It is well-established that unionized workers in the United States are covered by more extensive employee benefits than are comparable nonunion workers.1 Data from the March 2002 Current Population Survey (CPS), for example, show that unionized workers are 16.4 percentage points more likely than similar nonunion workers to be covered by an employer-provided health insurance plan, and 18.8 percentage points more likely to participate in an employer-sponsored retirement plan.2 What is less clear, however, is why this is so. In their seminal book What Do Unions Do?, Richard Freeman and James Medoff argue that greater benefits for unionized workers stem from two factors: 1) union bargaining power (what economists call the "monopoly face" or "monopoly effect" because they liken union bargaining power to that of a monopolist), and 2) union voice (or what is sometimes called the "collective voice" face).3
An important and longstanding question is, What is the relative importance of these two explanations? In particular, note that the welfare implications of each explanation are quite different. The monopoly face distorts competitive outcomes and reduces aggregate welfare, and monopolies of any kind are viewed as inefficient (and therefore less desirable) in standard economic thought. The collective voice face, however, can overcome market imperfections and increase aggregate welfare relative to what would result from individual, self-interested behavior. Thus, to evaluate accurately the aggregate welfare effects of labor unions on employee benefits, one must separate the positive and negative effects in empirical studies.
The key to isolating these two effects is to note that the monopoly effect increases total compensation while the collective voice effect tends to rearrange the total compensation package rather than to increase it.4 In statistical terms, then, holding total compensation constant in empirical analyses will separate out the monopoly and collective voice effects. Freeman and Medoff use the BLS Employer Expenditures for Employee Compensation data from the 1970s5 and find that the union effect on employee benefits is roughly equally split between a monopoly and a collective voice effect. Widely available data sets like the public-use samples of the Current Population Survey, however, lack good measures of total compensation. Consequently, researchers have only rarely tested the continued currency of these early results.
Recent research conducted by this author, which will be published later this year in the Journal of Labor Research, uses the BLS Employer Costs for Employee Compensation (ECEC) data from the March 2004 National Compensation Survey (NCS) to investigate whether the earlier results are still accurate 30 years later. The forthcoming analysis is based on 33,776 private-sector jobs in 7,863 establishments. The results suggest that the two-face framework of Freeman and Medoff continues to be relevant. More specifically, the analysis has three main findings: First, jobs that are represented by a union have total expenditures on nonmandatory benefit items that are 25- to 50-percent higher than similar nonunion jobs. Second, the union effect on benefits is particularly large for lower paid establishments and for small establishments. And third, the union effect on employee benefits consists of both a monopoly and a collective voice effect, though there is wide variation--depending on the specification, the collective voice effect might be as low as 25 percent or as high as 75 percent. But there always appears to be a nontrivial mix of both the monopoly and collective voice effects.
These preliminary results underscore the importance of empirical analyses of significant employment-related questions and provide an example of the diverse applications of the NCS and other BLS data sources.
1 See, for example, Thomas C. Buchmueller, John DiNardo, and Robert G. Valletta, "Union Effects on Health Insurance Provision and Coverage in the United States," Industrial and Labor Relations Review, July 2002, pp. 610-27; John W. Budd, "Non-Wage Forms of Compensation," Journal of Labor Research, Fall 2004, pp. 597-622; Richard B. Freeman, "The Effect of Unionism on Fringe Benefits," Industrial and Labor Relations Review, July 1981, pp. 489-509; Richard B. Freeman, "Longitudinal Analyses of the Effects of Trade Unionism," Journal of Labor Economics, January 1984, pp. 1-26; and Richard B. Freeman and James L. Medoff, What Do Unions Do? (New York, Basic Books, 1984).
2 Budd, "Non-Wage Forms of Compensation." The differences shown here are generated from probit models controlling for gender, marital status, ethnic background, education, potential labor market experience, part-time status, hourly paid status, employer size, public sector employee, industry, occupation, and region. Basic statistics can also be accessed from the Census Bureau's Federal Electronic Research and Review Extraction Tool (FERRET) at http://dataferrett.census.gov/. Statistics obtained using FERRET will differ slightly from the figures cited here because the cited figures are from probit models which control for differences in union-nonunion worker characteristics.
3 Freeman and Medoff, What Do Unions Do? An additional possibility is that unions promote awareness of existing benefits; see John W. Budd and Karen Mumford, "Trade Unions and Family-Friendly Policies in Britain," Industrial and Labor Relations Review, January 2004, pp. 204-22.
4 The monopoly-voice model can be applied to various aspects of unionism; see Freeman and Medoff, What Do Unions Do? For example, theoretically, the collective voice effect can lead to increased productivity (greater output at every level of labor input), which might, in turn, increase total compensation. It is nevertheless common to distinguish between the monopoly and collective voice effects on compensation by examining the extent to which the voice effect rearranges rather than increases the compensation package, for two reasons: 1) the power to capture increased productivity in the form of higher wages stems from monopoly power, and 2) when analyzing different compensation packages between union and nonunion workplaces at one particular point in time, any effects of unions on productivity have already occurred.
5 The BLS Employer Expenditures for Employee Compensation data were separate and distinct from the subsequent Employer Costs for Employee Compensation (ECEC) data. For a discussion of the differences between these two sources of data, see Felicia Nathan, "Analyzing EmployersECosts for Wages, Salaries, and Benefits," Monthly Labor Review, October 1987; on the Internet at http://www.bls.gov/opub/mlr/1987/10/art1full.pdf.