school or learns on the job, so that the net effect of E on INNOVATION is positive or negative
depending on whether a>b.
The intuition for (3) is simple. Traditional managerial techniques require on-the-job experience,
while novel managerial techniques are those accumulated through off-the-job specific training or
(business) schooling. This typically involves a trade-off, for a firm faces the decision to employ
today a relatively experienced but old-fashioned manager or tomorrow a relatively un-experienced
but well trained manager. It might also be that the marginal productivity of managerial experience
and schooling is different across different categories of firms, such as between the so called
innovative and non-innovative ones.
In practice, we empirically implement the logical framework described above in two stages. In the
first stage firms are all alike but they contemplate the choice of innovating or not. As documented in
previous studies, they are more likely to become innovative if they undertake R&D and if endowed
with enough cash-flows, as well as other location, size and industry time-invariant variables
(captured by fixed effects in our empirical analysis). But the firm’s propensity to innovate may also
be affected by experience-related variables, whose role has not been much investigated before.
Firms may innovate more if they do not employ too high a share of temporary workers as well as if
they are endowed with relatively young managers. Once firms have selected themselves into
innovative and non-innovative, experience as well as location, firm and size dummies are also
allowed to affect productivity growth in the second stage, while R&D and cash flow are not. The
exclusion of R&D and cash flows is thus our main identifying assumption.
3. Empirical specification and strategy
We search for the best empirical specification consistent with our data, starting from a baseline
specification with no asymmetry between innovative and non-innovative firms. This baseline
specification is the following: