A Multivariate Logit Analysis of the Outcomes of Arbitration Decisions in British Columbia
In: Evaluation review: a journal of applied social research, Band 14, Heft 3, S. 247-263
ISSN: 1552-3926
A major conclusion drawn from this study is that bivariate estimates of the probability of management or union wins in grievance arbitration cases can be seriously in error as a result of the failure of these estimates to adjust for the effects of variables not included in the bivariate comparison. In this analysis of 1,815 arbitration cases occurring in British Columbia in the 1982 to 1985 period, bias is identified in the bivariate estimates of the probability of management wins by industry and issue. In addition, from application of a multivariate logit model, the study estimates the effect on arbitration outcome of industry, union size, union arbitration proneness, issue, arbitration experience, union occupation, and year. Similarities to and differences from the findings of previous studies are identified.