There is some- question of whether political science would want to be a more experimental scieiuie, even were it possible- When a scientist condiit-ls .an experiment the purpose is lo hold as m;>ny factors as possible constant, ill order to permit a single independent variable to opcralc upon a single dependent variable. Unfortunately (or foi-tuna(ely). tlir real world or politics is not [hat sterile and controlled, ;ind tlierc are a liost of seemingly extraneous faclors tliat inlluence tlie w;iy people vote, or policies arc made, or interest groups lobby. Thus. ;iny great mvcslmem in experimentation might rob tlie discipline of much of its descriptive richness and attention to complexity, which ;ti-c important for understanding what makes politics in France so dillereni from politics in its neighbour Italy, much less politics in Nigeria. A great deal of political life involves (lie interaction of numerous forces, so that any artificial isolation of the causal factor would almost certainly be misleading as well as less interesting.
Real countries present both problem.1; and benefits for comparative politics. The benelits are obvious. Within those real countries occur the real, complex and convoluted sequences of events that arc of so much interest to the student of tlie subject. On tlie other hand, tlie complexity of real political life ine;ins (hat variables come to tlie researcher in large bundles of factors that arc almost inextricably intertwined. It is then up to the researcher to disentangle the sources of variance, to coniexlualise the findings, and to provide as useful a 'story' about politics as lie or she can.
If the claims of political science to be an empirical method are based largely on comparative analysis, so too is a good deal of the substance of normative political theory based implicitly on comparison. Normative polifical theory is directed at identifying and producing 'the good life' in tlie public sphere, and to some extent arguments about tlie desirability ofdinercnt forms of government are based on the comparative observations of llic propounders of those theories. Certainly, not all normative political theory is ili;it instrumental in its advocacy of particular solutions to (lie problem of government, but much of tlie analysis lias been. Thomas Hobbes had observed that the absence of effective government - his 'state of nature' - during the period of the English Civil War produced less desirable social outcomes than might occur under more effective government- r'rom tlial observation, lie extrapolaied to argue lor sirong government, even ;it the possible expense of some civil liberties. Even when there are no real-world observations of a pariicular version of 'the good life', normative analysts often engage Jn mental experiments based upon

their generalised empirical knowledge of politics, drawn largely from informal comparative analysis. This comparison could be conducted across time as well as across countries.
The real world of comparative government is therefore (lie laboratory (or political scientists lo delerininc what works and what docs not, as well as to demonstrate important theoretical relationships among variables. It is a laboratory in which the 'experiments1 are designed by other people - politicians, civil servants, interest group leaders - for their own purposes and not for those of social scientists. Therefore, tlie difficult but yet crucial task for the comparative analyst is to devise the methods necessary to construct meaningful theoretical and analytic statements about government and politics within those complex and largely unplanncd settings. We must develop (lie methodologies for entering into tlie real world ol politics in a number of different settings, and extracting meaningful information. That information needs to be descriptively accurate, but it also need to be more than that. Those statements must also link observations made in one political system with similar observations made in another, or link the observations in one system witli general propositions about politics. For a researcher to be able to sav what happens in a single country may be interesting, but it is generally insufficient to mean that he or she is really engaged in tlie study of comparative politics.
Studying the single country can however be argued to be comparative analysis if it is guided by implicit comparison with other systems, or if the research has a strong connection to theories based on comparison (Verba, 1967). Indeed, carefully selected and craftcd studies of individual countries can lie crucial for testing theory (see Key Text 5.1). One country may he tlic crucial case for testing a hypothesis - iftlic proposition works llicrc then il will work anywhere - so that the case-studv is the most cfiicicnt means of addressing tlie theoretical proposition. For example, elite llicory and elite pacts have become important means of explaining tile success ofdemocralisation in transitional regimes (Burton and Higley. 1987; Higlcy and Gunthcr, 1992; Zhang, 1994; Hagopian, 1990). These pacts involve members of tlic contending political factions reaching agreements to govern peacefully, even in llic face of intense disagreements between them, not only on policies but also on llie structure of government itself. II elite pacts were found to function lo manage a deeply divided society, for example, in Lebanon, then they might be expected to have the potential to work almost anywhere. Lebanon was discussed

as an example of conflict management through consociationalism prior to (lie outbreak of the Civil War. It now remains to be seen if other means of conflict management can work.
The virtue of comparative analysis, especially when the analysis is limned to a single case or a few cases, is that it forces greater specificity on (lie researcher. What are the factors in a country tlial produce certain observed patterns of behaviour? If tlie study is conducted outside a comparative framework, i[ is easy for [he rcsearclier to make a number of assumptions about the exceplionalism of tlie case- In a more comparative mode of analysis, tlie similarities as well as the differences become evident, and the researcher must think more clearly about the root causes of the performance of the system. Comparative analysis can also he useful even when one is thinking about political behaviour within one's own country. What is it that makes British or American politics different from those of other countries, and what evidence would a researcher need to muster in order to substantiate tlial argument about the differences? What are the 'shadow cases' that would be most useful when making judgments about British politics? Are they the Commonwealth countries, the USA, or tlie countries of continental Europe?
We do not need to apologise at all for the contributions of comparative politics, despite the obvious disability of moving the study forward and of making it even more useful for analysts and perliaps indeed for practitioners in (lie 'real world'. This body of research has been able to cope with immense changes in the surrounding political world - decolonisation and political development, the development of regional bodies such as the European Union, democratisation in Latin America, Southern Europe, and Central and Eastern Europe, and the emergence of die 'little tigers' in Asia - and lias something important to say about these changes. There lias also been tlie gradual accretion of a rich -store of knowledge about individual political systems, wliicli can serve not only as grist for theoretical exercises but also simply for a better understanding of those systems. Finally, there have been interesting and important theoretical developments. Tliese developments have not been uncon-trovcrsial, but again, they have added to the store of weapons at the disposal of those political scientists who go into the field attempting to understand politics and political behaviour in a comparative manner.
In (his book, tlie concentration on the development o! theory is in contrast to the roots of comparative political analysis, which lie more in llie description of formal, legal institutions and in prescribing

