the OLS case and .21 in the 2SLS case vis-à-vis .10 for their correspondent regressions). The
Sargan tests for over-identifying restrictions cannot reject the hypothesis of validity of instruments
only at the 5% level, with a p-value of .064 for the innovative firms and .72 for the non-innovative
ones.
The impact of mean age of the board members on productivity growth is negative and statistically
significant (γ1,OLS = -0.31, γ1,2SLS = -0.31) for innovative firms and positive and non significant
(γ2,OLS= γ2,2SLS = .06) for the non-innovative firms. Hence the statistical significance stays there and
the point-wise estimates are in the same ballpark (perhaps higher for the innovative firms) as the
OLS results previously shown in Table 4.
Again replicating the same pattern of results as in Table 4, the estimates for the capital share are
almost three times higher for innovative firms than for non-innovative firms. The OLS method
appears slightly overestimating α1,OLS = 0.256 and 0.087, respectively, for innovative and non-
innovative, while α1,2SLS = 0.20 and remains equal to 0.085 for the non innovative when the capital
stock is instrumented with the 2SLS method.
The impact of the share of temporary workers does not also differ significantly across estimation
methods but they differ in significance across categories of firms. The share of temporary workers
affects negatively productivity growth at the second stage for innovative firms (with coefficients of
some negative .11-.12, very similar to those found in Table 4), and it is -0.27 for the non innovative
firms. This result replicates that of Table 4, where the corresponding coefficients were slightly
smaller.
21