Exploratory Cluster Analysis of Attitude Structure in the New Zealand Electorate
Abstract
In replicating a national survey dealing with political attitudes in the New Zealand population (2) the authors had to consider the problems associated with classifying their own newly obtained sample of 451 individuals in terms of the responses elicited by an interview which included 35 attitude items and nine biographical measures. The attitude items were designed to measure responses to the political situation and New Zealand society in general, while the biographical measures were sex, age, marital starus, time spent in employment, number of children, educational qualifications, occupation, source of income, and political affiliations. The three responses allowed for each attitude item were (a) agree, (b) 'uncertain, and (c) disagree. Data were analysed using automatic interaction detection analysis (3) after an alternative clustering method (requiring considerable computational expense) did not uncover any structure in the dam (1). The 35 attitude items were used as the independent variables, with political affiliation being the marker (dependent) variable on which the splitting process was based. Separate analyses were run on rwo nonoverlapping subsets of the data (250 cases and 201 cases). Three clusters were identified as common to both analyses. People in cluster one were generally happy and satisfied with NZ politicians. People in cluster two were dissatisfied with NZ politicians but found politics boring. Those in cluster three were unhappy wich NZ politicians but found politics interesting. A stepwise discriminant analysis was then used, with the 26 son-political attitude responses as the independent variables, to identify non-political attitude differences between the three clusters. The remaining attitude which distinguished between the clusters was the belief that New Zealanders care about one another. Only 23% of cluster one disagreed with the statement that New Zealanders care about one another (the rest either agreed or were uncertain) while 53% and 45% of clusters two and three, respectively, disagreed with the statement. Political affiliation indicated that cluster one corresponded roughly to the right end of the political continuum while cluster three was at the lefr end. Members of cluster rwo divided their support over four political parties. A detailed description of the survey results are available from the authors.
Authors: Mark Chignell, Barrie G. Stacey
Published in: Psychological Reports (1980)