governmental structure based on ideas of democracy and 'good government'. This theoretical development in the discipline has been praised by some, especially those who fought so hard to break the domination of formalism (Eckstein, 1963). It lias also been condemned by more traditional scholars (Maclntvre, 1978), who argue that (lie differences among countries are so subtle, and so ingrained in cultural and historical factors, that they arc not subject to 'scientific' analysis; however, while attempting to be sensitive to the difficulties of comparison and to the subtleties of national differences, we will nevertheless operate within the more scientific paradigm.
Forms of Comparative Analysis
The task of understanding politics comparatively is made all tlie more difficult by the number of the different forms of analysis that are commonly labelled 'comparative polities' in textbooks and in political science curricula. On the one hand, there are studies ol'an individual country (other than that of the autlior and the intended audience) that are labelled 'comparative'. At [lie other end ofllie spectrum are studies in wliich individual countries are submerged in a multitude of other countries, and become 'data points' in a statistical analysis much more than they are vital political entities. Both types of studies have their appropriate place in the discipline, despite efforts by scholars from tlie other camp to weed them out, and both can claim to make important contributions to the study of comparative politics. The problem lies in attempting to understand both methods of analysis within that single rubric of'comparative', and to understand the contributions each type can make.
Despite their apparent dissimilarity, all these approaclies to comparative politics must confront a fundamental trade-olf between llie respective virtues of complexity and generalisation. That is, the more an approach (the case method for example) takes into account the context and the complexities of any one political system, or a limited range of systems, the less capable llial rcscarcli strategy will be of producing generalisations about politics. Similarly, the more an approach fsucli a.s statistical modelling) attempts to furnish generalisations and to le.st broad theories about politics, the less nuance about panicLihir political systems it is able to permit in its analysis.
This trade-oil'is inherent to llie research process of comparative politics, and (o some exicnl to oilier component parts ol' tlie social

sciences, and (lie researcher must make decisions about where along that undei 'lying continuum ofspccificily to pl;icc a particular piece or research. Dogan and Massy (1984; 127) note that 'at times conipar-ativists will emphasise similarilies, al linies differences. They will tend to look for differences in contexts that arc roughly similar, or ... will try lu find analogies in contrasting political systems.' Research also can be strengthened by the interaction of the two styles ofcompar-ison, altliougli that makes dillicultics for any individual scholar.
Another way to think about these fundamental trade-ofis in comparative analysis is to contrast conligurativc and statistical methods of analysis (sec Lijpliarl, 1971). In the configuralive approach. llic primary purpose is the thorough description of a case or cases, so lhat the consumer of the research will be capable of comprehending the logic of political life in that limited number of settings. Tlial understanding of the case will be, however, rather deep, or 'thick', and would include a number of aspects of social and political life found in those systems. The anthropologist Clifford Geertz (1973) used the phrase 'thick description' to describe interpretative work in his discipline. By this, lie meant using detailed description as the precursor of interpretation and theory development in the social sciences. The descriptions generated in a configurative analysis tend to range across institutional and micro-political 'variables', rather than just on the specifics of a case. That description would also involve an understanding of the social, cultural and economic context of politics. Tills is the kind of comparative analysis associated with the scholar who devotes his or tier career to the study of one or a few countries, but feels verv much at home in tlie culture and politics of lliat limited sample of countries. The understanding developed through tlic extended analysis of llie single case becomes almost intuitive, so lhat conveying it to others may become verv dinicult. One of the standard descriptions of science is thai its findings and methods must be 'intersubjectivcly transmissible'. If the findings are excessively intuitive, they do not necessarily contribute to political science.
One of the more interesting, and confounding, features of comparative analysis from the conliguralivc perspective is that researchers from dinerent Countries may apply (heir own particular national lens to ihe same dala. This i.s true not only in terms of the selection and treatment of the cases, but also in (lie implicit comparison with the home country Isee U'iarda, 1981}. Tocqucville'?; (1946) magisterial study of politics in llie United States was written very much vvim

the French society and French politics of his lime in mind. The strong position of American political science in the Held of comparative politics has meant lhat a large proportion of single country studies use (lie United States as the implicit if not explicit comparison. Edward Page (1990), liowcvcr, lias documented a British tradition in comparative politics lliat tends to be less theory-conscious and less quantitative than American configurative studies. Given lliat the individual scholar is ttie principal rcscarcli 'instrument' in tilis type of study, there may be a great deal of extraneous and error variance added to these studies (see Key Text 9.1), and the -study may be comprehensible only in the national context within wliicli it was written.
Slalistical explanation has a fundamentally din'erciH purpose. It attempts to test propositions about tlic relationship among political variables across countries and in a variety of settings. To some extent, the characteristics of these settings otiicr than those explicitly included in the model arc assumed to be irrelevant, so long as they do not confound measurement oflhe variables in which tlic rcscarclier is interested, or introduce oilier forms of extraneous and error variance (see below, pp. 30-2). Further, statistical explanations tend to leave unmeasured a number of factors that might be central to the more descriptive and convoluted explanations provided through config-uralive analysis. To the extent that cultural factors arc a part of the statistical explanation, they tend to be included only as residual explanations, and to appear only when alt oilier measured factors prove inadequate to explain tlic dependent variable in question. In fairness, llic-re have been some attempts to measure cultural factors more explicitly, and lo include them in tlic statistical models. For example, llie wealth of data coming from llie World Value Study is used increasingly to botli describe national political cultures and to attempt to explain oilier political phenomena (Inglehart, 1997;
Kaasc and Newton, 1995).
At tlie extreme, sialislicat analysis ceases to identify countries as countries, hut instead conceptualises them as packages of variables. Two major students of comparative politics, for example, argued lliat llie ultimate purpo.se of tlie comparative exercise should be to eliininale proper names of countries and to think primarily in terms of concepts and variables IPr/.eworski and Tcune, 1970: 30). 'lliat view i.s probably too exirrme lor inosi .students of comparative politics, even those inleresled in (|ii;intit;ilivc analysis- Tlie variable-driven perspective on comparison does, however, point to tlic need to

