Table 7: Exports (Model: IV 2SLSs)
|
Independent variables |
OLS |
SAR |
|
W_EXP |
0.04 * | |
|
Constant |
-2.38 NS |
-5.72 NS |
|
QLA |
1.27 NS |
1.16 NS |
|
QLB |
0.98 NS |
0.89 NS |
|
QLC |
-5.74 * |
-5.61 * |
|
E25 |
1.64 *** |
1.55 ** |
|
POP |
0.24 *** |
0.24 *** |
|
ESGT |
-0.02 NS |
-0.02 NS |
|
NRM |
-0.57 NS |
2.41 NS |
|
BCD |
62.61 *** |
59.79 *** |
|
BI |
6.41 NS |
5.88 NS |
|
BCND |
2.27 NS |
2.34 NS |
|
CTSPM |
0.00 NS |
-0.67 NS |
|
CTCAPM |
0.00 NS |
0.04 NS |
|
R_POS1 |
793.21 *** |
791.53 *** |
|
R NEGl_____________ |
________-340.50 *** |
-336.29 *** |
|
R2aj. / R2buse |
0.49 |
0.49 |
|
Jarque-Bera |
2499383055 *** | |
|
Koenker-Basset____________ |
_________166.96 *** | |
|
Specification Tests | ||
|
Moran |
1.63 *** | |
|
LM (erro) |
2.28 NS |
0.11 NS |
|
LM robusto (erro) |
0.09 NS | |
|
LM (lag) |
3.45 ** | |
|
LM robusto (lag)___________ |
__________1.25 NS |
*significant, 10%; **significant, 5%; ***significant, 1%
The specification tests have shown that the spatial lag model is the most appropriate13. As
anticipated, the location quotient variable (QL) indicates that industrial exports are negatively
correlated with type C agglomerations. By definition, type C are not exporting companies and therefore
their presence is small in environments where exports are higher. From the viewpoint of industrial and
regional policies, this is an important aspect, since large municipality-based concentrations of type C
companies do not share the same economic spaces with agglomerations of exporting companies into
which all type A and most type B companies fit. Such spatial “segregation” limits the spillover effects,
captured by the statistically significant spatial lag variable, which could help in the competitive
catching-up of type C companies.
13 Variables R_POS1 and R_NEG1 are dummies built from OLS residuals, intended to capture outliers that might affect
model estimations.
21