What precisely is the impact of unionization on wages? The estimated union/non-union wage differential appears to fluctuate widely across various empirical studies. For example, based on a time-series estimation of a macro-model, Minford (1983) arrives at a figure of 70 per cent for the union mark-up. Studies based on industry cross-sections, however, appear to converge on 25 per cent for the United Kingdom. (See for example, Mulvey, 1976; Metcalf, 1977; Parsley, 1980. An exception is Mulvey and Abowd, 1980, who, after bias correction, arrive at about 6 per cent.) But many think that such cross-section studies are unreliable (e.g. Nickell, 1982) and misleading (e.g. Geroski and Stewart, 1982). In fact, concluding from his extensive survey, Lewis (1983) argues that, in general, industry cross-sections measure more than the union/non-union wage differential and therefore report an upward biased mark-up.
This paper has presented results on the union/non-union wage differential using individual-level cross-section data. It appears that, once account is taken of working characteristics such as job demands and job severity, the wage differential becomes smaller. The mark-up for 1968-1969 is slightly higher
than the figure of 8 per cent for 1975 arrived at by Stewart (1983). But it is still pretty close to his estimate, much lower than that of 25 per cent associated with most British industry cross-section studies, and bears no comparision to the time-series estimate of 70 per cent obtained by Minford (1983). However, it is worth bearing in mind that the sample size we have utilized is relatively small. Our results on job security variables and job benefit variables also suggest that it might be worthwhile concerning ourselves with an explanation of the "enlarged wage differential" instead of the conventional "pure wage differential". This should be possible with the availability of more information, which would enable us to impute monetary values to all... [ABSTRACT FROM AUTHOR]