direct some scholarly attention away from the specifics of thick
description of individual countries into the development and testing of hypotheses that arc meaningful cross-nationally. As we will point out below (pp. 139-54), however, more formalised comparative analysis need not, and should not, abandon totally the traditional concern with real cases and real countries, nor need it be entirely
quantitative in order to be able to 'lest' hypotheses.
i
Quantity and Quality
Although the familiar con figurative, case-study approach to comparative politics and the statistical approach to the same phenomena may appear extremely different, the fundamental logic of research comparison that undergirds them is actually very similar. As King et ai. (1994; sec also Kriizer, 1996) have pointed out very forcefully, any meaningful differences between quantitative and qualitative research are almost always exaggerated by the advocates of both styles of enquiry. Conducting either style of social research properly depends upon the same issues of research design, each is prey to the same potential errors, and each requires interpretation that goes beyond their available data. In particular, both types of research lace the task ol developing research designs that maximise the observed variance in their dependent variables that is a function of the presumed independent variable or variables. Further, the research design - whether quantitative or qualitative - also must minimise error variance, or the observed variance resulting from random sources such as measurement error and unreliabititv.
As well as error variance, dealing with extraneous variance is a crucial problem lor research design. Qualitative researchers often do not think in terms such as independent and dependent variables, but their research problems can certainly be expressed in that m;mner. A scholar using the case-study method must be sure thai [lie causes to which he or she is attributing the observed outcomes are indeed the 'true causes', and not ;i function of other factors that might as easily produce [he observed outcomes. While statistical research can cope with the problem of other pos.sil.ile causes by introducing control variables to identify and quantify the confounding causes, qualitative research musi deal with tin- confounding factors through careful research design, greater allcinion lo ihc proper selection of cases, and fairness to all causes when doing the research. Of course, iflhe statistical researcher does not identify, via theory or personal insight,

(lie appropriate control variables then the analysis will be no more effective in sorting out confused causality than will the qualitative analysis. There is a fundamental dilemma here for the researcher. He or slic needs a theory to guide the research if it is to be truly comparative, but that theory may colour the research. We all have a tendency to sec what we want to sec, or what we set out to find.
The argument dial quantitative and qualitative research are more similar than dissimilar is not uncontroversiat. Advocates of both styles of enquiry argue for (lie superiority of their own method. In addition, scholars who have no particular methodological axe to grind raise some doubts about the similarity of the methods in practice. For example, Ronald Rugowski (1995) argues that die admonitions of King el al. (1994) noi to select cases on values of the dependent variable (see below, pp. 36(1') are apt for statistical analysis, but would make much case-research less useful than it is. The strength of much case-analysis is that it samples {sic) purposefully on the dependent variable to be able to test theory in [lie most diuicult setting- Likewise, Collier (1995) expresses some doubts about the assumed goal of generalisation implied by the wedding of the methodological traditions.
Although Rogowski and Collier do raise some questions about the similarity of quantitative and qualitative analysis as presented by King el al., they are si-ill working in ilie same positivist, empirical tradition of analysis. There arc other scholars who would argue tliat qualitative, interpretative analysis is the only real way to consider comparison (Maclntyre, 1978). For these scholars, all this discussion of variance would be irrelevant; what is important is individual interpretation of political events. This volume obviously takes a more posilivisi position, and argues thai we really need to remove, as much as possible, the individual and his or her idiosyncracies from the research if we want to be able to make statements about politics that are more general, more usable, and testable.
Types of Comparative Studies
The study of comparative politics is an extremely diverse intellectual enterprise. This diversity is both ;i strength and a weakness of comparison On the one liand, Wiarda (1986: 5) argues that the field is so diverse thai it can scarcely be called a field at all. On the oilier liand, Verba (1986: 3G) argues ih;it the heterogeneity of the field will

continue, and th;it this diversity is a source of strength rather than of weakness. Vcrba argues that the openness o(' (lie ticid to various theories and methodologies helps to maintain its vitality and its capacity to cope with a rapidly changing political world. Any premature closure of content or method might therefore, it is argued, inhibit intellectual progress.
At least five types of studies are classified as being components of comparative politics. We can question whether some of these studies arc truly 'comparative', given (ha( they cither focus on a single country other than that of the scholar doing the study or that they arc so general tlial individual countries get lost- All five types of research do, however, find themselves treated as part of comparative politics within the discipline. Tlieac live tvpcs arc:
(1) Single country descriptions of politics in X, whatever X may be (examples .ire Anderson, 1983; Fitzmaurice, 1981; Rose, 1989;
Lal, 1986; Ramage, 1995);
(2) Analyses of similar processes and institutions in a limited number of countries, selected (one expects) for analytic reasons (Bendix, 1964; Moore, 1966; Skocpol, 1979; Lipset and Rokkan, 1967; Tilly, 1975; 1993; Collier and Collier, 1991);
(3) Studies developing typologies or other forms of classification schemes for countries or subnalional units, using llie typologies both to compare groups of countries and to reveal something about the internal politics of each political system (Lijphart, 1990; Efazar, 1987; Bcbler and Scroka, 1990);
(4) Statistical or descriptive analyses of data from a subset of llie world's countries, usually selected on geographical or devclop-mcnial grounds, testing sonic hypoihe.sis about the relationship of variables within lhal 'sample' of countries (Lange and Garrett, 1985; Amcs, 19K7; Kaasc and Nevvion, 1995; Hyden and Brattou. 1992); and
(5) Statistic;!! analyses o( all countries of llie world attempting to develop paiicms and/or le-sl relationships across the entire range of political systems (Banks and Tcxtor, 1963; Banks, 1971: Rumrnet, t972; 1979: Russell, 19(34; Sullivan, 1996].
Although tliese arc very didcrrnt (ypes of research exercises, requiring a variety ofdinei-cnl research skills fbolh methodological and theoretical), they all claim ,to be viable components of comp;ir;ilive politics. Further, ihey ;ill (;in claim rightfully lo make some signili-

