In this paper, we compile a unique historical dataset that records strike activity in the British engineering industry from 1920 to 1970. These data have the advantage of containing a fairly homogenous set of companies and workers, covering a long period with varying labour market conditions, including information that enables the addition of union and company fixed effects, and providing geographical detail that allows a district-level analysis that controls for year and seasonal effects. We study the cyclicality of strike durations, strike incidence, and strike outcomes and distinguish between pay and non-pay strikes. Like the previous literature, we find evidence that strikes over pay have countercyclical durations. However, in the post-war period, the magnitude of this effect is much reduced when union and firm fixed effects are included. These findings suggest that it is important when studying strike durations to take account of differences in the composition of companies and unions that are involved in strikes at different points of the business cycle. We also find that strike outcomes tend to be more favourable to unions when the national unemployment rate is lower.