This paper uses industry price-cost margin data to estimate the extent to which unions reduce profits. Estimates allowing for the endogeneity of union status are contrasted with estimates that assume union status is exogenous and not determined in part by either profitability or industry structure. Endogeneity is found to be an important consideration in estimating the union impact on profits: two-stage estimates are considerably larger than OLS estimates. The final section explores the total estimated redistribution from capital to labor in the manufacturing sector. An important conclusion is that unions raise prices less than was previously believed.

A number of studies have found that unions lower profits, but controversy continues over whether the union impact is or is not greater in more concentrated markets. This empincal controversy is linked to two underlying issues: whether unions distort capital investment decisions, and the total volume of monopoly profits in the U.S. economy. Research on the supermarket industry reveals that unions lower profits substantially in this sector and that indeed their impact is greater when local markets are more concentrated. Recently there has been an explosion of research concerning the union impact on profits and the associated question of whether or not unions reduce profits more in noncompetitive than competitive sectors (for a review of studies, see Voos and Mishel, 1986). This has become an important empirical issue for a variety of reasons. One concern has been whether unions distort capital investment decisions. Insofar as unions merely capture monopoly rents, as opposed to reducing the rate of return in competitive industries, the presumption is that unions create less distortion. A separate issue has been whether or not previous research has underestimated the extent of total monopoly rents in the U.S. economy. This would occur if observed profits are less than total profits in noncompetitive sectors because in those sectors a portion of monopoly rents is being redistributed to unionized employees. While most studies find that unions lower profits, there is some controversy regarding whether or not unions reduce profits more in industries with less competitive market structures. Freeman (1983) and Karier (1985) both find that unions have a much larger effect in concentrated industries. However, Hirsch and Connolly (1985) question that conclusion and Clark (1984) reports contrary evidence with regard to market share. All previous investigations of this issue utilize manufacturing inter-industry data, cross-sections of predominantly-manufacturing firms, or 2-digit SIC manufacturing industry observations disaggregated by state. This paper reports the results of research designed to ascertain whether unions reduce profits in one relatively homogeneous nonmanufacturing industry, supermarkets, and if their impact is larger when local markets are more concentrated. Lamm (1982) and Marion et al. (1979a and 1979b) have explored the impact of unions and of concentration, respectively, on food prices. Moreover, excellent data on supermarket profits exist. The data give our study several unique qualities. Because observations are on firms in specific geographic markets in a single disaggregate industry, any estimated union profit effect cannot be said to result from a correlation between unionism and unobserved industry specific factors. The study is also the first to be based on evidence from the nonmanufacturing sector. The Data Utilized and the Model Estimated The data on which this research was based were gathered under subpoena by the Joint Economic ComReceived for publication July 8, 1985. Revision accepted for publication October 17, 1985. * University of Wisconsin and Industrial Union Department, AFL-CIO, respectively. Special thanks to Frederick Geithman, for many helpful suggestions and for research assistance, and to Bruce Marion for making the data available to us and for commenting on our results. All opinions, as well as errors, are the authors' own. Funds for a portion of this research came from the University of Wisconsin Graduate School. During a portion of this research, Mishel was affiliated with Cornell University.

The 1980s have witnessed considerable experimentation in compensation practices, worker participation, labor-management cooperation, and the organization of production. Collectively, such practices at the local or plant level have become known as One concern that has been raised repeatedly both within the academic community and within the labor movement, is how, and in fact whether, such innovations can coexist with unions and collective bargaining as currently constituted (see for example, Charles Heckscher, 1988; Thomas Kochan, Harry Katz, and Robert McKersie, 1986). Profit sharing, gainsharing, and other compensation changes pose certain issues for unions-how to define profits or gains to protect workers and minimize management finagling, for instance-but these probably require mainly increased access to financial information and the technical sophistication to deal with that information. Labor-management cooperation programs that involve union leaders or activists in joint problem solving with management could create the perception that the union is cozying up with management and is unwilling to push worker demands or grievances. But as with profit sharing and gainsharing, the union sector has long experience with cooperation committees and can presumably make the necessary political adaptations. Innovations that establish a second channel of communication between workers and managers are more problematic according to some observers insofar as they possibly could lead to less worker interest in unionism. Some could preempt the union's expansion into new areas of collective bargaining or might threaten to supplant bargaining over traditional issues. And they might exacerbate existing divisions within the workforce or the union (Donald Wells, 1987). Further issues arise with the more extensive innovations like production that combine the radical reduction of separate production classifications, team decision making in work or overtime assignments, and pay by knowledge. Experience in some auto plants indicates teams can lead to fundamental changes in the operation of the grievance system and the union's day to day role in the plant (Katz, 1985). Indeed it has been argued that team production systems require a modification of the system of job unionism- management manages and the union grieves (Kochan et al., p. 161)-which has predominated in the United States since World War II. The concerns discussed above do not mean that these programs offer no benefits to workers and unions. Benefits such as increased security, enhanced satisfaction, reduced alienation, and greater control over the work environment may result from many of these programs. In fact, it is the differential weighting of these potential benefits and costs that has in part led to the variety of union responses to workplace innovations.