Comment R O N A L D G. E H R E N B E R G
Cornell University Bernard Lentz has presented an interesting paper in which he seeks to integrate the economic analysis of trade unions and the theory of investment in human capital. Lentz views an individual's decision to join a union as being based upon a comparison of the expected present values o f his or her net economic benefits from union and nonunion employment. Lentz's emphasis on viewing the decision to join unions in a life-cycle context is a welcome one and it can be favorably contrasted to several recent sophisticated econometric papers on the topic which ignore life-cycle considerations. ~As I will discuss below, however, I believe that his underlying theoretical structure must be generalized in a number o f directions, including the consideration o f nonpecuniary aspects o f employment, and his empirical work needs to be strengthened. At its current stage, the empirical evidence in support of his approach is at best suggestive. At the onset of my comment though, I want to stress one important issue which Lentz neglects. He notes that his model implies that the decision to join a union is a "productive investment" from the perspective of the individual, however the model yields no such implication about the "social productivity" o f the decision to join a union. Clearly, this is an important issue. Given the current controversy over labor law reform legislation we should be interested in knowing whether, on balance, unions have led to an increase in the overall level of output in the economy. While proponents of unions may argue that the benefits derived from unions are primarily redistributive, in the form o f improved earnings and conditions of work for their members and in the f o r m of the social programs and protective labor legislation that unions lobby for, evidence that unions have a net positive "social marginal p r o d u c t " would strengthen the case for labor law reform. Up until recently, most analytical labor economists tended to view unions, on balance, as having a negative influence on the level o f productivity. 2 This view derived, in part, from the belief that unions limit management flexibility and from popular evidence of restrictive work rules, including manning requirements, that a number of craft unions historically demanded, and won, in industries such as railroads, construction, printing and shipping. 3 Recently, however, analytical labor economists have begun to rediscover the possibility that unions may also have positive influences on productivity. Discussions o f this possibility and empirical evidence to support its existence are found in papers by Richard 'See, for example, John Abowd and Henry Farber (1977), and Henry Farber and Dan Saks (1978). 2Seethe excellentdiscussion on this topic by Everett Kassalow in his conference paper. 3TheSeptember 1978newspaper strike in New York City was based almost solelyon the controversy over one such requirement. JOURNAL OF LABOR RESEARCH Volume 1, Number 2 Fall, 1980
RONALD G. EHRENBERG
409
Freeman, James Medoff, Charles Brown, and their associates at H a r v a r d . ' Their analyses are based heavily upon unions' roles as institutions o f collective voice operating within structured internal labor markets and they conclude that unions can increase productivity in a number o f ways. 5 First, by providing workers with a direct means to voice their discontent to management and by establishing job rights based upon seniority, unions may reduce worker discontent and voluntary turnover (i.e., quitting). Reductions in job turnover increase firms' incentives to provide firm-specific training for their employees and will lead to increased productivity. 6 Moreover, seniority systems weaken the extent of rivalry between inexperienced and experienced employees and consequently increase the amount o f informal on-the-job training which the latter are willing to give to the former. 7 Second, by increasing the economic rewards to employment, and providing a grievance mechanism, unions may increase worker morale, motivation, and effort. Finally, unions provide an explicit mechanism by which labor can point out possible work rule changes or changes in techniques of production that will benefit both labor and management. Clearly, the net impact of unions on productivity must be ascertained empirically, as these potential positive productivity affects o f unions may be offset by the aforementioned negative influences. In a recent paper, Charles Brown and James Medoff provide strong evidence that productivity in the U.S. manufacturing sector was some 20 to 30 percent higher in unionized firms in 1972 than in nonunion firms, even after controlling for the capital intensity of production, the quality of labor, and the two-digit industries in which the firms were located. ~ As with all empirical studies, their results are not totally unambiguous. Moreover, they unfortunately are restricted to manufacturing and provide no information about the effect of unions on producticity in those sectors o f the economy in which unions have been conjectured to have the most adverse influence. 9 Obviously, more research on the influence o f unions on pro'See, for example, Charles Brown and James Medoff (1978), Richard Freeman (1976; 1977) (forthcoming), James Medoff (1979), and Richard Freeman and James Medoff (1979). ~For a discussion of unions' role as an institution of collectivevoice, see Freeman and Medoff (1979) and Albert Hirschman (1973). For a discussion of internal labor markets and the roles unions play, see Peter Doeringer and Michael Piore (1971) and Oliver Williamson, Michael Wachter, and Jeffrey Harris (1975). ~See Freeman (1977) and Medoff (1979) for evidence that, ceterisparibus, quit rates are lower for unionized workers. 7Doeringer and Piore (1971). ~Brown and Medoff (1978). These estimated productivity differentials are roughly equal in magnitude to estimated union/nonunion earnings differentials that previous investigators have found. This provides an answer to the question of how it is possible for both high wage (union) and low wage (nonunion) firms employingthe same "measured quality" of labor to coexist in the same industry*. 9Additional evidence on positive union productivity effects in the cement industry is presented in Kim Clark (1980). Lest one conclude that unions always increase productivity, the evidence available for coal mining in the U.S. suggeststhe opposite. See Marie Connerton, Richard Freeman and James Medoff (1979).
