Article ID: | iaor20012415 |
Country: | Netherlands |
Volume: | 97 |
Issue: | 1 |
Start Page Number: | 15 |
End Page Number: | 29 |
Publication Date: | Dec 2000 |
Journal: | Annals of Operations Research |
Authors: | Owsiski Jan W., Zadrony Sawomir |
Keywords: | statistics: multivariate |
The paper presents a simple exercise of application of cluster analysis to the set of voting data for the members of Polish Parliament of the previous term, which started in 1993. Each MP is characterized by the vector containing the specification of MP's behaviour during individual votings, distinguishing the following categories: ‘For’, ‘Against’, ‘Abstained’, ‘Not voted’, and ‘Absent’. The distances between the ‘MP descriptions’ containing the thus defined categories of behaviour during votings (forming the matrix of dimensions: number of MPs × number of votings) constitute the basis for the analytic procedure based upon clustering. We are looking for the clusters of similar MP descriptions which are, simultaneously, possibly different between clusters. The analysis performed does not account for the membership of the MPs in the political groupings (parties and alliances) within the parliament, and one of the main goals of the analysis is just to identify the degree of agreement between the results of the cluster analysis and the membership in such groupings. This is done after the procedure of clustering has been performed. Another aspect being of interest in the analysis is dependence upon (sensitivity to) the parametric definitions of distances with respect to the categories adopted in the description of voting (e.g., ‘is absence the same as not voting?’). The results of analysis are meant to show to what extent the actual behaviour of the MPs reflects their political membership and what are the actual relations (in terms of true behaviour) between political groupings represented in the parliament. The objective of the altogether modest study was two-fold: (1) to assess the capacity of the (definite algorithms of) cluster analysis in identification of the group structure of attitudes expressed through voting behaviour, and, given that such a capacity exists, (2) to compare the structure obtained with the formal membership in the political organizations, and also, though only marginally, with the structure of the political scene assumed or perceived in the mass media or through the political pronouncements.