Considerable attention has been focused in recent years on the empirical analysis of the effect of trade unions on the structure of earnings. Such empirical investigation usually entails measurement of the union/non-union differential in wages or earnings. At the level of the individual this involves comparison of the earnings of union members with estimates of what their earnings would be were they not union members, or, equivalently, examination of differences in earnings between individuals comparable with regard to all other relevant characteristics who differ in their union membership alone. In the United Kingdom all previous attempts at estimation of this differential have used aggregate data at either the industry or occupation level.1 The conventional methodology involves regression of the logarithm of the average wage on either the proportion of workers who are union members or the proportion covered by a collective agreement, together with a vector of industry characteristics that are deemed to determine the non-union wage. However, several problems are immediately apparent with this methodology. First, the model ignores the possibility of variation in the differential across industries, which has potentially serious consequences. Second, the vector of other characteristics customarily used is somewhat limited and raises doubts concerning the adequacy of the standardization for what might loosely be termed labour quality. Third, only a sub-sample of industries is ever used owing to lack of data. Those industries excluded are generally smaller (in terms of employment) that those included. If the differential varies with the size of the industry a further bias will be induced. All of these problems are either removed or lessened by the use of individual-level data. (For an analysis of their impact at the industry level see Geroski and Stewart, 1981).2 As a result, it is now widely accepted that the appropriate way to measure such differentials is to use micro-data sets at the level of the individual worker. Until recently suitable data have not been available for the United Kingdom and hence such analysis was not possible. This paper presents the first estimates for the United Kingdom based on individual-level data. If the value of the union/non-union differential were the same for all workers in the unionized sector, analysis at the industry level and the individual level would produce similar results. However, the Marshallian laws of derived demand suggest that this will not be the case. Despite this clear prediction of variation in the differential, there has been little systematic investigation of the variation. This paper also presents such an analysis. Since variation in the differential across industries results in aggregation bias in industry-level studies, such an analysis is of considerable relevance to the evaluation of the evidence from such studies, as well as being of interest in its own right. The paper is laid out as follows. Section I presents a method of estimating the union/non-union differential at the individual level while Section II
Stored only when both papers are in the corpus — 117 citations exist upstream.