Two stylized facts have attracted a lot of attention recently: i) the inability of real wage growth to track real labour productivity growth (mostly in the US), and ii) the (worldwide) decrease in the labour share.
A simple statistical decomposition of aggregate data shows that the disconnect between pay and productivity growth is equivalent to a change in the labour share, computed from aggregate data.
A first contribution of our paper is to point out that the labour share computed on aggregate data is different from the labour share computed using micro data. We then show that firm heterogeneity in terms of technology, market power, capital and wages plays a significant role in explaining in the first -hence the productivity-pay disconnect- even in the simple case of Cobb-Douglas production functions with homogenous parameters. Moreveor, firm heterogeneity also matters in explaining the latter -the labour share as correctly measured- as long as we depart from the Cobb-Douglas assumption.
This contributes to the large and growing literature exploring the determinants of the observed decline in the labour share, which so far has mainly looked at aggregate factors.