Friday, January 21, 2011

2010 Union Data

BLS The Bureau of Labor Statistics has just released its 2010 figures for union coverage.   Overall union density went down from 12.3% to 11.9%; in the private sector, union density went down from 7.2% to 6.9%, and in the public sector, it went from 37.4% to 36.2%.  Among the other highlights:

   --The union membership rate for public sector workers (36.2  percent) was
     substantially higher than the rate for private sector workers (6.9 percent).

   --Workers in education, training, and library occupations had the highest
     unionization rate at 37.1 percent.

   --Black workers were more likely to be union members than were white, Asian,
     or Hispanic workers.

   --Among states, New York had the highest union membership rate (24.2 percent)
     and North Carolina had the lowest rate (3.2 percent).

For more  detailed compilations of the 2010 union membership figures, as well as historical data, you can see


Government Reports | Permalink

TrackBack URL for this entry:

Listed below are links to weblogs that reference 2010 Union Data:


Jeff's brief post omits the most significant numbers: First, the decline in union density from 12.3% to 11.9% overall; 7.2% to 6.9% in the private sector; and, most surprisingly, from 37.4% to 36.2% in the public sector.

Second, the 612,000 decline in absolute numbers of union members (about 4%). The drop in numbers occurred in both sectors, 339,000 in the private sector and 273,000 in the public sector. Puzzlingly, the decline in union membership far exceeds the loss of jobs in the economy last year, so clearly something other than the bad economy is involved.

The BLS is an expert agency, but I've always found its union membership figures pretty mushy. BLS reported a substantial gain in union membership (over 700,000) in 2007 and 2008, but apart from some SEIU gains in the public sector, there were no major unionization campaigns that could account for the supposed increase. National-level yearly jumps and plunges like that just haven't happened. Data for individual states are also suspect. The BLS report shows an increase in 2010 of 5,000 union members in my small state of South Carolina, for example, but there was no significant union organizing that year and the few unionized businesses did not expand.

Finally, the public sector estimates seem to rely on a strange definition of a union. The BLS report is not clear, but it may use the same classification that Jeff's dad, Barry, does at That seems to count as a "union" organizations that do not engage in collective bargaining. For example, there is NO state or local public sector collective bargaining in SC (or in many other southern states), yet claims that 13.1% of the public sector employees belong to unions and 16.3% are covered by contracts. (Unionized federal employees could only account for a small fraction of the claimed public sector union members.) The only explanation is that the stats count membership in the SC Education Association and perhaps the SC State Employees Association, neither of which bargain with anyone. That distorts the unionization numbers because the only thing that distinguishes a union from a simple association of employees is the use of collective bargaining.

So take those statistics with a big grain of salt. They are useful for trends --- and the trends in density and membership have been steadily down since the mid-1950s --- but not for precise calculations of smaller groups.

Posted by: Dennis Nolan | Jan 22, 2011 6:32:14 AM

Thanks Dennis, leaving off the overall numbers was just a late night oversight.

Posted by: Jeff Hirsch | Jan 22, 2011 1:29:28 PM

Dennis Nolan’s comments on union membership and coverage estimates raise several issues concerning their reliability. I share his concern about estimates at the state level for public employees, but not his concern about the national figures.

To address his comments one must first describe the data source and survey questions on which union estimates are based. Union data from BLS and all figures shown at (produced jointly by Dave Macpherson and me) are compiled from the January through December Current Population Surveys (CPS), the monthly household survey administered by Census to about 60,000 households. The CPS is the source of our monthly unemployment figures, plus a host of demographic and labor market information. Union membership and coverage figures are based on two questions (unchanged since 1977) asked to employed wage and salary workers. First, each worker is asked if on your primary job you are “a member of a labor union or of an employee association similar to a union?” Second, those who are not members are asked, “On this job, are you covered by a union or employee association contract?” Both BLS and measure union members as those who answer “yes” to the first question and measure “coverage” by including the sum of members plus covered non-members. Thus, the coverage measure can include members not covered by a collective bargaining agreement (say, teachers in South Carolina who join the NEA). It is widely recognized that these measures do a far better job measuring private than public sector unionization.

Professor Nolan’s concern about state level estimates for public sector workers is appropriate, not only because of shortcomings in the survey questions, but because of small sample sizes and reporting error. Sample sizes on which state union estimates are based are often small; hence year-to-year changes generally reflect more noise (sampling and reporting error) than they do signal. This is why BLS does not provide disaggregated union estimates such as those compiled at UnionStats. And this is why UnionStats reports the sample sizes on which all estimates are based, warning users that estimates based on small samples should be used with caution. (For example, the estimate of 2010 public unionization in SC is based on 378 surveyed workers who “blow up” into a population estimate of 316 thousand.) Combining the imperfect definitions of membership and coverage for public sector workers, the relatively small sample sizes on which estimates are based, and reporting error by households, Dennis is correct to attach limited weight to public sector estimates by state. One should be even more reluctant to give weight to year-to-year changes in estimates. Trends should be based on, say, three year averages (to increase sample size) compared over long time periods (thus increasing the ratio of signal to noise).

As for the national figures, sample sizes are sufficiently large to produce minimal sampling error. Dennis finds puzzling that union membership estimates in 2010 show a larger decline than total employment. Two things are going on here. First, declines in public employment produced a disproportionate decline in total membership since the public sector is so highly unionized compared to the private sector (the public sector accounted for 24% of the employment decline, but 45% of membership decline). More generally, what are observed each year as increases or decreases in employment (or membership) reflect the relatively tiny differences between immensely larger flows of job loss and job creation. Even when the economy is strong and there is strong growth in net employment, extremely large numbers of jobs are being destroyed (Schumpeter’s creative destruction). Thus, union membership can expand absent major new organizing if already organized companies are growing. More common has been declining union density during periods of economic expansion – since most newly created jobs are born nonunion, unions have not been able to organize at a rate that will maintain density (i.e., replace lost union jobs plus organize a sufficient share of newly created jobs).

Posted by: Barry Hirsch | Jan 22, 2011 3:26:27 PM

Post a comment