Public infrastructure capital, scale economies and returns to variety

higher in magnitude (0.630 and statistically significant). The same can be said for the findings for the
regression utilising productive public capital. Its coefficient is again higher than the respective
estimate at regional level (0.583 and statistically significant for the sectoral panel). A crucial difference
from the regional analysis can be found if the results for regional and sectoral social infrastructure
regressions are compared. At the regional level the coefficient for social public capital is not
statistically significant, while when sectors are considered it is both significant and high (0.789) -
seemingly more so than is the case for productive infrastructure. The cost function analysis results
have shown that social infrastructure does have a positive effect in reducing private costs at the
sectoral level. Thus, both the model results and those of the cost analyses suggest that some sectors
are highly affected by social public capital. This effect cannot detected by the regional (prefectural)
panel where manufacturing industry as a whole is examined. If there were available sectoral data at
regional level, this discrepancy (between sectoral and prefectural results) would be probably solved.

Table 9 Infrastructure effects on total output: Greece panel for sectors, 1982-1991

Dependent Variable: ln of Total Manufacturing Output (GPV)___________________________________________________________________________

*** Statistically significant at 1% level, ** Statistically significant at 5% level, * Statistically significant at 10% level

Even more interesting is a comparison of the results for the two regional-sectoral panels -
those for the metropolitan area of Athens panel (table 10) and the Rest of Greece panel (table 11).
The results for Athens, in all three regressions for total, productive and social public capital, generate
an insignificant coefficient for public capital. This is in contrast with all other results for the different
spatial levels. The estimates for labour approximate roughly similar levels (positive and statistically
significant) to the respective results for the regional and other sectoral panels. In contrast, the
coefficients for the manufacturing establishments variable are much higher in magnitude than for any
other panel. In all three regressions for Athens these estimates are around 1.6, which implies that the
number of manufacturing firms generates a greater volume of industrial output. But more important
for this research is the finding that, for Athens, infrastructure in all its guises seems to play no role, as
in all the regressions its estimates are statistically insignificant. One potential explanation involves the
fact that the Athens economy is much more advanced in comparison with that of the rest of Greece.
These results seem to corroborate, at first sight, the argument that infrastructure investment has a
smaller effect in more advanced economies that already endowed with a sufficient infrastructure
capacity (see for instance Holtz-Eakin 1990). Conversely, there is the other possibility that in reality

23

<<   <   20   21   22   23   24   25   26   >   >>