cant contributions to die development of thai field of enquiry. There is often a tendency of scholars to select one mode of enquiry and defend it as the best viable approach to enquiry, or even the only one, but at least for this one social science a great deal is to be gained from eacli research strategy. Were there time and money enough, the optimal strategy would appear to be to move back ;ind forth beiwccn large-scale statistical studies and the more descriptive analysis of individual countries, refining concepts and theories as one goes. Unfortunately, one individual scliolar rarely lias the skills necessary to do all those things, so that this iterative research agenda must be shared, often in an extremely uncoordinated and haphazard manner.
Single Country Studies
The single countrv case-study has perhaps (lie least claim to advancing the scientific status of comparative politics, although il is also perhaps the most common form of analysis in me discipline. The obvious weakness of this approach is thai il is not really comparative, but ratlier is the explication of politics 'someplace else'. As Sartori (1991: 243) has said:
a scliolar who studies only American presidents is an Amcncanist, whereas a scholar who studies only French presidents is a comparativist. Do not ask me how this makes sense - it does not. The fact remains that a field called comparative politics is densely populated by non-comparativists, by scholars wlio have no interest, no notion, no training comparing.
Despite Sarlori's negative views, corn [in ra live politics convenlionally has been dominated by studies of politics 'somewhere else'. There are now any number of books labelled 'Politics in Country /V; an analysis of the electronic card catalogue at my univcrsily library revealed 371 separate books wilh such a title, or 'The Government and Politics of Country X\ covering over fifty dinereiit couniries, ;>nd even more monographs and case studies of particular peritHls of time, events, and/or institutions in a particular country. In almost every country, the politics of that country is taught as tlie basic class in political science, with comparative politics beins; politics occurring everywhere else - even if it only a single country. Governments ol (lie United Stales and UK .ire rarely taught as components of comparative analysis in their home country, even though each is just another

example of an industrialised democracy that needs lo be understood
comparatively.
Limitations or language, educalion and funding tend to force scholarship in the direction of the single country monograph, and, despite some almost inherent weaknesses, this style of research docs make numerous important contributions. In llie first place, these individual Country volumes can be the grist for die mill for scholars more interested in direct comparison and theory. This is especially true if a scries of books is written with a common theoretical framework in mind, or if any one country study is written in a self-consciously theoretical manner by itself. The series of country study books appearing in the 1960s and 1970s under die general editorship ol Gabriel Almond and Lncicn Rye provides a good example of the possibility of implicit comp;irisun through a series of single country studies. These books included Rose (1989), Ehrmann (1992), Edinger (1977), Barghoorn and Remington (1986) and Kolhari (1970). They were later compiled into Almond et al. [ 1973). Another example of a coherent series of books would be that on 'policy and polities' in a variety of developed political systems, including Ashford (1981, 1982), Hecio and Madsen (1987)^ and Tuohy (1992).
A second possible virtue of die -single case-study is to explicate a concept thai appears to be particularly evident in one national setting and to use the country study to develop that concept. For example, the Netherlands is (or at least was) an archetypical case ofconsocia-tionalism. Lijpharl's (1975a) analysis of Dutch politics provided a great deal of general information not only concerning that country, but also in particular about the Dutch approach lo managing deep-scaled social cleavages. That concept lias since been applied, intellectually if not pi-aclically, to other political systems (sec p. 94). Further, the coiisociational concept lias been applied to tlie Netherlands several additional times to assess the degree of political change within that system (Mierlo, 1986; Toonen, 1996). Concept-defining studies sucli as Lijpbart's may engender unwise conceptual stretching in the future (sec pp. 86-93), but they also help fill the conceptual storehouse of comparative politics.
Another useful example of the concept-defining study would be Robert Piitnain el a/.'s work on Italy (1993) and the need for a 'social capilal' or 'social infrastructure' in society (or successful democracy. Putnam's research monitoi-rd iliL- development uf social capital -ilieaiiing membership in organisations ol all sorts - across time and across different regions ol' Italy, finding tliat diose regions without

adequate 'social capital' round it difficult to have viable democratic instilinions. This concept has proved to be useful well outside the IxHindiiries of that one system. Putnam has used it to identify patterns of social and political behaviour in the United Stales that he finds potentially damaging to democracy there (Putnam, 1993). Similar analysis of social capital and die social basis of democracy has also been done for other European countries (Percz-Diaz, 1994), the Unilcd Slates (Pulnam, 1995) and for some African countries (Gyimah-Boadi, 1996). The concept of social capital may generate some difficult empirical and llieoretica! questions (Tarrow, 1996), but it has proven to be useful for comparative political analysis.
Process and Institution Studies
Anollier long and honourable strand of comparative political analysis is to select a small number of instances of a process or an institution thai appear similar (or at least appear "comparable') in some important ways and then use those instances to illuminate the nature of either die process or the institution itself, or the politics of the countries within which it occurs. In practice, these case studies are often capable of saying a good deal about the process, as well as a great deal about the countries. Further, time becomes an important element of the analysis, pointing to additional possibilities of comparisons across lime as well as across political systems.
The purpose of such studies is somewhat different from that of the whole-system comparisons already discussed. It is not lo describe and (implicitly I compare wliole systems, but rather to develop lower-level comparisons of a particular institution or political process. To some extent, these institutions and processes are assumed to lie almost independent of the setting within which they occur. Tims, these Hfholars arc employing comparative data to develop a theory (implicit or explicit) of aspect of political life. For example, (lie monumental historical studies mentioned above tend to develop theories ol fundamental social revolution (Skocpol, 1979; Hepci, 1991), or of the development of die welfare stale (Flora and Heidenheimcr, 1981), or of tax policy (\Vebbcr ;md VVildavsky, 1987). The particular settings within which these evcnis occur are generally less important than the nature of the events themselves in the development oflliosc theones-
Although less grand than the historical studies of social dcvelop-ineni, the comparative analysis of public policy formulation and implementation would be a clear cast- of applying a process model

