This paper reports individual-level estimates of union/non-union wage differentials, using coverage information from the New Earnings Survey Panel Dataset. There are no existing panel estimates for the United Kingdom. Of a number of findings, the more important are (i) fixed-effects estimates are about one-half the equivalent cross-section estimates; (ii) the biggest differentials are for 'company/district/local only' agreements; and (iii) the differential is counter-cyclical. Also the effect of the 1979/82 recession was probably dampened by anti-union legislation, and the upward trend between 1975 and 1995 is due to the decentralization of collective pay bargaining. © Blackwell Publishers 1998.