Manufacturing Productivity with Worker Turnover

86 Pages Posted: 25 Sep 2018

See all articles by Ken Moon

Ken Moon

University of Pennsylvania - The Wharton School

Patrick Bergemann

University of Chicago Booth School of Business

Daniel Brown

UC Berkeley

Andrew Chen

Apple Inc.

James Chu

Stanford University, Department of Sociology, Students; Stanford University - Freeman Spogli Institute for International Studies

Ellen Eisen

UC Berkeley

Gregory Fischer

London School of Economics & Political Science (LSE)

Prashant Kumar Loyalka

Stanford University - Freeman Spogli Institute for International Studies

Sungmin Rho

Graduate Institute of International and Development Studies (IHEID)

Joshua Cohen

Apple University

Date Written: September 11, 2018

Abstract

Once endemic in early manufacturing, high worker turnover has re-emerged as a challenge for modern manufacturers. A long-standing literature suggests or presumes that standardizing practices and deskilling assembly tasks shield manufacturing productivity against turnover. We investigate this persistent common wisdom using a uniquely rich dataset drawn from China-based FATP (final assembly, testing, and packaging) facilities that produce millions of units of consumer electronic goods per week yet exhibit high worker turnover exceeding 300% annually. Combining data on weekly rolling-horizon production plans, 52,214 workers' compensations and assignments, and production lines' volumes and yields, we find that worker turnover significantly impacts productivity, conservatively incurring $103-146 million in material costs alone over the production life cycle of a single device model absent active mitigation by the firm. Turnover disrupts critical workflows and relationships, which are often neglected even as firms track individual employee performance with increasingly granularity. To study managerial prescriptions, we extend the classical production planning problem to include endogenous worker turnover as an Experience-Based Equilibrium. Advances in approximate dynamic programming and reinforcement learning are applied to estimate and simulate our model. We find that well-calibrated increases to worker compensation reduce the manufacturer's labor-inclusive, variable production costs by about 5%, or $135 million. The wage increase improves workforce and yield stability and reduces underage costs, which are incurred from lost sales when missing production targets.

Keywords: Data-driven workforce planning, Empirical operations management, Employee turnover, Experience-Based Equilibrium, Production planning, Productivity, Stochastic optimization, Structural estimation

Suggested Citation

Moon, Ken and Bergemann, Patrick and Brown, Daniel and Chen, Andrew and Chu, James and Eisen, Ellen and Fischer, Gregory and Loyalka, Prashant and Rho, Sungmin and Cohen, Joshua, Manufacturing Productivity with Worker Turnover (September 11, 2018). Available at SSRN: https://ssrn.com/abstract=3248075 or http://dx.doi.org/10.2139/ssrn.3248075

Ken Moon (Contact Author)

University of Pennsylvania - The Wharton School ( email )

Jon M. Huntsman Hall
3730 Walnut St.
Philadelphia, PA 19104-6365
United States

Patrick Bergemann

University of Chicago Booth School of Business

Daniel Brown

UC Berkeley

Andrew Chen

Apple Inc.

James Chu

Stanford University, Department of Sociology, Students ( email )

Stanford
United States

Stanford University - Freeman Spogli Institute for International Studies ( email )

Stanford, CA 94305
United States

Ellen Eisen

UC Berkeley

Gregory Fischer

London School of Economics & Political Science (LSE)

Houghton Street
London, WC2A 2AE
United Kingdom

Prashant Loyalka

Stanford University - Freeman Spogli Institute for International Studies ( email )

Stanford, CA 94305
United States

Sungmin Rho

Graduate Institute of International and Development Studies (IHEID) ( email )

PO Box 136
Geneva, CH-1211
Switzerland

Joshua Cohen

Apple University ( email )

Register to save articles to
your library

Register

Paper statistics

Downloads
85
rank
277,913
Abstract Views
387
PlumX Metrics