across a range of countries. Much public policy analysis in political
science uses a process model, beginning witli agenda-setting and ending with evaluation and feedback iJone.s, 1984; Peters, ]994b). TIlis niodel posits thai the same stages will have to occur in all political systems, so thai comparison can lie made on just how the process unfolds in dilfercnl -settings. This process model is therefore very similar to tlie functionalist tlieorics of politics (Almond and Colcman, i960; Almond and Powell, 1966 - see Key Text 1.1). TIlis approach argued thai there were certain requisite functions that all political systems liad to perform, and comparison therefore was focused on how this took place. Although the heyday of functionalist analysis was during the 1960s, much of its logic remains operative, albeit disguised, in ronlemporary thinking about governing.
Comparative studies ofinslHulions arc somewhat dillercut, in thai they often have a discernible functionalist bias built into them. Legislatures must legislate, ;>nd tliereforc they will tend to perform approximately tlie same tasks, although perhaps in markedly different ways. This presumed commonality of activity may, however, be overstated. Polsby (1975), for example, distinguishes 'transformalive' legislatures from 'arena' legislatures. The former, most notably the Congress of the United Slates, actually make independent judgments about proposed legislation, and frequently impose significant transformations on tlie intentions of llic political executive. Arena legislatures, on tlie other hand, are merely fora within which tlie constitutional formalities necessary for approval of the wishes of the executive take place - the British Parliament is a prime example (but sec Norton, 1993). Legislative bodies of both tliese types of do legislate in some sense of tlie term, but are they really engaged in the same activity? The answer to dial question may depend upon whether tlie researcher is interested in legislatures as institutions, or is more interested in the process of policy-making,
Typology Formation
A tliird way of approaching comparative politics is to develop classification schemes and/or typologies of countries, or perhaps of difFerent components of llic political system. For example, Lijpharl (1968) attempted to clii-s'iil'y democratic political systems according to the degree ot fragmentation ol'tlieir elites and tlie fragmeniaiion of llie mass political culture. This emss-cla.ssilication (l-'igure 1.1) yielded four categories of democratic political systems. Lijphart later

Key Text 1.1 Gabriel Almond and BIngham Powell, and Structural Punctionalism
One of the dominant approaches to comparative politics during the 1960s and into the 1970s was 'struclural-functionalism'. This approach was largely borrowed form anthropology and sociology. The basic premise was that all societies (or polities in our case) had certain core functions thai they had to perform. Polities could then be compared on the basis of how they performed those functions and what structures (institutions) they used to perform them.
The principal virtue of structural-functionalism was that it could be used to compare any political systems from the most primitive to the most advanced. All these political systems had to perform certain functions - things such as 'interest articulation', 'rule-adjudication' and 'political communication' in the Almond and Powell version of the approach. The acceptance and elaboration of this approach was in part a response to the critique of traditional comparative politics that it was culture bound and built entirely upon the experiences of wealthy European and North American societies.
Gabriel Aimond had sketched the structural-functional approach to politics in his earlier book (1960) with James Coleman, but the Almond and Powell volume elaborated the approach. It added another category of comparison - capabilities such as extraction and regulation - that focused on the policy-making role of government. The Almond and Powell version of structural-functionalism also made the developmental aspects of the mode! clearer. In particular, they argued that three basic variables - cultural secularisation, institutional differentiation, and xxxx -could be used to measure the degree of political development.
The critiques of this approach have been numerous. It has been argued that Western cultural and political values were smuggled into a presumably value-free model of political development. It was also argued that by being so general this and similar approaches to comparison said nothing about the way in which politics actually worked in any real country. To the critics the analytic language merely got in the way of meaningful comparison rather than aiding in the process. Finally, the model told the person using it little or nothing about the way in which inputs were converted into outputs. The institutions of government such as legislatures and bureaucracies were merely parts of a 'black box' that would remain largely unopened until the resurgence of institutional analysis. Still, tor an era in which there were massive changes occurring in African and Asian countries this general model did a great service.

FIGURE 1.1 Types of democratic systems
Political elite
Competitive Coatesceni
Hofmigeinoas Cemripetal Depotiticiscd (United Kingdom) (Sweden)
Political cult and society tureFragmented Cenlrifugal (Iialy) Consociationai (Netherlands)

Source: Derived from LiJphart (1968).
(1984) compared democratic systems using the rubric of consensual and majoritarian political systems. Majoritarian systems are those of tlie Westminster type, in which a single party is usually able to gain control of government, and in which politics is conducted in an adversarial manner. Consensus democracies, on the other hand, tend to have coalition governments, in which the style is cooperative and consensual, so thai policies might vary less than in adversarial regimes-Typologies imply the interaction of two or more variables to produce a classification system. For example, Lijphart's typology of democratic systems was more than a simple classification system. It required that each political system be classified on two variables, and that they then be placed in a group based on their scores on those two variables. Thus, a typology is llie beginning of a theory about the subject mailer used to classify the cases. It argues that there are at least two variables that are crucial for understanding the phenomenon being studied, and that it is possible to derive useful measures of those variables for purposes of classification. Further, the typology argues that a second, more useful, order of classification can emerge from inn-grating the first. The degrees of homogeneity of elites and masses may be interesting in themselves, but they become mucli more useful for explanatory and theoretical purposes when they arc combined.
Other forms of classification (taxonomies) arc not so elaborate, but rather may be only simple listings of ihe major types in a class. Taxonomies assume a scheme that uses variables to classify the cases (laxsi), although unlike ly|"ol"gics these do not involve the interaction

of the varinhles. l-'or example, Sartori (1966) argued that political party systems in Western F-uropc fell inio one of three categories: two-party, moderate multi-party, and extreme multi-party. Similarly, Peters (1995al lias argued that the relationships between bureaucracies and iiiiciTsl groups around the world can be seen as falling into four basic classifications: legitimate, clienlela, paraiitela and illegitimate. Berg-Schlosser (1984) provided a classification of African political systems based on a number of political variables. He then uses this (-ht.ssification scheme (Table I.I) to predict the economic and political performance of these systems, finding that more democratic regimes perform rather well in comparison with others. These various classilicaiion systems help to explain (lie behaviour of political systems, or parts of systems, and their capacity to perform their tasks of governing. Perhaps the extreme version ofllie taxonoinic approach to comparison is to create a description of single type of political system in which the scholar is interested and then to use a simple dichotomy for analysis: a case is either 'A'' or 'non-A'1. This method may be seen as a derivative of the methodology of 'ideal types' advocated by Max Weber (Gerth and Mills, 1958: 59-61) and used by him in studying religion, bureaucracy and modernisation. For example, O'Donnell (1978) developed the concept of'bureaucratic authoritarianism' to describe the politics of major Latin American political systems. This concept v^as then used to analyse politics in the various countries in the region, and the dynamics of their development by comparing their actual performance with the conceptual model. This concept did prove useful for describing the Latin American cases, but h;is not been able to travel well outside that region (Bunce, 1995: 88-9).

Regional Statistical Analyses
A fourth variety of comparative analysis encountered in the literature is the statistical study of a selected group of countries, usually the population of countries within a particular region. Tlie purpose ol this approach to analysis is to lest .sonic proposition about politics within that region. Tlie goal is usually to make a generalisation only-about that one region, and, if successful tliere to extend that analysis later to be a proposition about politics more generally. As with the single country study mentioned above, a particular region may be selected as the trial locus for analysis because the traits in question arc manifested most clearly (here,
Take, for example, the huge iiumhcr of .studies that have been undertaken concerning [he welfare slate in Western Europe (for example, Hicks and Swank, 1992; Esping-Anderson, 1990 - sec Key Text tO.l; 1996; Lcibfricd and Pierson, 1995). These studies tend to examine the political and economic roots of the welfare stale in lliese countries, as well as looking at patterns of expansion or contraction of public expenditures across time. Some scholars, for example, Esping-Anderson, may attempt to differentiate groups of countries within the broader European area, although most attempt to derive generalisations about the entire group of countries (see Castles and Milchcll, 1993). Clearly, the welfare state came into full flower in Western Europe and tlie nature of its politics and economics may be seen most clearly tliere.
Could, however, tlie same sets of variables that influenced development of social programmes in Europe also have influenced their development in North America (Kudrle and Marmor, 1981), or in Latin America, which in sonic wavs was also an innovator in tlie welfare state (Rosenbcrg, 1976)? Thus, valuable as these studies may be for the analysis of a single region, their utility could be extended to creating more general theories about subsets of (lie political world, but defined in functional as opposed to geographical terms.
One problem that these regional studies present is that they tend to encourage conceptual stretching (see below, pp. 86-93), once they arc extended beyond the original area of enquiry. Tlie welfare state has a particular conceptual meaning for Europeans, and perhaps for Canadians, but llic interpretation of dial term may be substantially different in the United Stales. This diflerence may not !)<o too much of a problem if we only examine aggregate data - (or example, levels of public expenditure lor social programmes but may become severe if

we attempt to understand comparative survey data on attitudes towards taxing and spending (sec tlie 'beliefs in government' studies, for example, Borre and Scarborough, 1995). Even within Europe, it is not always certain that sonic fundamental political and social terms are interpreted tlie same way, especially between Northern and Southern Europe (Lijpliai-E, 1990). Thus, while continents appear homogeneous on tlie map, and university departments often organise thfir teaching of politics in thai way, there may be sufficient variation in political life to make comparative research difltcult and perhaps even misleading.
Another way to say that this approacli (and the next) encounter conceptual stretching is to say that they trade complexity for an attempt to generalise. The first two approaclics to comparison tended to lie very ricli in complexity, and were able to deal with the nuances of ttie political systems in winch they were working. In tlie third approach, [lie trade-oiT began to be evident- The creation of typologies and ideal types tends to collapse subtle diHerenccs among countries into simple dichotomies. This is an example of Sartori's (1990) 'degreeisrn', in which variables that are inherently continua are collapsed into dichotomies by selecting an arbitrary point along the dimension as a threshold value (sec below, pp. 103 5). As we move into the regional and then global studies, the purpose of analysis becomes statistical hypothesis-testing, and the real and important differences among countries may be manifested only wlicn the data from particular countries appear as 'outliers' in llic distribution of residuals of a regression equation. If llic case is clearly deviant, then the theory almost certainly provides no means of explaining it, and the researcher usually must resort to some variables not included in the theory, often something amorphous, such as "political culture'-
Giobai Statistical Studies
The final option for comparative analysis that we will discuss is the global statistical study. This docs for tlie entire population of countries wlial [he regional studies do for a subset. They measure political and socio-economic variables and then apply statistical tests to tin; relationships among those factors. Some of llic early instances of this style of research remain classics in tlie lieltl. Perhaps llie best example was A Cr^s-f'nfify Survey by Hanks and Texior (t9(i3}. which attempted to classify all tlie (then) countries of (lie worki on a number of political variables, and then created huge i-on-flatioii

matrices linking all those variables. A similar elfort, (tie World Handbook of Podlicai and Social Indicators (Russctt, 1964; Taylor and Hudson, 1972), developed by [he Yalc Political Data Program, did much the same thing but used primarily aggregate, interval-level data as opposed to the nominal data of the A Cross-Polity Survey. Rudolph Rummel (1972; 1979) also engaged in a huge exercise in data collection, the Dimensions of JValions Project, although much of the emphasis was on the international behaviour (for example, war and other acts of aggression) of the countries rather than their domestic politics. More recently, Wallerstein fl980; Hopkins and Wallerstcin, 1980) has taken a more radical approacli, arguing that in essence the A" of comparative politics is one - a single world system. Despite that claim, his work still tends to use nation-states as a unit of analysis, albeit within the context of a dominant world system.
These statistical exercises are about as far removed intellectually from the single country monographic study as they can be, but they still are often described as comparative political studies. The strategy oftliese excursions into data analysis was to accumulate as much data as possible, and then to utilise those data to identify some generic truths about politics across the entire range of countries, usually at a single point in time. For example, the World Handbook did include substantial amounts of time-series data on socio-economic variables, although somewhat less data on political variables. These studies performed some very important functions for comparative research in their lime. For example, they pointed out that it was possible to apply the same ideas and measurements to the developing countries of Africa, Asia and Latin America as could be applied to Western Europe and North America, Pertiaps even more basically, these studies attempted to demonstrate that it was possible to measure empirically some of the important aspects of government that had previously been left largely 10 normative analysis, for example, democracy (sec also Cnudde and Neubauer, 1969; Botlen, 1980;
1990).
This style of research did make its contributions, but it also revealed a number of serious deficiencies. One of these is the measurement problem mentioned above in reference to regional studies (sec below, pp. 93-4). Almost by definition/these global studies involve substantial conceptual stretching. It is difiicult for any empirical measure of a concept such as 'democracy' to travel welt across a range of cultures with a wide range of political and social histories. Even aggregate dilta may IM- somewhat suspect in these

