Conventional coalition theories have applied only to formal or executive coalitions and have tended to be very one-dimensional in their approach. They do not do justice to the strongly informal element in Italian coalitional behaviour, where a multi-dimensional framework of analysis is required. Because of Italy's system as a partitocrazia the key variables are inevitably party-political in nature. In this context, there is a strong case for looking at local coalitional behaviour and its relationship with national politics. In particular, it may be examined whether national patterns have simply been replicated locally. or alternatively whether local government has allowed experimentation in new alliance formations as a possible precursor for national politics with implications for policy change. Despite the strong control that national party élites have over local politics, Italy has experienced left administrations sub-nationally: however, these have made no dramatic impact because of the institutional and political constraints on local government and also because Italy's worsening economy has further reduced policy options at this level. This analysis therefore supports the thesis of that country's ‘ungovernability’, and the answer to the ‘Italian crisis’ must clearly come from Rome and not the provinces. For general comparative purposes, there is nevertheless a strong case for considering sub-national politics in coalition studies not least because this level of activity invariably enters the strategic considerations of national party leaders.

The article criticizes stratification measures of social class and class voting, demonstrating that they may produce biased images of the relationship between social class and party choice. Firstly, it is argued that the conventional ‘Alford index of class voting’ is typically not comparable, neither over time nor across countries. Applying a materialist concept of social class, the author then concludes that class voting in Denmark has in fact increased, although Denmark is generally listed as the ‘prototypical’ evidence of the opposite. This difference in perception is due to significant changes in class behaviour which are concealed by stratification measures of social class.

Pre-adult socialization is one way in which political ideologies are formed. Parents can affect the ideologies of their offspring through direct transmission or through the indirect mechanism of family social milieu. This paper analyses parent–adolescent pairs drawn from a larger sample survey study of eight western nations carried out in the mid-1970s. Three varieties of ideological perspectives are addressed: the traditional left–right continuum, pro– and anti-establishment attitudes, and material–postmaterial value structures. In general the familiar and historically rich orientations of the left–right dimension are transmitted much more successfully than are the remaining two, which are newer and less encapsulating. Regardless of particular ideology, the direct transmission model is significantly more faithful to the findings than an indirect model based on the family's social class and religious composition. The distortion in the reproduction of ideologies between the generations indicates the importance of intra–individual development and macro–level forces as sources of political ideologies.

By juxtaposing observed voting behaviour and voting expected due to partisan affiliations, normal vote analysis attempts to classify elections at the aggregate level and to estimate the impact of short-term variables on the vote. Apart from graphical illustrations, this latter task is performed by the effect coefficients proposed by Boyd. This contribution demonstrates that these coefficients under certain circumstances can exaggerate the effect of short-term influences on the vote, proposes a modified formula, and shows how normal vote analysis can be adapted to comprise the notion of statistically controlling for third short-term forces. Finally, an attempt is made formally to relate the effect coefficients of normal vote analysis to the decomposition of variance in multiple regression analysis. It is demonstrated that such a formal relationship cannot be established, so that in terms of variance decomposition normal vote analysis is incapable of precisely and numerically comparing short-term vs. long-term influences on the vote, or of comparing short-term effects across different variables or even across different samples.

Debate about the interpretation of the Irish party system has been brought to a head by the publication of two studies that devote considerable attention to the issue and that differ radically in their conclusions (Carty, 1981; Garvin, 1981). The question is of interest beyond the ranks of specialists in Irish politics because the overriding issue has been the extent to which the Irish party system conforms or fails to conform to comparative models. Both the systems approach (Blondel, 1969; Sartori, 1966; 1976) and the social cleavage approach (Lipset and Rokkan, 1967) have figured in this debate, the latter tending to occupy centre stage. Indeed the Carty-Garvin divergence relates precisely to the applicability of the Lipset and Rokkan framework.