Considering Jæger’s results, I tested for curvilinearity
by including GOVERNMENT OUTLAY SQUARED in
my multivariate models. No such effect was found
when controlling for level 1 factors. Note also that
the between-country variation is fairly high, which in
itself justifies a multilevel analysis approach. This was
confirmed by calculating how much of the variance
is accounted for by the country-level residual, which
ranged from 5.6 to 13.0 percent in the six models.