Janda and Gillies: "How well does 'region' explain political party characteristics?" --> back to first page
Notes
  1. The data were collected under NSF grants GS-1418, GS-2533, and GS-27081 for the International Comparative Political Parties Project at Northwestern University, with Kenneth Janda as Principal Investigator. The data have been deposited with the Inter-University Consortium for Political and Social Research as ICPSR Study 7534 (Janda, 1979). The variables and coding procedures are discussed thoroughly in Janda (1980).
  2. This usage of analysis of variance differs from Cox's description of the aim of regionalization as the allocation of places to regions 'in such a way as to minimize the within-region variance! between-region variance ratio within the constraint of the number of regions required' (1969: 70). Although Cox also employs the analysis of variance model, his approach seeks to delineate regions. This study evaluates the utility of regions once regions have been delineated on other grounds.
  3. The parties were selected for study if they met minimum criteria for strength and stability. For legal parties, we required that they win at least 5 per cent of the seats in two elections from 1950 to 1962. For illegal parties, we required evidence of support from at least 10 per cent of the population over five years. The selection criteria are discussed in Janda (1980: 5-7).
  4. The eta-squared statistic is also known as the correlation ratio. It is the ratio of the explained sum of squares to the total sum of squares in an analysis of variance. It expresses the proportion of variation in the dependent variable that is due to the groupings on the independent variable.
  5. According to the theory of regional explanations of party politics, region constitutes the independent variable and party characteristics the dependent variable. In the conventional use of discriminant analysis, however, the nominal variable (region) constitutes the dependent variable. In truth, discriminant analysis is blind to causal ordering, and the authors can conceive of the analysis in the reverse direction.
  6. Difficulties have been noted in using the 'three world' classification scheme. Roth and Wilson suggest that this scheme is 'neither neat nor analytically precise' (1976: 5). The problem lies in part in the fact that classification into the First and Second Worlds is based primarily on a country's political system and its dominant ideology, while Third World countries are defined by the extent of their social, political, and economic development. However, Horowitz (1969: 39-46) discusses the three worlds in terms of four factors: economy, polity, society, and military. The First World is 'dominated by the United States, including allies in Western Europe and satellites in Latin-America and elsewhere'. The key traits of First World countries are an industrialized, capitalist economy, parliamentary democracy, a highly urbanized society, and a professionalized military that executes the orders of the political elites. The Second World is 'dominated by the Soviet Union, including allies andlor satellites in Eastern Europe and parts of Asia'. Second World nations are characterized by an industrialized, socialist economy, democratic centralism, high urbanization, and a professionalized military that works with the political elites. Third World nations are 'non-aligned and non-satellite nations with a general tendency toward clustering in Africa, Asia and Latin-America--a spectrum conventionally covering Algeria to Yugoslavia in economy and India to China in polity'. In spite of the variation, Horowitz suggests that Third World nations are characterized by low development, a mixed economy tending toward socialism, mass democracy, an urbanizing society, and a politically active military.
  7. The discriminant analysis rontine in SPSS was used for this analysis. The canonical functions were not rotated. Unfortunately, the SPSS program does not calculate the total structure coefficients reported in this paper. They were computed by correlating the variables with a composite score created from the unstandardized discriminant function coefficients.


Janda and Gillies: "How well does 'region' explain political party characteristics?" --> back to first page