(quadratic). The F-statistics for testing the joint significance of the individual effects are
given under the F-value column of Table 2. The results strongly suggest the presence of
an individual heterogeneity in the data. The values associated with upper turning points in
the cubic function are slightly higher for P and slightly lower for the N and DO
pollutants.
Table 3 shows the two-way fixed effects model. Notice that in some cases, the
parameter estimates produced by the two-way model are higher compared to the one-way
model; in other cases, however, these numbers are smaller. We also found that almost all
the coefficients of all SC variables are significant in the N equation. The F-statistics
indicated the presence of both individual and time specific effects. The turning points
produced by the two-way method are higher than those produced by one-way model,
especially for N. The EKC curves associated with the cubic functional form of all these
pollutants for both one way and two way fixed effect models are shown in Figure 3.
The regression results for the random effects model are given in Tables 4 and 5.
The Hausman statistics reported in Tables 4 and 5 are lower than the critical values from
a chi-squared table, except for N of the one-way random effects model and for N and DO
in two-way random effect models. Thus, the hypothesis that the individual effects are
uncorrelated with the other regressors in the model cannot be rejected.
As shown for the nitrogen equation in Table 4, the coefficients for social capital
in both quadratic and cubic forms are statistically significant at the 1% level for N. These
empirical results provide evidence of an U-curve relationship between social capital and
nitrogen level. Using the quadratic specification, we obtained a turning point of 0.03.
Although the coefficients on population density associated with all pollutants possessed
15