University of Ulster
This paper introduces for the first time in Northern Ireland an electoral ecology approach. The returns of the 1991 Census of Population and the 1993 local election results are merged and examined in order to explore the demographic and socioeconomic characteristics of voting behaviour.
The first section of the paper deals with the traditional approaches such as the studies on the mechanics of the electoral system and the analysis of elections as the investigation of political and social attitudes. These have been the main thrusts of electoral research to date in Northern Ireland.
The statistical association of electoral and socio-economic data has been called 'electoral ecology' since it takes area/units, electoral constituencies (wards, etc), as the unit of analysis rather than the individual voters.
The division of Northern Ireland into 101 District Electoral Areas (DEA) facilitated the introduction of this approach since it made possible the use of the necessary multivariate analysis as a valid electoral and statistical exercise. The DEAs therefore are the unit of analysis.
The new approach is placed in the context of the two principal models of voting behaviour which have developed in recent years. The first is based on party identification, the life-long attachment by individuals to a party. The second is based on a rational choice approach where voters weigh up all the issues and approaches before coming to a decision. The former model has always been seen as depicting well the Northern Ireland situation.
The paper concentrates on linking the demographic indicators from the census with key electoral factors such as turnout and party preference. The analysis is set against the important demographic shifts which have taken place in the last decade. These include the overall rise in the proportion of the population that is Catholic and in particular, the rise of the Catholic middle class.
The paper deals at length with the way in which the census data and ward data was linked. This was of central importance since the wards (the building blocks for the census analysis) had been reorganised between the taking of the census and the voting in the elections.
Prior to the results being presented the paper discusses the 1993 local election campaign and the importance of both the battle between the unionist and nationalist parties and within the respective camps.
A core section of the paper concentrates on voting preferences as expressed at the DEA level. Factor analysis was the key technique adapted with factor scares subsequently utilised in multiple regressions as a guide to the type of variables which could best predict potential support for the five main parties (UUP, DUP, Alliance, SDLP and Sinn Fein).
Using principal components analysis three factors were extracted which explained 82.7% of the total variance. The factors were rotated to maximise the loadings using orthogonal rotation.
The factor analysis pinpointed variables which would be most useful in building a more focused linear model for support for each of the main political parties. The highest predictive power was achieved in respect of Sinn Féin where 77% of the variance in their vote was explained by the model.
The second major focus of interest in the analysis in the paper is the way in which social and economic factors influence turnout rates at local government elections in Northern Ireland.
Fourteen demographic variables formed the basis of a multiple regression equation using turnout as the dependent variable. A model with a high goodness of fit R2 = .78 was developed.
The paper highlights that the three key influences on turnout are the percentage Protestant/ Catholic population; the number of self employed individuals in the District Electoral Area and the geographical region. A dychotemous dummy variable coding for each DEA as East/West of the Bann was introduced here.
The paper concludes that whilst religion is a dominant factor in party support, the nature of intra unionist and nationalist competition is changing. The traditional social and economic profile of both unionist and nationalist voters is changing within an unstable political environment.
The research provides the benchmark to measure on a longitudinal basis, the electoral response (using the census data in tandem with the 1997 local government electoral data) to the unfolding political landscape.