circumstances. For example, a measure such as gross national product may not tap adequately economic activity in any country with a large subsistence economy and a great deal of household production. Even in highly industrialised countries, llicre is a great deal of household production, but if it occurs at relatively equal levels across these countries then (lie measurement is not confounded seriously. When comparing these countries with very poor countries, however, there is marked measurement error. In addition, aggregate statistics can have their own social and cultural elements thai may reduce their comparability across countries (Desrosieres, 1996). A concept such as unemployment, for example, appears to be directly comparable across systems, but may contain strong national biases in measurement (Moore and Richardson, 1989).
Thus, these research eITorls may have produced a number of significant correlations statistically, but it is not always very clear what the substantive importance of these efforts was. Such problems of interpretation are even more true given that they have tended to encourage relatively mindless, atheorctical exercises in correlating all variables with all other variables, in the knowledge thai there probably would almost certainly be some significant correlations-Rummel's use of factor analysis tended to carry this approach to higher levels of statistics, but not always a higher level ofsubstance-
Summary
From the above discussion, it may appear that comparative polilics is a hopeless enterprise. Each of the approaches considered above appears filled with Haws and insurmountable intellectual objections. On the one liand, that apparent hopelessness is very real, but yet it is no reason not to carry on doing comparative politics. Any approach or method we may select for conducting our research will have some real and important problems. However, these weaknesses do not mean tliat we should not choose it. What they do mean is that we need to be cognizant of the strengths and weaknesses of each candidate method and then to be wary of the results our analysis may give us. If we are aware of llie inherent Haws of each method, we can compensate for those problems in interpretation, if not dirccily in the analysis.
The dilliculiies of each of these methods also points to the possibility and desirability of 'triangulation', or ttie use of multiple methods and approaches (Webb^rt/., 1967; Roth, 1987) for research. l-'or example, it'a country is a deviant case on a statistical distribution,

then the obvious comparative question is, Why arc they? A researcher may then want to engage in a more detailed analysis ol'llic one case, and in so doing not only illuminate llial one case, but also illuminate and expand the theory being tested through statistical analysis. Unfortunately, relatively few researchers bother lo extend their analysis in this direction. Most statistical researchers simply dismiss the outliers as deviant cases, so long as tlie overall results are statistically significant. At most, llic outliers may receive a sentence or two- Likewise, the case-study expert on one country or a limited number of countries may never know whether Ins or licr case is an outlier or not, because of no! thinking in those statistical terms.
The Content of Comparisons
Up to this point, we have been looking at the various methodological approaches to comparative analysis. We should also point out that, despite its ccntrality in the field, cross-national analysis is not the only way in which to perform comparative political analysis. The logic of a comparative analysis is almost always exactly the same, regardless of what the geographical or temporal focus of the comparison may be. Each of the dilfercrit possible foci and loci of analysis will, however, impose some costs and provide some benefits for tlic research exercise. The researcher must determine what advantages, in terms of testing or developing theory, can lie gained by using one or another of llie various sites and styles of comparison. He or slic must also determine what are tlie costs, in terms of possible threats to validity, tliat are associated with cacti stvic of reseairli.
Although we usually ihink primarily of cross-national comparisons - the whole system bias described by Lijpliarl (1975b: 166) comparative analysis across subnational units within a single nation-state is also a fruitful form of political analysis. Indeed, if we are interested in employing cornparalivc analysis for (lie purposes of testing political theory then in many ways thi;' lorm of analysis lias even greater potential than lias tlie cross-national one. The principal virtue of analysis across subnational governments is that it holds constant, or minimises, a number of factors that might otherwise confound it. l-'or example, V. 0. Key's classic analysis /1949) of politics in tlie Southern states of tlie United Slates points out how holding a number of factors constant enables tlie researcher lo understand belter llie -subtle differences in these polhical systems, even at a more dcscriplive level.

The comparative state policy literature (Dye, 1966; Sharkansky, 1968; Brace, 1993) further demonstrates how this locus of analysis enables researchers to uncover important relationships among variables that might not be possible at the national level. Tins style of comparative policy researcli lias been emulated in other countries with botli federal fFricd, 1976) and non-federal systems (Alt, 1971).
It would l)e easy to assume that the cultural and social factors that may confound comparative analysis can be eliminated in an analysis of units witliin a single nation-stale. That is not necessarily true. For example, the stales within the United States share many cultural and socio-economic features, but also have distinctive political histories and cultures (liat may influence policy choices and their internal politics (Elazar, 1987; Kincaid, 1990). Further, even though there may be relatively limited variance in some factors, in contrast to a comparison between tlie United States and, for example, Upper Volta, there may still be significant variance. That is, in a way, an opportunity, however, given that tlic greater commonality among the slates may provide an opportunity for more precise and sensitive/ forms of measurement than would be possible in cross-national comparisons.
Cross-Time Comparisons
In addition to comparisons across geographical units, comparison can also be made wilhin the same political unit across lime. Even more than with the case of subnational units witliin a single country, this strategy holds (relatively) constant the cultural and social factors that may confound analysis of political relationships. Certainly, social and cultural patterns within a political system may change gradually across time, but time also can be included as one component of the analysis. This contextual change can be controlled for statistically (Aclien and SIiively, 1995), and also can lie controlled for in qualitative research through a thorough understanding of the case. Most forms of comparative analysis focus on llic characteristics describing political units and tlic relationships llial exist among variables measured williin those units. Tlie cross-lime method, however, looks at rc[ationsliips among variables within individual units across lime.
To [lie extent that cross-system comparisons are made with time-series analysis, those comparison;, would be made across tlic p.itlcrns