410
JOURNAL OF LABOR RESEARCH
ductivity is needed. However, by providing tentative evidence on the "social productivity" of unions in manufacturing, Brown and M e d o f f have weakened the case against labor law reform. Let me return now to the issues that Lentz does consider. At the theoretical level, his focus on the necessity o f viewing the decision to join a union in a lifecycle context is an important contribution that leads to implications concerning the relationship between an individual's age and sex and his or her probability of voting for a union. These implications derive at least partially from the fact that the shape of age-earnings profiles differ between unionized and nonunionized employees. This point had been documented by at least two previous sets of investigators, however both failed to draw the implications for union membership which Lentz does.I° Lentz's focus on comparing the present value o f net economic benefits of union and nonunion employment is probably sufficient if one seeks only to provide an explanation for why the probability of union membership varies with age and sex. However, if one's ultimate goal is to build a model which can be used to predict the determinants of union membership over time or across areas, or the outcomes of particular NLRB elections, it seems clear that a more general framework is required. Allow me to suggest three extensions which I believe are needed. First, his focus on net economic benefits ignores the possibility that noneconomic conditions of work may vary between union and nonunion employment. For example, Greg Duncan and Frank Stafford (forthcoming) have recently shown that for blue-collar workers, unionized firms tend to have more structured work settings, less flexible hours, faster work paces, and less employee control over the assignment o f overtime hours, than do nonunion firms. This may arise because production settings which give rise to interdependence among workers and the demand for specific work requirements by employers, also give rise to unions. That is, the decision to vote for a union may be heavily influenced by these nonpecuniary conditions of employment. While after the fact it is possible to attempt to estimate whether a fraction o f any observed union/nonunion earnings differentials merely compensates union members for their " p o o r e r " working conditions and to place a monetary value on these job characteristics, such estimates tend to be imprecise. This suggests that rather than attempting to obtain dollar values for these noneconomic conditions of employment, as Lentz attempts to do for grievance procedures, his model should be expanded to allow these job characteristics to be entered as separate arguments in individuals' utility functions. '1 Second, Lentz's model fails to distinguish between the net economic benefits of union employment for existing union employees and the net economic benefits to individuals contemplating joining a union. Since there is often a
~aSeeFarrell Bloch and Mark Kuskin (1978) and George Johnson and Kenwood Youman (1971). 'Duncan and Stafford estimate that 2/5 of the observed union/nonunion earnings differentials in their sample are compensating wage differentials for nonpecuniary job characteristics.
