There is by now ample empirical support for the claim that unions do affect labour contracts1. These results indicate that unions do indeed have some bargaining power. In order to develop public policies towards unions, it becomes important, therefore, to understand the ways in which their bargaining power operates. Several theoretical speculations have related bargaining power to both product and labour market structures2. Yet, there has been very little empirical corroboration of the arguments developed3. A major impediment to empirical testing has been the lack of quantitative information on the industrial relations network. In recent years, however, the US Bureau of Labour Statistics has collected large bodies of data on collective bargaining contracts. In this paper, one of these data sets is used4 to generate quantitative measures of the determinants of union power. The paper is organized as follows. In the first section, the major theoretical arguments concerning the determinants of union power are reviewed, and the data used to capture the sources of union power are presented. Section II includes a description of the methodology and of the basic sample used. In Section III, the effect of union power determinants on the wage differential between union and non-union establishments is estimated in a cross-section of a sample of 4073 United States establishments in manufacturing industries alone. The effect of the indicated determinants on both union and non-union wages, not a central concern of this paper, is also reported. In conclusion, some policy implications are briefly indicated.