itively correlated with the plants’ share of skilled workers. However, the correlation disappears
when plant-specific fixed effects are included in the estimating equation. Hence, the evidence re-
ported by Pavcnik does not allow to conclude that plants investing in new technology become
more skill-intensive over time. However, the strong evidence reported in the paper concerning the
skill-biased nature of investment in new technology still helps explain within-industry skill upgrad-
ing, as firms investing in new technology, thanks to their investment, expand their employment
share relative to non-investing firms. In other words, it is the reallocation of resources towards
more skill-intensive plants, rather than plants’ skill upgrading, which might explain why industries
become more skill-intensive over time.
5.1.2 The role of exporters
While Pavcnik (2000) focuses on the role of imported technology, a related paper by Bernard
and Jensen (1997) centers instead on the role of exporting plants for skill upgrading and wage
inequality.21 Contrary to previous studies on the determinants of skill upgrading in manufacturing,
that analyze within-industry and between-industry shifts in employment using fairly aggregated
data, Bernard and Jensen look at the contribution of individual (exporting and non-exporting)
plants to the aggregate increase in the relative demand for skilled labor. In particular, they ask
whether skill upgrading and the rise in the skill premia stem from within-plants increases in the
relative demand for skilled labor or from a reallocation of resources toward the more skill-intensive
plants. The question is of particular interest since it can shed light on the relative contribution of
trade and technology to the increased demand for skilled labor. More precisely, within-plants skill
upgrading can be mostly attributed to skill-biased technical change, i.e., to changes in production
practices (such as the widespread introduction of computers and related technologies) that have
increased the relative demand for more educated workers. On the other hand, between-plants
employment shifts can be mostly attributed to cross-plants demand shifts, and in particular to
21The data used by Bernard and Jensen come from U.S. manufacturing plants and cover the period 1973-1987.
Sample plants account for almost two thirds of total manufacturing employment in the U.S..
36