This paper sheds light on the importance of aggregation bias in the analysis of wage shares developments over time and across countries. We focus on five European countries and the United States and show that the trend decline in the aggregate wage share observed in these countries over much of the 1980s and 1990s partly reflects changes in the sectoral composition of the economy. The application of a fixed-weight aggregation method changes the profile of the observed wage share in a significant way: in particular there is no longer sign of an overshooting of the wage share levels of the early-1970s. Error-correction wage equations based on the adjusted wage shares generally have a better regression fit and show long-run elasticities of real wages to unemployment that vary less across countries and are substantially lower than those obtained with observed shares. These results are broadly confirmed by wage regressions using sectoral data and the Pooled Mean Group estimator ...