The Effect of Unions on Earnings and Earnings on Unions: A Mixed Logit Approach

A LARGE NUMBER OF STUDIES have attempted to quantify the impact of unions on the wages or earnings of workers. See, for example, Lewis 9 or Ashenfelter and Johnson 1 for critical summaries of many of these studies. Several different research methodologies have been employed to ascertain the extent to which unions have raised relative wages. Empirically, time series and cross-sectional data at the firm, industry and economy-wide level have been examined. Theoretically, most of the analyses have been partial equilibrium in nature, although the recent papers by Johnson and Mieszkowski 7 and Diewert 4 investigate the impact of unionism in a general equilibrium setting. Virtually all of these studies have found a positive, significant effect of unions on wages, although there is considerable variation in the estimated size of the effect. Common to all of these studies is the assumption that unionism exerts a unilateral and exogenous effect on wages. Unfortunately, the interesting issue of the determinants of union membership has been relatively unresearched, and by and large the matter of membership remains unrelated to the effects of unionism on wages. This is at least potentially a serious matter since it seems clear that there may be an effect of wages on unionism as well as an effect of unionism on wages. This may occur because relative wages affect the attractiveness of various industries to a potential union organizer, or because they may affect the probability of a worker voting for a union in a representation election. That relative wages may affect the probability or extent of unionization has been previously noted-see, for example, Reder 12 and Wachter 20. From a statistical point of view, this would imply that union membership, or extent of unionization, would more properly be viewed as jointly or simultaneously determined with wages, rather than being treated as exogenous. Nevertheless, most studies of the effects of unions on wages or earnings have treated unionization as exogenous. A notable exception in this regard is the recent work of Ashenfelter and Johnson 1, which examined at the industry level the effect of unionization on wages and the effect of wages on unionization. In this paper we perform a similar analysis, except using individual observations. The two endogenous variables