Cyclical Variation in Individual Conditional Strike-Settlement Probabilities

Empirical work on strikes has tended primarily to concern itself with the strike incidence, with the recurring finding that strikes occur procyclically. Less attendion has been paid to strike duration and the cyclical fluctuations it might display, although a variety of theoretical models have been proposed suggesting that a strike duration will vary with the business cycle. This paper presents an empirical investigation into this issue, using Canadian microdata to estimate flexible models of conditional strike-settlement probabilities. Two contrasting hazard models are estimated and compared. Both utilise a flexible polynomial representation for the duration dependence and explicitly incorporate unobserved heterogeneity. The first is a betalogit model recently proposed by Kennan (1984), and the second is a discrete analogue of the familar proportional-hazards model with a Gamma mixing distribution. Score tests for a number of potential specification errors are also constructed for both models. The models are estimated on a sample of individual contract strikes in manufacturing over wage issues during the period 1971 - 1980. The results strongly support the hypothesis that strike duration varies countercyclically, with conditional settlement probabilities at the peak of the cycle being up to twice as high at the trough. A supplementary finding is that cyclical sensitivity is even more marked in the paper and printing industries, and that individual conditional settlement probabilities there are much lower, ceteris paribus, than elsewhere in manufacturing. The hypothesis of homogeniety across industries and unions in the remainder of manufacturing is accepted by the data