A better way to display these correlations is through simple least-square re-
gressions on the pooled data. Throughout, all the variables are in logs, except for
dummies; further, to alleviate simultaneity concerns, all the right-hand side variables
are lagged Hve years. Column (1) of Table 3 reports the results of regressing real
output per worker (GDPW) on patent rights (IPR) the Sachs and Warner openness
index (OPEN), an interaction term between IPR and OPEN, an interaction term
between IPR and country size (POP) and country size itself (POP). The regression
also controls for the two important determinants of productivity, physical (KL) and
human (HL) capital per worker. According to (26) the two interaction terms should
have a positive sign. Consistently, column (1) shows that the coefficient on both
interactions is positive and precisely estimated.
Table 3: Panel Analysis
OLS(1) |
OLS(2) |
LSDV(3) |
LSDV(4) |
LSDV(5) | |
IPR |
-1.941 |
-5.723 (1.568)*** |
-0.407 (0.875) |
-0.464 |
-0.904 (0.488)* |
OPEN |
-0.437 |
-0.719 |
0.041 (0.098) |
0.038 |
0.153 |
IPR*OPEN |
0.801 |
0.556 |
0.216 |
0.219 |
0.385 |
IPR*POP |
0.163 |
0.393 |
-0.005 (0.074) |
- |
- |
POP |
0.207 |
-0.452 |
-0.013 (0.113) | ||
KL |
0.400 |
0.453 |
0.323 |
0.321 | |
HL |
0.164 |
0.214 |
-0.037 |
-0.042 (0.024)* |
- |
R2 |
0.83 |
0.82 |
0.58 |
0.58 |
0.39 |
No. of Obs. |
306 |
318 |
318 |
318 |
318 |
F-test[country effects] |
- |
- |
31.02 |
39.06 |
122.44 |
Hausman χ2 |
176.16 |
48.56 (0.000) |
8.63 |
LHS: real GDPW. All variables, except dummies, in logs. RHS variables are lagged (5
yeras). Column 1 uses the Sachs and Warner Openness index. Columns 2-5, use the trade
share in GDP. Standard errors in parenthesis (robust, in OLS regressions). Constant not
reported. *, ** and *** indicate significance at 10%, 5% and 1% level.
25