A Theory of Parties and Electoral Systems

3. An Extensive Test

An Extensive Test ONCE a theory has been developed deductively, it must be tested empirically. In doing this, one may pursue either, or both, of two complementary strategies. On the one hand, the theory may be tested intensively. The greater the number of hypotheses derived from the theory that can be confirmed,the greater will be the confidencethat can be placed in the theory, In particular, as the intermediate steps by which a final conclusion is reached are themselves confirmed, confidence is increased in the explanatory power of the theory, as wellas inits predictivecapacity. On the other hand, a theory may also be tested extensively. The nature and costs of intensive analysis necessarily limit consideration to a smallnumber of cases. While the fit of a variety of data to the predictions of a theory is important in buildingconfidence, so too is the fit of data from a variety of cases taken from the theory's presumed domain. Although fewer, and more general, hypotheses can be tested extensively, the overall validity of the ultimate predictions of a general theory is as important as its detailed fit to a few cases in arguingforits utility. The three chapters following this one are devoted to an intensive test of the theory of electoral systems and party structurejust presented, based on data from three European countries. The present chapter reports an extensivetest of the theory's generalpredictions based on data from a random sample of fourteen Western democratic nations. In constructing a theory one deliberately oversimplifies and abstracts from reality. Simple and clear-cut theoretical concepts replace far more complex realities, Many variables are assumed to be constant when in fact they are not; others are assumed to be irrelevant although in reality they should not be ignored completely. An intensiveanalysis allows one tomake the qualifications and modifications necessary to have the theory fit a particular case* Extensive analysis does not. The fact that the same operational indicator of a theoretical construct may not be the most appropriate for all cases is translated into measurement error. While nonconstant "constants" and ignored variables may average out, in doing so they contribute to unexplained variance. Thus in extensively testing a 35 3 36 A Theory of Parties and Electoral Systems theory, predictions of fact become hypotheses about tendencies, likelihoods, and proportions. Because ofthis, and the limitednumberofcases, the actual hypotheses to be tested differ slightly from those listed at the end of chapter 2, The propositions considered in this chapter are: 1. Parties competing in PR systems will be more likely to be ideological in style than parties competing in plurality systems. 2, Parties competing in smalldistricts will be less issue oriented than parties competing in large districts. 3. Parties competing in systems with ordinal choice will be less issue oriented than patties competing in categoric choice systems. 4, Parties competing in systems with an intraparty choice allowed to voters will be more likely to be factionalizedor fractionalized than parties in systems without this kind of choice. Since these hypotheses involve a number of quite elusive concepts, like ideology, many of which can be measured only indirectly, the observed relationships occasionally may be rather weak, They should, of course, all be in the hypothesized direction. It is important that one be clear as to what the theory does and does not seek to explain. The theory tries to explain whether or not parties will be ideological in their approaches, not what ideology or combination of specific policy positions any particular party will adopt. It tries to explain whether or not parties will be issueoriented, not whichparticular issues will be important at any given time. It tries to explainwhetheror not parties will be cohesive, and if not whetherthey will be factionalized or fractionalized, not why specific politicians adhere to one or another particular faction. The indicators of party characteristics employed in this chapter are derived from the data file ofthe International Comparative Political Parties (ICPP) project. The ICPP data are based on approximately a 50 percent stratified random sample of party systems in existence between 1950 and 1962.1 The fourteen countries considered here are the modern Western democracies that fell into this sample. Altogether, data concerning fortynine parties are included, all those winningat least 5 percent of the seats in the lower houses of their national legislatures in at least two successive elections, from 1950 to 1962, Separate coding of variables was done for each of two subperiods, 1950 to 1956 and 1957 to 1962.2 One of...