III. HUMAN RESOURCES AND INTERNATIONAL SECTIONS
REFEREED PAPERS

Estimating Returns to Managers From Employee Unionization

TODD FISTER
University of Illinois at Urbana-Champaign

Abstract

      This paper finds that unionization appears to reduce manager pay, but seems to have no effect on manager employment. The paper relies on fixed-effects and instrumental variable techniques, the latter using company age as an instrument for unionization. The results show that OLS regressions may be biased by unions self-selecting into firms with few managers and high manager pay. Likewise, the fixed effects technique suggests that managers in unionized establishments earn less relative to the frontline employees, although this result appears to be biased by unions selecting into firms that pay managers more relative to the other employees.

 

      This paper analyzes how unionization affects manager-level pay and employment, using a fixed effects and instrumental variable research design. There is a large literature on unionization and covered worker wages (for recent examples, see Batt 2001; Budd and Na 2000; and Hirsch and Schumacher 2001.), firm investments (Bronas and Deere 1993, 1994; Cavanaugh 1998; Connolly et al. 1986; Hirsch 1991, 1992), and financial performance (Abowd 1989; Becker 1995; Becker and Olson 1992; Ruback and Zimmerman 1984). Little work has examined how unionization influences manager-level outcomes. If unionization reduces manager pay, then managers will try to prevent union representation and may use company resources to protect their current pay levels. In one of the few papers on the subject, the authors find that unionization at the industry level is associated with fewer managers and lower manager wages, leading to the conclusion that unionization reduces the need for manager monitoring and that unions shift firm rents from managers to workers (DiNardo, Hallock, and Pischke 2000).

      The empirical difficulty in determining returns from unionization is that unobserved variables, especially worker ability and firm rents, are likely correlated with both union coverage and wages. To control for unobserved heterogeneity specific to the firm, this study first uses a fixed-effects model to identify the manager pay premium within the firm. The second analysis relies on an instrumental variable (IV) method to identify the relationships between unionization and manager pay and employment. The IV method uses firm age as an instrument.

Specification

Manager pay is assumed to vary with unionization (U_i) and a set of control variables (X_i):

      For ordinary least squares to estimate an unbiased coefficient on unionization (), the error term must be independent of unionization. Theoretically, we would expect unions to organize in firms that pay high production-worker wages and employ high-ability production workers. To the extent manager pay and manager ability are correlated with the same for production workers, unionization will be correlated with the manager-pay error term, and this correlation biases the coefficient. The error term likely contains both establishment-specific components (geographical location, firm rents) and manager-establishment components (unobserved manager human capital, the managerial labor market), so the error term in equation 1 can be decomposed into an establishment-specific error term () and a manager-establishment error term ():

      The establishment-specific error term can be eliminated by comparing manager wages to the wages of other employees in the firm, such as production workers. By definition, managers and production workers in the same firm have identical establishment-specific error terms. This is a type of fixed effects model that relies on multiple pay levels per firm instead of the more typical use of multiple time periods per firm. To eliminate the establishment-specific error term, I first need to specify a pay equation for another level of employee, such as production workers:

      Taking the difference of equations 2 and 3 leads to the following equation,

      The establishment-specific error terms cancel out, so the remaining variables estimate how unionization and other firm characteristics affect managers and production workers differently, plus the random errors. It is possible to estimate equation 4 and obtain unbiased results if the error terms are not correlated with unionization. Again, that assumption is too restrictive. For example, firms that pay managers significantly more than production workers may cause workers to perceive compensation inequities and, therefore, organize a union.

      An instrument variable approach first estimates predicted unionization and then uses predicted unionization to identify from equation 5. The variables in the first stage include all the covariates from the base specification plus the instrumental variables, Z_1i:

      By assumption, the instrumental variables Z_i is independent from the error term in the manager pay equation, cov(Z_1i , -) = 0. Using predicted unionization in the second stage results in a union variable that is independent of the error term, so its expected value is now the expected value of the true relationship.

