Firms that responsively match supply to demand may pay a steep price for internally generated disruptions to productivity. In the case of responsive manufacturers, their performance depends on variable workloads of tasks being completed by coordinated groups of workers—and thus may suffer when workers leave. Using novel data tracking the staffing and productivity of 52,214 assembly line workers in a China-based factory assembling millions of devices weekly, we present empirical evidence that worker turnover conservatively causes USD 146-178M in greater yield losses alone, primarily by severing knowledge-sharing and relationships between assembly line co-workers. To study these effects both empirically and prescriptively, we extend the classical production planning problem to include endogenous worker turnover as an Experience-Based Equilibrium. Advances in reinforcement learning and approximate dynamic programming are used to estimate and simulate our model. While the firm determines lines’ workrates in response to their productivities, we address this endogeneity using the firm’s rolling demand forecasts as instrumental variables (demand shifters). Our main findings demonstrate considerable value from a less turnover-prone workforce. When wages are low, the firm institutes a buffered workforce and overtime hours as means of hedging against productivity disruptions. Consequently, a counterfactual wage increase not only smoothens production but also makes the optimal workforce leaner. Mainly by reducing underage, higher wages decrease the manufacturer’s variable production costs (including wages) by up to 21%, or USD 594M for the product we study.
Recommended citation: Moon, K. et al. (2020). "Manufacturing Productivity with Worker Turnover" Working paper.