Although the pooled OLS regression is a useful way to summarize partial corre-
lations in the data, it may place too much weight on cross-sectional variation and
suffer from omitted variables, particularly given the small number of covariates. In
this respect, a LSDV regression with country fixed-effects has more advantages, as it
controls for omitted variables that change very little over time and that may be cor-
related with other regressors, such as institutional and geographical characteristics
of countries. However, since this estimator uses only within-country variation, the
Sachs and Warner index of openness, with its almost nil time variation, is here inad-
equate. The analysis therefore continues using the trade share in GDP as a measure
of openness. Before moving to the fixed-effects regression, Column (2) shows again
the pooled OLS estimates with the new trade measure and it confirms the previous
findings: the two interaction terms are positive and significant at the 1% level.
Columns (3)-(5) report the results from the LSDV fixed-effects estimator. Col-
umn (3) includes all the right-hand side variables. The interaction term between
patent rights and openness is still positive and significant. On the contrary, the
coefficient on country size is now very small and not statistically different from zero.
This is not very surprising, given that population varies mostly across countries
(Table 1 shows that the cross-sectional standard error of POP is almost three times
its mean). It suggests that only the large cross-sectional variation of country size
may have a significant impact on the effectiveness of IPRs, which is not inconsistent
with the theory. Column (4) reports the estimates after dropping the size variables,
whose contribution to explain changes in productivity over time has been found sta-
tistically small. Finally, Column (5) isolates the effects of patent rights and trade,
the main variables of interest, by dropping all the other covariates. In all cases, the
coefficient on the interaction term between openness and patent rights is consistently
found to be positive and statistically different form zero.28 To conclude, given that
in all the specifications the coefficient on the interaction term is found to be positive
with significance levels always below 4%, there seems to be fairly robust evidence
that patent laws are more correlated with high productivity in open countries.
A few calculations on the coefficients in Table 3 can help to understand the
magnitude of the effects and if the estimates across specifications are comparable.
Consider first the impact of intellectual property protection. For the average country,
Columns 1-3 imply that a 10% increase of the index of patent rights is associated
28Adding a time trend affects the results only marginally and turns out not significant.
26