political Studies, 8, 158177. Lipset, S. M. (1959). Some Social Requisites of Democracy: Economic Development and Political Legitimacy. The American Political Science Review, 52, 69105. Lipset, S. M. (1990). Continental Divide: The Values and Institutions of the United States and Canada. New York and London: Routledge. Manheim, J. B. & Rich, R. C. (1991). Empirical Political Analysis, 3rd edition. New York: Longman. Martz, J. D. (1994). Comparing Similar Countries. In M. Dogan & A. Kazancigil (Ed.s), Comparing Nations. Concepts, Strategies, Substance, pp. 239259. Oxford: Basil Blackwell. Meckstroth, T. (1975). Most Different Systems and Most Similar Systems: A Study in the Logic of Comparative Inquiry. Comparative Political Studies, 8, 132157. Mould, W. B. (1974). On Getting Nothing for Something. A Note on Causal Models of Political Development. Comparative Political Studies, 7, 139164. Naroll, R. (1965). Galtons Problem: The Logic of Cross-Cultural Analysis. Social Research, 32, 428451. Nurmi, H. (1974). Causality and Complexity. Turku: Annales Universitatis Turkuensis, Series B, Tom 131. Page, E. C. (1990). British Political Science and Comparative Politics. Political Studies, 38, 438452. Peters, B. G. (1998). Comparative Politics. Theory and Methods. London: Macmillan Press. Przeworski, A. & Teune, H. (1970). The Logic of Comparative Social Inquiry. New York: John Wiley. Ragin, C. C. (1987). The Comparative Method: Moving Beyond Qualitative and Quantitative Strategies. Berkeley: University of California Press. Ragin, C. C. (2000). Fuzzy-Set Social Science. Chicago: The University of Chicago Press. Rose, R. (1991). Comparing Forms of Comparative Analysis. Political Studies, 39, 446462. Ruin, O. (1969). Political Science in Sweden in the Post-War Period. Scandinavian Political Studies, Yearbook Series, 4, 171182. Sanders, D. (1994). Methodological Considerations in Comparative Cross-National Research. International Social Science Journal, 46, 4349. Sartori, G. (1991). Comparing and Miscomparing. Journal of Theoretical Politics, 3, 243257.

International Journal of Social Science Methodology 61


Smith, T. (1978). A Comparative Study of French and British Decolonization. Comparative Studies in Society and History, 20, 70102. Smithson, M. (1987). Fuzzy Set Analysis for Behavioural and Social Sciences. New York: SpringerVerlag. Vanhanen, T. (2003). Democratization. A Comparative Analysis of 170 Countries. London and New York: Routledge. Zelditch, M. Jr. (1971). Intelligible Comparisons. In I. Valier (Ed.). Comparative Methods in Sociology, pp. 267307. Berkeley: University of California Press.

Anda mungkin juga menyukai