trade-induced demand shifts.
The decomposition performed by Bernard and Jensen reveals that between-plants shifts explain
46% of the total increase in the relative demand for skilled labor and 58% of the total increase in
the skill premium. These results stand in sharp contrast to previous studies carried at the industry
level, where the within-industry component explain virtually all of the increase in the relative
demand for skilled labor (Katz and Murphy, 1992; Berman, Bound and Griliches, 1994), and hence
suggest that decompositions based on industry-level data hide substantial within-industry plant
heterogeneity and potentially overestimate the importance of skill-biased technical change.
In order to determine the role of the export status of firms for the dynamics of relative wages
and employment, Bernard and Jensen look at the contribution of exporting plants to the within
and between increases. They find that non-exporters have a within effect on relative employment
21% larger than exporters, whereas the between effect on employment is entirely explained by the
exporters. As for wages, the role of exporters is even stronger. The within effect on wages is 26%
larger for exporters than non-exporters, whereas the between effect on wages is entirely explained
by the exporters. These results suggest that the rise in wage inequality is due to employment
gains at exporting plants, even though skill upgrading is taking place at both exporters and non-
exporters.
Finally, in order to test more directly the role of technology and product demand shifts for labor
market dynamics, Bernard and Jensen regress the within and between components of relative wage
and employment increases on changes in export sales, domestic sales and technology variables, such
as the change in the R&D to sales ratio or in computer investment. The main results are that
between-plants changes (both in wages and employment) are strongly positively related to increases
in both foreign and domestic demand, with the coefficient of the former three times larger than
the latter. The impact of technology measures on the between components is instead weaker.
These results suggest that the between-plants movements of workers and wages, which are crucial
for the increase in the skill premium, are largely determined by demand shifts across plants and
37