primary and reduce that for tertiary education with deepening globalization. The testable
implication of this theoretical result is that the marginal eect of globalization in developing
countries, δ + 7, should be significantly positive for primary and significantly negative for
tertiary education.
Even though the model does not oer a clear prediction for secondary education expendi-
tures, we also consider the impact of globalization on this type of educational program. With
regard to this type of education expenditures, we presume that the eect of globalization
will be less pronounced than for primary and tertiary education in both industrialized and
developing countries.
Our preferred measure for public education expenditures for the three types of educational
programs, i. e. the dependent variables in equation 10, are the expenditures per student,
normalized through division by GDP per capita.
We prefer these measures to using “budget shares” (i. e., primary, secondary, and tertiary
education expenditures as share of total education expenditures) because we must scale gov-
ernment spending on the three dierent educational programs by the number of students in
these programs. This is done automatically when we use expenditures per student as depen-
dent variable. We conduct, however, further below a sensitivity analysis by using the share
of primary, secondary, and tertiary education expenditures in total education expenditures as
dependent variable. A concise description of the dependent variables and their sources can
be found in table 1.
With respect to the control variables in equation 10, our measure of globalization is the
KOF-index developed by Dreher (2006). The KOF-index is based on three sub-indexes which
capture the extent of economic, social, and political globalization. These three sub-indexes
are aggregated into one single measure of globalization. We prefer this measure to the simple
trade openness variable (exports plus imports divided by GDP) because it takes into account
that globalization is a multi-dimensional phenomenon that has broader implications than only
to increase international trade. In view of the research questions in this study, an appropriate
measure of globalization should, inter alia, additionally capture to what extent English is
used as a second language since the ability to communicate in the current Lingua Franca will
directly impact international mobility.
We use a diverse set of additional control variables, which are are listed in table 1. Most
importantly, we include total education expenditures per student as percent of GDP per
capita. This variable has to be controlled for since our aim is to investigate how the relative
importance of primary, secondary, tertiary education expenditures (i. e., the composition of
education expenditures) has changed with globalization.
We additionally include (i) the size of the population in order to capture economies of scale
in the provision of public goods; (ii) total government expenditures as share of GDP in order
to control for the economic importance of the public sector; (iii) GDP per capita in order to