The Surprising Impacts of Unionization on Establishments: Accounting for Selection in Close Union Representation Elections∗

Regression discontinuity studies of the effects of union representation rely on quasi-random variation in vote tallies near the majority-rule threshold. It is precisely in the closest elections, however, that incentives may be most powerful for manipulation. This study shows evidence for such contaminating selection in close union elections. First, unions lost close elections at an unexpectedly high rate, even those involving a large number of voters. Second, prior to the election, establishments where the union lost by a narrow margin employed fewer workers, had a smaller payroll, and had a more dispersed distribution of workers earnings than establishments where the union barely won. As a result, level comparisons between establishments where the union barely won and barely lost are misleading. Regression discontinuity panel estimates which account for pre-election differences suggest that union representation had a negative effect on payroll, average worker earnings, and employment, in marked contrast to previous findings. This effect operated largely through compositional changes: higher-paid workers were more likely to leave, while younger and lower-paid workers were more likely to come following a union victory. The union effect on the earnings of workers who stayed was minimal. JEL Codes: J01, J51 ∗This research uses data from the Census Bureau’s Longitudinal Employer Household Dynamics Program, which was partially supported by the following National Science Foundation Grants: SES-9978093, SES-0339191 and ITR0427889; National Institute on Aging Grant AG018854; and grants from the Alfred P. Sloan Foundation. Any opinions and conclusions expressed herein are those of the author and do not necessarily represent the views of the U.S. Census Bureau. All results have been reviewed to ensure that no confidential information is disclosed. The author is indebted to JP Ferguson, Jim Davis, Javi Miranda, and Kristin McCue for help with the data. The author also gratefully acknowledges support from NSF Grant SES-0922355. Word count (text and appendix): 7,973. †181 Faculty Office Building, Provo, UT 84602, Fax: (801) 422-0194, Tel: (801) 422-4049, E-mail: [email protected].