Labour market theories allowing for search frictions make marked predictions on the effect of the degree of frictions on wages. Often, the effect is predicted to be negative. Despite the popularity of these theories, this has never been tested. We perform tests with matched worker-firm data. The worker data are informative on individual wages and labour market transitions, and this allows for estimation of the degree of search frictions. The firm data are informative on labor productivity. The matched data provide the skill composition in different markets. Together this allows us to investigate how the mean difference between labor productivity and wages in a market depends on the degree of frictions and other determinants. We correct for worker self-selection into high-wage jobs. Using within-market variation, we also investigate the extent of (and explanations for) positive assortative matching.