Essays on econometric analysis of strike duration data

The dissertation consists of three essays on econometric analysis of strike duration data. Early empirical studies on strike lengths focused primarily on descriptive analysis. Following the developments in bargaining theory and economics of strikes, several researchers analyzed the effects of cyclical conditions, size of the firm, issues of strike and other pertinent variables on strike lengths. These studies were conducted mainly in a univariate parametric framework. Often the empirical results have been found to be contradictory. While the choice of data set may be an important contributory factor, other possible explanations include misspecification of the distribution function and application of a univariate framework when the true underlying process is multivariate. In this thesis, we examine strike lengths in Canada during the period 1981-2002 by relaxing some of these assumptions.;The second essay proposes a bivariate duration model of strike lengths based on the war-of-attrition framework. Separate hazard rates are estimated for workers and management. Both are found to exhibit negative duration dependence although workers appear relatively more conciliatory at the outset of a strike. The results are able to reconcile many of the inconsistencies of previous empirical work. In particular, workers and management are found to react asymmetrically to certain covariates including local unemployment conditions and the issue for the strike. Strike lengths are found to be countercyclical.;The third essay extends the structural model to a continuous time framework. We show that it is possible to develop a simple continuous time version of the standard binary choice model which maintains its connection to economic behaviours. We find that the effects of various regressors are qualitatively similar to our findings in the second essay. The study also shows that the duration dependence parameters are sensitive to the choice of unobserved heterogeneity distribution. We find the sensitivity to be significantly less in the structural models.;In the first essay, we consider semiparametric estimation of discrete duration models with idiosyncratic errors with unknown distribution. The information matrix is derived. In the case of separable duration dependence, the information matrix is singular. In that situation the information matrix for the regression component of the hazard function is derived. We develop N --consistent estimators for those instances when the information matrix is nonsingular. Less-than-- N --consistent estimators of the duration dependence component of the hazard rate of the separable model are developed. Simulations and an application to strike-length data illustrate the viability of the approach.