thill are manifested within individual countries across lime. If statistical analysis of developmental patterns is done within countries, the outcomes of that analysis, such as [lie degree of fit with the statistical model, or the slope of the regression line (as a measure of efTect) could be compared. Diagnostics might also be used to determine if there liad been significant changes in the relationships of variables across lime. Less quantitative analysis might examine the apparent impact of different factors - independent variables - at different stages of the development process. This mode of analysis can establish some less formal sense of changing relationships among variables across time.
Analysis across time remains, for the most part, underutilised in political science. This is often a function of inadequate time-series data for oilier tlian a few variables, Ibr example, voting or public expenditure. While governments have collected a great deal of economic data across time, they have been less helpful to tlie political science researcher, and methods such as survey research tend to have been poor at preserving time-series on all but a few key variables. It is important to note, however, thai less statistical time-scries analysis is making a major impact, in the hands of the 'historical mslitutional-ists' (Thelen e( ai., 1992; Pierson, 1996) and their considerations of the enduring impact of policy and institutional clioices.
Tliere arc several ways to combine the cross-national and llie cross-temporal modes of analysis. In particular, the pooled time-series approach to statistical analysis (Stimson, 1985; Alvarez, Garrett and Lange, 1991) includes data from a number ofcountries across a signilicant period of time- This method provides several benefits for comparative analysis, one of the most important ofllicm being simply expanding the number of observations, thereby improving the capacity of researchers to make reliable statistical estimations. In addition, pooled time-series permits the analyst to lest directly whether the principal source of variance is the country or some other factor. It may be that relationships of variables within countries are more significant in explaining outcomes than tlie differences across countries. The same factors that create the benefits for comparative analysis also create some statistical problems (see Key Text 6.1)
but tins still offers a means of expanding the reach of comparative analysis.
Finally, we come to the question of whether all comparisons involving lime need to use exactly the same slices of chronological time. The most appropriate comparisons for theoretical development

may be made with events and structures occurring in different time periods, albeit periods dial are similar in their basic characteristics. For example, the coiiieniporacy problems of modernisation and democralisation in llic countries of Eastern and Central Europe (Agh, 1994; Roskin, 1991; Rose, 1992} arc functionally equivalent to the problems of Western Europe in llie late nineteenth century, or to immediately post-colonial Africa and Asia. Tlie researcher will have to establish tlie relevant criteria for comparability, and then attempt to extract the theoretical and substantive meaning from the comparison.
Summary
To he effective in developing theory, and in being able to make statements about structures larger than an individual-or the small group, the social sciences must be comparative. We have been aware of this truism for some lime, and a great deal of good comparative analysis has been available in all the social sciences, but especially in political science. Much of that comparative analysis lias however been done, using more intuition about the logic of comparison than any self-conscious effort to compare efTcctively and efficiently. This volume is intended to identify tlie questions that a comparativist sliould at least think about before embarking on data collection and analysis, whether the data to be used are quantitative or qualitative. It also will discuss some of the techniques (in a non-technical manner) available to the researcher and what they can, and cannot, add to the study of comparative politics.
Although there is some agreement that comparative politics is essential to the development of tlie discipline, there is less agreement over what constitutes acceptable versions of this approach. Definitions of'good research' are largely a (unction of who is being asked, and in wliat context. The traditional view of comparative politics emphasised the description of one or a limited number of political systems, and a place certainly remains in the discipline for that such research. More contemporary researcli is often more directly comparative, involving a number of countries, and uses a variety of quantitative leflmiqiics in place of the more detailed description common in traditional analysis. Botli of these -styles have their place, and increasingly the discipline is attempting to reconcile tlie research traditions and (lie research melhocl-s ofbolh. Slill, there are adherents

of each who will argue for exclusivity rather than a more catholic
approacti.
As we assess the quality of comparative research, we must remember tliat 'quality* is to some degree a function of the goals being pursued by [lie researcher. One purpose of comparative analysis is to verily propositions, and to demonstrate lhal certain relationships among variables hold true in a wide variety of settings; the intention here is to demonstrate similarity and consistency- On the other hand, good comparative research, and indeed most comparative research, is concerned with differences and demonstrating that what occurs in one setting most certainly docs not occur in another. This latter strategy of differentiation may appear to be only an old-fashioned description in new clothing. It can lie as 'scientific' as the strategy of finding similarities, but merely looks for different types of propositions and different types of variables.
While there will be a number of questions raised, the answer to most of them is, 'It depends'. There is no perfect way to conduct comparative analysis, only strategies that are belter or worse under given circumstances. Furthermore, almost any choice of a research strategy involves sacrificing some virtues in order to achieve more of otiicrs. What is perhaps most important, therefore, is to understand fully what values arc being traded for what others in choosing any analytic approach, The researcher can then know where tlie analysis is likely to be weak, and where the results have to be regarded with even greater scepticism than is truly necessary for all social science research. Scepticism is almost always justified, but careful consideration of the probable impacts of the tools used for analysis helps to make the scepticism useful.
The one tiling that should be universal in studying comparative politics, whether in one country or in many, and whether using statistical or qualitative methods, is a conscious attention to explanation and research design. What is the thing the researcher wants to explain - what is tlie 'dependent variable'? When (lie problem is clear, the question then becomes, What is the presumed 'cause' of tlie phenomenon in question? It is better for the development of the discipline of political science if thai presumed cause is related to some broader theoretical concerns, but there may be interesting research in which that initial connection is tenuous at be.sl. Filially, what evidence is needed to prove' the connection belwcen cause and cncct, and how can lliat evidence l>c mustered? How can we be as sure as possible alioul llie quality of our evidence.

• [ На главную ]
• [ Библиотека ]
• [ Reader ]
• [ Программы ]
• [ Абитуриенту ]
• [ Фотоальбом ]
• [ О ФП ]
• [ Гостевая книга ]
• [ Ссылки ]
• [ Обратная связь ]