Strike Data with a Crisis Point

The paper is concerned with the analysis of strike data in which the distribution of short strikes differs from that of long strikes. It appears through visual inspection and asymptotic procedures that for Israeli strikes in the years 1965-1992, the hazard function is exponential for strikes lasting less than 40 days and that it is Weibull with a Weibull parameter greater than unity for longer strikes. The economic interpretation of the phenomenon is discussed. As there is typically only a small sample of long strikes available (Kiefer 1988), classical asymptotic tests are unlikely to convey the correct message. We suggest a test statistic for the hypothesis that a break does not occur. A new F-based expansion for the small sample distribution of the test is derived. The test rejects the hypothesis of no break for the data under investigation. While conventional approximations to the distribution of the test statistic are shown to break down catastrophically, the F-expansion appears to be highly accurate.