A structural labor demand model is developed that allows for worker heterogeneity regarding firing costs and productivity. It is estimated for a firm in demise when 3,650 workers were made redundant in a restructuring following bankruptcy. This was the largest mass layoff in the history of the Netherlands. The model produces sharp predictions on how firing thresholds depend on individual worker characteristics. The signs of the estimated coefficients are consistent with these predictions. The model correctly predicts 68% of individual worker displacement. The results provide new in-depth knowledge on how firing costs influence personnel decisions.

The costs of hiring a worker generally differ in size from the firing costs. This article investigates optimal labour demand schedules for production and non-production workers of firms that operate under uncertainty and face asymmetric costs of adjusting their workforce. In the empirical part generalised methods of moments (GMM) estimates of the structural parameters of the Euler conditions for production and non-production workers are presented, together with specification and structural stability tests, using time series data of the Netherlands and U.K. manufacturing sectors. We find that asymmetric adjustment costs play an important role in the explanation of unbalanced labour demand between upward and downward movements of the business cycle. Moreover, hiring costs exceed firing costs of production workers, whereas firing costs exceed hiring costs of non-production workers.

Firms may adjust capital and labor sequentially or simultaneously. In this paper, we develop a structural model of interrelated factor demand subject to nonconvex adjustment costs and estimated by simulated method of moments. Based on Norwegian manufacturing industry plant-level data, parameter estimates reveal cost advantages for adjusting capital and making net changes in labor simultaneously. Factor demand models with fully specified interrelated adjustment costs structures perform best to describe the dynamic panel data.