When dealing with infrastructure investment the positive externality associated with a
centre is not the only benefit. Therefore the total welfare gain of the positive externality could
be worth calculating as one of the benefits.
When comparing the size of the estimates in the regression the most important
contributions to the average wages are gender, education, some sectors, and age. As
mentioned the fixed effects model is rejected, which means that regional unemployment has a
small but significant estimated parameter. An interpretation could be that higher
unemployment lowers wages because of competition for vacancies.
3.5 Pecuniary externalities: Wage differentials and commuting distance
The second hypothesis of this paper is that commuting distance affects wages. To test this, a
regression analysis on data set 2 is carried out.
We recall that data set 2 has information about both place of residence and place of work.
Both place of residence and place of work could be used as a fixed effect, but when
comparing adjusted R2, place of work is chosen. However, an F-test3 does not support the
municipalities as fixed effects and because of that they are abandoned.
The same problems concerning grouped data are present in this regression as well. The test
described rejects the hypothesis that there is no heteroscedasticity4 in all three models with the
explanatory variables of table A.2 in appendix A. Again, the results of the FGLS regression
are preferred to WLS and OLS because the t-value of β is numerically smaller in the
heteroscedasticity test.
Gender, education, age, year, unemployment by place of residence and commuting
distances are all significant in the model using FGLS, WLS, and OLS. Estimates and standard
deviations are presented in appendix A, table A.2. Comparing the estimated parameter with
the regression using data set 1, the estimated parameters of gender have increased by 17%
(using FGLS) and other changes have also occurred. The estimated parameter of
unemployment is still small, though it has increased. An explanation could be that
unemployment by place of residence is used and in the regression using the first data set
unemployment by place of work is used.
Commuting distance has a positive effect on wages. The estimated parameter is around
0.03 (using FGLS), which means that if commuting distance doubles the average wage would
increase by 3%.
4. The LINE-model - Modelling externalities and transport
Transport system changes have a direct impact on the costs of transportation, either by
reducing transport costs as in the case of transport system improvements or by increasing
transport costs as in the case of taxes on transport activities such as road pricing. Changes in
transport costs have in turn direct effects on commodity prices and income: Transport cost
changes influence directly the prices of commodities, because transport cost is a gross margin
added to commodity trade. Transport costs are an addition to the price when shopping for
commodities or when comsuming as a tourist. In both cases changes in transport costs are
added to the price of the commodity after transportation to the buyer. Changes in transport
costs also have a direct impact on disposable income net of commuting costs.
3 See appendix B, II).
4 See appendix C, II).
11