      I propose using capital equipment age as an instrument for unionization. An effective instrument is correlated with unionization but uncorrelated with the manager pay premium over production workers within the same firm. Capital equipment age is a straightforward instrument. Assume that a certain percentage of firms are unionized each year (with probability p) and that unions tend to persist once organized. The decision for workers to unionize will not be perfectly independent for firms (each year, every firm faces unionization with probability p), nor will it be perfectly dependent (a single firm faces its own probability p_f each year). Older firms, then, are more likely to have been unionized in past periods, and that unionization has likely persisted to the present period. However, firm age has no impact on manager pay, because compensation will be determined by broader labor market forces and individual human capital.

Data

      In 1996, the Center for Educational Effectiveness at the University of Pennsylvania and the U.S. Census Bureau conducted a representative survey of American establishments. These data were intended to improve knowledge about training and education within firms but also contained responses about establishment and worker characteristics. An establishment is any nonheadquarters business in a single location. For example, establishments include doctors' offices, law firms, single-employee service firms, restaurants, retail stores, warehouses, factories, and transportation companies. Establishments exclude corporate headquarters, nonprofit operations, and government and military offices. A chain of five separate restaurants would count as five establishments, although they might have a common legal owner and shared management.

      The Census Bureau surveyors used established contacts to gather information through a series of questions in a telephone interview. The response rate was 77 percent for a total sample size of 3,081 establishments. Respondents could chose not to answer certain questions, so casewise deletion for missing data reduces the sample size to 1,361 establishments. There are no major differences in variable means between the total and reduced samples.

      The variables for this analysis are log manager pay, log production worker pay, unionization, log book value of capital, industry, multiestablishment status, and capital age. The key dependent variable is the difference between log annual manager pay and log annual production worker pay. I converted the pay measures into natural logs to fit the convention of the compensation literature and to reduce skewness in the data. The key independent variable is unionization, measured as the percentage of nonmanagement and nonsupervisory employees covered by a union contract. Industry is measured as a dummy variable for 20 broad industry categories (such as "health services" and "food/tobacco"), and multiestablishment status is a dummy variable for whether the establishment is part of a multiestablishment company. Capital age is measured as the percentage of fixed capital purchased 10 or more years ago.

Analysis and Discussion

      The fixed effects regression, shown on the 1st column on Table 1, estimates that a 100 percentage point increase in unionization leads to a 11 percent lower manager-pay premium over production workers, controlling for capital, industry, and multiestablishment status. This analysis does not control for the nonrandom distribution of unions into firms. To eliminate biases from unobserved heterogeneity, the instrumental variable regression uses variation in unionization caused by variation in firm age to identify the causal effect of unionization on manager pay. This regression, shown in the 2nd column of Table 1, estimates than a 100 percentage point increase in unionization leads to a 92 percent lower manager-pay premium, more than eight times as large as the OLS coefficient. I performed a Hausman test to determine if the differences in coefficients are statistically significant. The null hypothesis that the coefficients are not different can be rejected at a .001 significance level.

      The second set of analyses examines whether unionization reduces the portion of employees who are managers using an OLS and IV analysis. Shown in the 3rd column of Table 1, the OLS analysis finds that there are fewer managers in unionized firms, which is consistent with prior industry-level empirical analyses. Using firm age as an instrumental variable for unionization, I find that there are more managers in unionized firms, but this is not significant at conventional levels. There is no evidence that unionized firms employ fewer managers.

      These results suggest that unions select into firms that pay managers more relative to production workers and that unionization itself has an effect on relative managerial outcomes. The first result is not surprising. If workers join unions because of distributive equity concerns (such as "unfair" compensation across levels), then unions will be more likely in firms with high manager pay relative to production worker pay. Likewise, managers in high-manager-pay firms may have been appropriating rents from firm owners, and unions would recognize that they could negotiate those rents to workers. The second result is consistent with two models: input substitution and agency theory. If unions increase worker wages to higher-than-market levels, then efficiency wage theory predicts that workers will shirk less. This shirking effect means that the firm requires less manager monitoring and, therefore, can hire lower-ability managers for less pay. A second possibility is that managers appropriate rents from owners before unionization but must share those rents with employees after unionization. A third possibility is that managers appropriate rents from prior investments but that owners invest less after unionization occurs. Manager pay would fall because the total firm rents fall.

