Measuring Union and Nonunion Wage Growth: Puzzles in Search of Solutions

This paper presents conflicting evidence on trends in private sector union and nonunion wages. The BLS quarterly Employment Cost Index (ECI), constructed from establishment surveys, uses fixed weights applied to wage changes among matched job quotes. The ECI shows a substantial decrease in wage growth for union relative to nonunion workers. The annual Employer Costs for Employee Compensation (ECEC), drawn from the same survey data as the ECI, provides wage level estimates constructed from the full sample of job quotes using current sector weights. Surprisingly, the ECEC shows no trend in relative union-nonunion wages. Household evidence from the Current Population Survey (CPS) can potentially reconcile the conflicting ECI/ECEC evidence, but it is first necessary to account for Census earnings imputation procedures that bias the level and trend in the CPS union gap. CPS wage data absent controls, which should provide results similar to that from the ECEC, instead displays a steep decline in relative union wages, similar to that seen in the ECI. Regression estimates using the CPS, similar in spirit to the ECI, indicate a union premium that declines modestly over time. Union and nonunion wage growth is next calculated from the CPS using methods roughly consistent with the ECI and ECEC. These results shed rather dim light on the sources of ECI/ECEC differences. The contribution of the paper is its unearthing of puzzling and apparently contradictory evidence on union and nonunion wage trends (as well as on economy-wide wage growth). Although our analysis helps account for specific elements of this puzzle, a comprehensive solution remains elusive. We conclude that there has been closing in the unionnonunion wage gap since the mid-1980s, but the magnitude of the closing is anything but clear.