RONALDG. EHRENBERG
411
queue for jobs in the union sector and some unions (e.g., craft unions) derive their strength through explicit restrictions on entry, the net economic benefits will be less for potential new entrants to the union sector than it will be for otherwise identical existing union members. Any complete model of union membership and growth must take account o f this problem; not all employees who desire jobs in the union sector will be able to obtain them a n d / o r union membership. ~2 It should be stressed that such an extension of Lentz's model is not required to analyze N L R B election data. It is required, however, to analyze the growth o f union membership in the aggregate. In particular, since union entry restrictions decline as employment increases, ceteris paribus, union membership, as a fraction of the labor force, should increase during expansions. This provides an alternative explanation for the " p a r a d o x " that Lentz observed (p. 4) - - the extent of unionism growing between 1939-41 and 1945-49 at the same time that the u n i o n / n o n u n i o n wage differential was falling. ~ Third, while Lentz is correct in asserting that any study o f the decision to join a union must consider the level of union dues, initiation fees and assessments, his model does not allow for the possibility that these variables are endogenously determined. ~' In particular, one might suspect that the level of dues which a union leadership can set is at least partially related to the present values of the net economic and noneconomic gains that individuals obtain f r o m union membership. More generally, this suggests that any model of union m e m bership should focus on the supply o f union services in addition, as Lentz does, to the demand for union services. ~ W h a t is required is a behavioral model o f union leaders which, in conjunction with the d e m a n d side o f the problem, would lead ultimately to a system of equations in which both union m e m b e r s h i p and dues levels would be simultaneously determined. Let me not turn briefly to the empirical portion of Lentz's paper. A substantial part deals with his attempts to ascertain whether the age-sex c o m p o sition of the workforce influences the p r o p o r t i o n o f the workforce that is unionized, as his model predicts. These attempts are at best suggestive. Some (Table 1) are based solely on simple correlations. Others (Tables 2 and 3) are based on regression analysis, however these are fraught with simultaneity problems. For example, in Table 2 the fraction of workers unionized in a state is specified to be a function of the age-sex distribution of employees in a state, while in Table 3 the direction o f causation is reversed. Yet, the equations underlying b o t h tables are estimated by ordinary least squares. Similarly, in Table 3, the age-sex
'2Abowd and Farber have analyzed this problem. ~The growth in unionization during this period is also explained quite well by the model presented in Orley Ashenfelter and John Pencavel (1969). ~'John Pencavel (1971) has presented evidence, from aggregate British time-series data, that the long-run elasticity of the percent of the labor force unionized with respect to the level of union dues lies in the range of - .36 to - .46. ~I have made this point previously in Ronald Ehrenberg and Steven Goldberg (1977).
412
JOURNAL OF LABOR RESEARCH
distribution of employment in a state is specified to be a function o f average hourly earnings in the state and the state's unemployment rate. However, both o f the latter variables are themselves influenced by the age-sex distribution o f employment. These problems suggest that Lentz needs to substantially reanalyze his data before anyone will find his results convincing. In addition, he should attempt to provide an explanation for why other studies have not uniformly supported his hypotheses. For example, Farber and Saks f o u n d that the probability an individual voted for a union in an NLRB election was independent of the individual's sex. Lentz spends almost half the paper discussing ways to value the various nonwage compensation items, including fringe benefits and union security arrangements, that enter, in principle, into his calculation o f the net economic gain from union membership. While I do not always agree with his methods, my comments are minor ones and I have communicated them to him privately. I would like to focus, however, on one rationale for these calculations. From Lentz's paper I infer, although he never explicitly states it, that he believes that the omission of fringe benefits and working conditions from comparisons of union and nonunion workers' earnings has led measured union/nonunion wage differentials to understate the true u n i o n / n o n u n i o n total compensation differentials. Presumably, this would occur because tastes for various fringe benefits differ across individuals and there is no easy way to communicate the preferences o f the average employee to the employer in a nonunion firm. As such, compensation in these firms would be more heavily weighted towards money wages, rather than fringe benefits, than they would in unionized firms. ~6 It is not clear, however, that this argument is correct and that estimated union/nonunion earnings differentials do understate true u n i o n / n o n u n i o n total compensation differentials. As noted earlier, Duncan and Stafford found that several nonpecuniary conditions of employment for blue-collar workers are poorer for union than nonunion employees. Indeed, they calculate that twofifths of the estimated union/nonunion earnings differential reflects compensation for these poorer conditions of employment. In addition, in an unpublished Cornell masters thesis, David Rogers (1978) has shown that several nonwage benefits for police and fire fighters tend to be lower in cities where police and firefighter unions exist than they do in nonunion cities, ceteris paribus." A possible scenario for both of these results is that firms (cities) in which working conditions (public employee benefits) are initially poor are the ones most susceptible, other things equal, to a successful union organization drive. When the time comes to negotiate a contract in such a firm (city) the union strives to improve the working conditions (benefits). However, if employers are resistant to yielding on these issues, the union may maximize the welfare o f its members by settling for larger wage increases, rather than substantial improve"This line of reasoning goes back at least as far as Richard Lester (1967)and some empirical support for it is found in Freeman (forthcoming). "Casey Ichniowski (1980) finds conflicting evidence for firefighters.
