Cluster-Bloc Analysis and Statistical Inference*

Publisher: Cambridge University Press

E-ISSN: 1537-5943|66|2|569-582

ISSN: 0003-0554

Source: American Political Science Review, Vol.66, Iss.2, 1972-06, pp. : 569-582

Disclaimer: Any content in publications that violate the sovereignty, the constitution or regulations of the PRC is not accepted or approved by CNPIEC.

Previous Menu Next

Abstract

Cluster-bloc analysis is a useful method of examining the voting records of a legislature, in order to find what subgroups of members regularly vote together. Agreement scores are calculated for every legislator with every other legislator. Then when a group is found to have all its members in high agreement with each other they are referred to as a cluster bloc. These groups, which are discovered empirically, are not necessarily the same as the formal caucus groups. So far each researcher has had to use his own judgment as to what constitutes “high agreement,” but it can be shown that the cutoff points can be established statistically, against the null hypothesis of random voting. Since each score can be tested for significance, it is possible to use statistically based indices of cohesion for the legislature or any specified subgroup and indices of adhesion between the various subgroups. Examples are given for the African group in the UN General Assembly.