Methods
Data
Our source of data on nurses is the merged outgoing rotation groups of the 2003 through 2006 Current Population Surveys (CPS), which is collected by the United States Bureau of Labor Statistics. Union membership data were first collected on a regular monthly basis beginning with the 1983 CPS. From the CPS, we extracted all RNs (Census Occupation Code = 3130) working in hospitals (Census Industrial Code = 8190) whose incomes were directly reported (and not imputed). Although the CPS is in principle a longitudinal dataset, with each participant appearing twice in the outgoing rotation group, we ignore the repeated observations, both in terms of identification (Hirsch & Macpherson 2000) and in terms of clustered errors. The CPS sample used for this analysis, after applying restrictions described later, had 831 RNs who were unionised and 3403 non-union nurses. Because we pool nurses from multiple years of the CPS, there is not a straightforward comparison of the sample size to the national population of RNs, but the unionisation rate and the share of RNs employed in hospitals is consistent with other data sources, such as the 2004 National Sample Survey of Registered Nurses (Health Resources and Services Administration 2006). The hourly wage for RNs is computed as usual weekly earnings divided by usual weekly hours. We use the CPI-URS, a time-consistent measure of the US consumer price level, to convert all values to real 2006 dollars. Educational attainment was based on credentials received by nurses; however, we cannot observe whether the completed degree
was in the field of nursing. We code nurses with no degree beyond secondary as Diploma nurses and other nurses by the highest degree attained. We used potential experience – i.e. age minus six minus years of education – as a proxy for years of work experience. We applied the following additional restrictions to our sample. In addition to requiring that all necessary variables be available for all observations, we restricted the sample to nurses who usually worked at least 20 hours per week, and whose real hourly earnings fell between the Federal minimum wage and $100 per hour (in 2006 dollars).