R O N A L D G. E H R E N B E R G
413
ments in working conditions (benefits). '~ As a result, we would simultaneously observe positive union/nonunion earnings differentials and negative union/nonunion working conditions (benefits) differentials. Hence, the true union/nonunion total compensation differential may be overstated by the measured earnings differentials, not understated. Let me conclude by noting, as my comments should have indicated, that Lentz apparently was unaware of much of the recent literature on empirical analysis of trade unions at the time he originally wrote his paper. Since much of this research is still unpublished, one should not take this to be a criticism of him. However. I want to stress how exciting and interesting I find this literature to be. The Harvard group (Freeman, Medoff, and associates) is subjecting to empirical quantification long-standing institutional hypotheses, using bestpractice middle-brow econometric techniques, as well as case studies. The Princeton group (Abowd, Farber, Saks, et al.) has stressed building rigorous behavioral models that lead to the specification of sophisticated nonlinear models which must be estimated by maximum likelihood methods. Lentz and other participants at this conference " n e w " to the analysis of trade unions would profit from immersing themselves in this literature before continuing their research in the area.
REFERENCES Abowd, John and Farber, Henry. " A n Analysis of Relative Wages and Union Membership: Econometric Evidence Based on Panel Data" (mimeo, 1977). Ashenfelter, Orley and Pencavel, John H. "American Trade Union Growth, 1900-1960," Quarterly Journal o f Economics (August 1969). Bloch, Farrell and Kuskin, Mark. "Wage Determination in the Union and Nonunion Sectors," Industrial and Labor Relations Review (January 1978). Brown, Charles and Medoff, James. "Trade Unions in the Production Process," Journal o f Political Economy (June 1978). Clark, Kim. "The Impact of Unionization on Productivity: A Case Study," Industrial and Labor Relations Review (July 1980). Connerton, Marie, Freeman, Richard and Medoff, James. "Productivity and Industrial Relations: The Case of U.S. Bituminous Coal" (Mimeo, 1979). Doeringer, Peter and Piore, Michael. Internal Labor Markets and Manpower Analysis. D.C. Heath (1971). Duncan, Greg and Stafford, Frank. "Do Union Members Receive Compensating Wage Differentials," American Economic Review (forthcoming). Ehrenberg, Ronald and Goldberg, Steven. "Officer Performance and Compensation in Local Building Trades Unions," Industrial and Labor Relations Review (January 1977). Farber, Henry and Saks, Daniel. "Why Workers Want Unions: The Role of Relative Wages and Job Characteristics." Princeton University Industrial Relations Section Working Paper No. 112 (May 1978).
'aEmployers may stand firm on working conditions if they view control over them as management's perogative. Similarly, they may prefer to grant wage increases over benefit increase if they view the cost of the latter as being open ended.
414
J O U R N A L OF LABOR R E S E A R C H
Freeman, Richard. "Individual Mobility and Union Voice in the Labor Market," American Economic Review (May 1976). "The Exit-Voice Tradeoff in the Labor Market: Unionism, Job Tenure, Quits, and Separations" (mimeo, 1977). "The Effect of Trade Unions on Fringe Benefits" (forthcoming). Freeman, Richard and Medoff, James. "The Two Faces of Unionism," Public lnterest (Fall 1979). Ichniowski, Casey. "Economic Effects of the Firefighters' Union, "Industrial and Labor Relations Review (January 1980). Johnson, George and Youman, Kenwood. "Union Relative Wage Effects by Age and Education," Industrial and Labor Relations Review (January 1971). Hirschman, Albert. Exit, Voice and Loyalty. Harvard University Press (1973). Lester, Richard. "Benefits as a Preferred Form of Compensation," Southern Economic Journal (April 1967). Medoff, James. "Layoffs and Alternatives Under Trade Unions in United States Manufacturing," American Economic Review (June 1979). Pencavel, John. "The Demand for Union Services: An Exercise," Industrial and Labor Relations Review (January 1971). Rogers, David. "Municipal Government Form, Unions, and Wage and Non-wage Compensation in the Public Sector" (unpublished Cornell University M.S. Thesis, 1978). Williamson, Oliver, Wachter, Michael, and Harris, Jeffrey. "Understanding the Employment Relation: Analysis of lndiosyncratic Exchange," Bell Journal o f Economics (Spring 1975).