When international mobility is prohibitively costly and therefore the extent of globalization
low, the tax rate can be set at t* = 1. The government would then choose an allocation of
дн and gL such that the relative share of дн is given by дн/дь = χ[A]1/^(ан/аь). Thus, the
market income of the high-ability individuals will be relatively large. After redistribution,
however, the income distribution would be reversed. In fact, the high-ability individuals will
be completely expropriated.
This conclusion is of course unrealistic if it is taken literally. It does, however, mirror a
particular pattern of scal policy that can be observed in several OECD and many developing
countries. In the typical European welfare state, for example, higher education is provided
either free of charge or is highly subsidized by the government, even though the majority of
the electorate does not directly benet from these expenditures. At the same time, average
and marginal personal income tax rates are relatively high. At rst sight, this pattern of scal
policy makes little sense. According to our model, however, this policy might maximize the
consumption possibilities of the income strata that have a low level of ability if tax competition
is very low.
As argued by Poutvaara and Kanniainen (2000), such a social consensus to nance higher
education publicly may break down once the highly skilled can emigrate more easily. This can
also be shown formally in our model by dierentiating the rst order condition in equation 8
with respect to the level of globalization:
dA ( дн )=λ × ( a *m+f dχr ), (9)
d-∕ι. gl d-∕ι. UïT.
with Λ = (1/0) (t*[A]χ[A])(1→)/^ (ан/аь).
Since Λ is unambiguously positive, the sign of expression 9 depends upon (dt* [A]∕dA)χ[A])+
t* (dχ[A]/dA). First, note that we have shown in section 2.2 that dt*∕dA < 0 because global-
ization intensifies tax competition. The sign of dχ[A]∕dA, however, is less obvious. In fact,
as argued above, it diers between industrialized and developing countries.
In developing countries, globalization decreases the returns to high-skilled and increases
that to low-skilled labor. This implies that dχ∕dA < 0. Therefore, since both dt*∕dA < 0
and dχ∕dA < 0, expression 9 will be unambiguously negative in developing countries. In other
words, developing countries will shift their educational policies in favor of lower education with
deepening globalization.
In industrialized countries, on the other hand, globalization increases the returns to high
skilled and reduces that to low-skilled labor. This implies that dχ∕dA > 0. Therefore, the
sign of expression 9 is indeterminate because dt*/dA < 0. In other words, governments in
industrialized countries face conictive incentives with respect to their educational policies.
Whether globalization will induce them to shift their educational programs toward lower or
higher education, or not at all, is therefore an empirical question. Given that industrialized