Knowledge that Transforms

We present a bargaining model of union contract negotiations, in which the union decides between two threats: the union can strike, or it can continue to work under the expired contract. The model makes predictions about the level of dispute activity and the form disputes take. Strike incidence increases as the strike threat becomes more attractive, because of low unemployment or a real wage drop. We test these predictions by estimating logistic models of dispute incidence and dispute composition for negotiations from 1970 to 1989. We find support for the model's key predictions, but these associations are weaker after 1981.

The variance of arbitrated wage settlements is significantly lower than the variance of negotiated wage settlements in a sample of teachers' contracts, even when possible selection of contracts into arbitration is accounted for. This result is consistent with the hypothesis that arbitrated settlements do not incorporate all of the information that is reflected in negotiated settlements.

This article considers a model in which the productivity of a worker depends on his experience, ability, and position in the firm. It is shown that workers are sorted in positions based on comparative advantage. Furthermore, workers are induced to retire when productivity is equal to the value of time after adjusting for the reallocation of other workers by positions that would occur if the worker retired.

Statistical analysis of the job-search model has isolated the role played by the minimum observed wage in identifying and estimating behavioral parameters. Estimators based on order statistics, however, are influenced by measurement error in ways that estimators based on averages are not, and there is ample evidence that wage data contain measurement errors. We propose a model in which worker behavior is captured by the prototypal search model but wages are observed with error.

Following permanent layoffs most women search for new jobs, but some withdraw from the labor force. We develop a joint model of the choice to undertake postdisplacement job search and unemployment durations for searchers and estimate it, using data from the 1988 Displaced Worker Survey. Maximum likelihood estimates of this "split-population" model show that labor force withdrawal is an important factor explaining the distribution of postdisplacement jobless spells. The model also allows us to distinguish the effect of any covariate on the decision to engage in postdisplacement search from its effect on search duration. Single-population models obscure this distinction.

Estimates of the intertemporal labor supply behavior of males in Canada using micro data are reported. Individuals make the intertemporal labor supply decision on the basis of annual hours and weeks. Precision of the parameter estimates is improved by using tenure variables as instruments for the wage. Further, the age and tenure variables are allowed to have taste parameters in the structural equations. The evidence suggests that this is required only for the two age variables. Elasticity evidence suggests that evolutionary changes in the wage cause changes in the number of weeks with the elasticity being 0.6 and statistically significant.

We study wage bargaining in which the union is uncertain about the firm's willingness to pay and threat payoffs vary over time. Strike payoffs change as replacement workers are hired, as strikers find temporary jobs, and as inventories or strike funds run out. We find that bargaining outcomes are substantially altered if threat payoffs vary. If dispute costs increase in the long run, then dispute durations are longer, settlement rates are lower, and wages decline more slowly during the short run (and may even increase). The settlement wage is largely determined from the long-run threat, rather than the short-run threat.