      One criticism is that this negative pay effect is driven by unionization increasing production worker pay while manager pay remains flat. If this interpretation is correct, then it has its own interesting implications. The first is that pay for any single level within a firm does not depend on pay at other levels. Managers do not receive a pay raise simply because employees lower in the organization receive a pay raise. This seems counter to social-psychological concepts like status, fairness, and equity, and economics concepts like internal labor markets. The second is that manager quality does not seem to rise with production worker quality. If wages for workers rise significantly, workers will queue up to receive above-market union wages, and the firm will select the most able employees. This will result in higher worker quality. If manager quality and worker quality are complementary inputs to the production process, then manager quality should increase with worker quality, and manager pay should rise. These empirical results suggest that manager pay, in fact, does not rise with unionization.

      If unionization does cause lower manager pay due to substitution or agency controls, then there is an incentive for managers to prevent unionization, even if unionization does not affect company profitability. This could help explain the consistent negative reaction that managers have towards employee unionization. Alternatively, the second interpretation--that managers do not benefit, while workers benefit greatly--could cause status, fairness, and equity problems within the firm, leading to managers who seek to prevent unionization for nonfinancial reasons.

References

Abowd, John M. 1989. "The Effect of Wage Bargains on the Stock Market Value of the Firm." American Economic Review, Vol. 79 (September), pp. 774-800.

Baldwin, Corliss Y. 1983. "Productivity and Labor Unions: An Application of the Theory of Self-Enforcing Contracts." Journal of Business, Vol. 56 (April), pp. 155-85.

Batt, Rosemary. 2001. "Explaining Wage Inequality in Telecommunications Services: Customer Segmentation, Human Resource Practices, and Union Decline." Industrial & Labor Relations Review, Vol. 54, no. 2 (March), pp. 425-49.

Becker, Brian E. 1995. "Union Rents As a Source of Takeover Gains Among Target Shareholders." Industrial & Labor Relations Review, Vol. 49, no. 1 (October), pp. 3-19.

Becker, Brian E., and Craig Olson. 1992. "Unions and Firm Profits." Industrial Relations, Vol. 31, no. 3, pp. 395-415.

Bronas, Stephen G., and Donald R. Deere. 1991. "The Threat of Unionization, the Use of Debt, and the Preservation of Shareholder Wealth." Quarterly Journal of Economics, Vol. 106 (February), pp. 231-53.

Bronas, Stephen G., and Donald R. Deere. 1993. "Unionization, Incomplete Contracting, and Capital Investment." Journal of Business, Vol. 66 (January), pp. 117-32.

Bronas, Stephen G., and Donald R. Deere. 1994. "The Effects of Unions on Firm Behavior." Industrial Relations, Vol. 33 (October), pp. 426-51.

Budd, John W., and In-Gang Na. 2000. "The Union Membership Wage Premium for Employees Covered by Collective Bargaining Agreements." Journal of Labor Economics, Vol. 18, no. 4 (October), pp. 783-807.

Cavanaugh, Joseph K. 1998. "Asset-Specific Investment and Unionized Labor." Industrial Relations, Vol. 37, no.1, pp. 35-50.

Clark, Kim. 1984. "Unionization and Firm Performance: The Impact on Profits, Growth, and Productivity." American Economic Review, Vol. 74 (December), pp. 893-919.

Conolly, Robert A., Barry T. Hirsch, and Mark Hirschey. 1986. "Union Rent Seeking, Intangible Capital, and the Market Value of the Firm." Review of Economics and Statistics, Vol. 68 (November), pp. 567-77.

DiNardo, John, Kevin F. Hallock, and Jorn-Steffen Pischke. 2000. "Unions and the Labor Market for Managers." Unpublished working paper, February.

Hirsch, Barry T. 1991. Labor Unions and the Economic Performance of Firms. Kalamazoo, MI: W.E. Upjohn Institute for Employment Research.

Hirsch, Barry T., and Edward J. Schumacher. 2001. "Private Sector Union Density and the Wage Premium: Past, Present and Future." Journal of Labor Research, Vol. 22, no. 3 (Summer), pp. 487-518.

Ruback, Richard S., and Martin B. Zimmerman. 1984. "Unionization and Profitability: Evidence from the Capital Markets." Journal of Political Economy, Vol. 92 (December), pp. 1134-57.