In this paper we analyse data from the 1980—4 WIRS panel for the light they can shed on the causes of the decline in aggregate trade union density in the UK. We argue that, contrary to conventional wisdom, intra-establishment union density did, on average, decline between 1980 and 1984. This suggests that a traditional compositional change story is untenable. We proceed to investigate the correlation of intra-establishment change with variables suggested by the various hypotheses advanced to explain declining density over the period. We find some evidence for an influence of intra-establishment compositional changes, but the most statistically significant variables are wages and unemployment. These take the wrong sign, however, to support the business cycle interpretation of falling aggregate union density.